Tracking Chinese Development Finance:

An Application of AidData’s TUFF 3.0

Methodology

Samantha Custer, Axel Dreher, Thai-Binh Elston, Brooke Escobar, Rory

Fedorochko, Andreas Fuchs, Siddhartha Ghose, Joyce Jiahui Lin,
Ammar A. Malik, Bradley C. Parks, Kyra Solomon, Austin Strange,
Michael J. Tierney, Lydia Vlasto, Katherine Walsh, Fei Wang, Lincoln
Zaleski, and Sheng Zhang

AidData, William & Mary

Acknowledgements
We owe a debt of gratitude to the large team of faculty, professional staff, and research
assistants who have helped refine and implement the 3.0 version of the Tracking
Underreported Financial Flows (TUFF) methodology. The construction of the 3.0 version of
AidData’s Global Chinese Development Finance (GCDF) dataset was only possible because of
their collective efforts over the past two years. We specifically want to thank Abena Konadu,
Abigail Taylor, Aldrin Yashko, Alec Veit, Alejandro Schnapp Mezerhane, Alex Park, Ally Swindell,
Alyson Reynolds, Andrea Phelps, Andy Shufer, Anna Garrett, Annabel Richter, Anne Paxton,
Annika Suri, Arianna Lagamma, Audrey Sun, Bailey Garber, Ben Greenberg, Beyoncé Lightfoot,
Bree Jiang, Caitlin Connor, Caitlin Noe, Camilla Yeleussizova, Carina Rosenberg, Carolin
Helmholz, Caroline Elszy, Changlin (Bob) Zhang, Check King Tay, Christian Moore, Claire
Rudinsky, Claire Schlick, Claire Wyszynski, Clara McAllister, Collin Absher, Connor Stanton,
David Zhu, Dorothy Gao, Dory Gilmer, Elise Tsao, Elizabeth Ott, Emma Bury, Falan Kifle, Gelila
Yonas, George (Teddy) Chunias, Gujie Shen, Gwyneth Smith, Harin Ok, Harrison Abramsohn,
Helen Tieder, Henry Tieder, Holi Raparaoelina, Jack Mackey, Jackson Humphreys, Jason Bao,
Jennifer Luo, Jiaju Du, Jiayi Xu, Jiexi Lin, Joanne Lee, Joey Lindsay, Joseph Bohley, Juhwan
Cho, Kate Hentz, Katelyn Oxer, Kathryn Spurrier, Kathryn Ziccarelli, Kayla Aviles, Kelly Shinners,
Kerim Gales, Kien-Tam Elston, Laina Lomont, Laura Lam, Lexi Olsen, Liang Geng, Lillian Butler,
Maddie Sharpe, Marianna Bonilla, Mary Trotto, Max Berckmueller, Meveryn Chua, Mia
Maccario, Micah Katahara, Michael Olyniec, Mujia He, Nancy Lang, Nate Spiers, Nico Rullo,
Noah Tran, Olivia Schlamp, Peter Awabdeh, Pooja Muthuraj, Pooja Tanjore, Prestin Tran, Redeit
Hailu, Riely Capece, Riley Companion, Rin Tsujikawa, Rodrigo Arias, Ryan McCormich, Sabrina
Levri, Sailor Miao, Sam LeBlanc, Samantha Rofman, Sarah Kennedy, Saranga Bansal, Sean
Nguyen, Shannon McGuire, Sophie Adams-Smith, Sophie Alden, Sophie McGehee, Sophie
Strauss, Sophie Torres, Susannah Matthews, Taylor Perry, Tom Mooney, Victor Gedeck, Will Wu,
Ye Xiao, You Wu , Yufei Tang, Yunlang (Bella) Li, Yuxi (Felix) Tian, Yuxin Qin, Zheyi (Serena) Liu,
Zhiying (Cindy) Zhou, Zipu Liu, Ziqi (Samuel) Yan, and Zupeng Zeng for their careful application
of the 3.0 version of the TUFF methodology and construction of the 3.0 version of the GCDF
dataset. Nicholas Costa, Taylor Liegel, Milo Yen-Goossens, Rushika Devineni, Shiying Cai, and
Eli Svoboda also provided valuable data management software support.

Over the last ten years, a large group of coauthors and collaborators—including Christian
Baehr, Ariel BenYishay, Richard Bluhm, Mengfan Cheng, Carrie Dolan, Seth Goodman,
Sebastian Horn, Alysha Gardner, Anna Gelpern, Roland Hodler, Brook Lautenslager, Jacob Hall,
Alexandra Joosse, Scott Morris, Edwin Muchapondwa, Daniel Nielson, Julius N. Odhiambo,
Paul Raschky, Carmen Reinhart, Daniel Runfola, Christoph Trebesch, Rachel Trichler, and Lukas
Wellner—has helped refine and expand the methodology and dataset by piloting coding
procedures, recommending new sources and methods, and scrutinizing preliminary records.

We gratefully acknowledge financial support from the Ford Foundation, the Smith Richardson
Foundation, William and Flora Hewlett Foundation, and USAID. We also acknowledge that
previous versions of the dataset would not have been possible without generous financial
support from the John D. and Catherine T. MacArthur Foundation, Humanity United, United
Nations University-WIDER, the Academic Research Fund of Singapore’s Ministry of Education,
and the German Research Foundation. AidData’s research is guided by the principles of
independence, integrity, transparency, and rigor. A diverse group of funders support AidData’s
work, but they do not determine its research findings or recommendations.

2
Citation
Custer, S., Dreher, A., Elston, T.B., Escobar, B., Fedorochko, R., Fuchs, A., Ghose, S., Lin, J.,
Malik, A., Parks, B.C., Solomon, K., Strange, A., Tierney, M.J., Vlasto, L., Walsh, K., Wang, F.,
Zaleski, L., and Zhang, S. 2023. Tracking Chinese Development Finance: An Application of
AidData’s TUFF 3.0 Methodology. Williamsburg, VA: AidData.

3
Table of Contents
Introduction 6

Section 1 - AidData’s Global Chinese Development Finance Dataset, Version 3.0 11

1.1 - Scope Parameters of the 3.0 Dataset 11
1.2 - Key Features of the Dataset 13
1.3 - Guide to Using the Dataset 16
1.4 - Field Definitions 19

Section 2 - Capturing Chinese ODA and OOF 52

2.1 - Measuring Concessionality and Intent 53
2.2 - Measuring Emergency Rescue Loans and the Cumulative Stock of Official Financial
Flows from China to LICs and MICs 55
2.3 - Categorizing Chinese Lending to Different Types of Borrowers 56
2.4 - Identifying when China’s Borrowers are Experiencing Financial Distress 58
2.5 - Chinese ODA and OOF Agencies and Instruments 59
2.5.1 - Sources of Chinese ODA and OOF 59
2.5.1.1 - Overview 59
2.5.1.2 - China’s Ministry of Commerce (MOFCOM) 60
2.5.1.3 - Export-Import Bank of China 62
2.5.1.4 - China Development Bank (CDB) 66
2.5.1.5 - State-Owned Commercial Banks 67
2.5.1.6 - People’s Bank of China (PBOC) 68
2.5.2 - Recipients of Chinese ODA and OOF 69
2.5.3 - Chinese ODA and OOF Financing Mechanisms 70
2.5.3.1 - AidData’s “Flow Type” Categorization 70
2.5.3.2 - Financing Agreements 71
2.5.3.3 - Loan Flow Type 73

Section 3 - TUFF 3.0 Data Collection Process 93

3.1 - Stage 1: Identifying New Projects and Sources 93
3.1.1 - Official Country Profiles 94
3.1.2 - DNA/Factiva Articles 95
3.2 - Stage 2: Record Enhancement and Verification 97
3.2.1 - De-duplication and Detailed Searches 97
3.2.2 - Dataset Variables and Structure 99
3.3 - Stage 3a: Project-Level Quality Assurance 117
3.4 - Stage 3b: Dataset-Level Quality Assurance 119

4
Section 4 - Geospatial Data Collection Process 120
4.1 - Stage 1: Geospatial Data Collection and Precision Level Labeling 121
4.2 - Stage 2: Geospatial Data Enhancement and Quality Assurance 122
4.3 - Stage 3: Geospatial Data Cleaning and Dataset Generation 122

Appendices 124
Appendix A: Classification of Official Finance 124
Appendix B: Geographic Coverage of the 3.0 Dataset 125
Appendix C: TUFF Source Prioritization Protocol 133
Appendix D: AidData’s Deflation Methodology 133
Appendix E: Health of Record Scores 136
Appendix F: AidData’s Global Chinese Development Finance Dataset, Version 3.0 at a
glance 138

References 140

5
Introduction
To help those who seek to understand the nature, distribution, and effects of development
finance from emerging donors and creditors, AidData developed the Tracking Underreported
Financial Flows (TUFF) in collaboration with an international network of researchers from
Harvard University, Heidelberg University, the University of Göttingen, the University of Cape
Town, Brigham Young University, and William & Mary. The methodology codifies a systematic,
transparent, and replicable set of procedures that facilitate the collection of information about
aid and credit from official sector donors and lenders who do not publish comprehensive or
detailed information about their overseas activities. It does so by synthesizing and
standardizing vast amounts of unstructured, open-source, project-level information published
by governments, intergovernmental organizations, companies, nongovernmental organizations,
journalists, and research institutions.

The methodology was first introduced in April 2013 as a way of tracking Chinese
government-financed development projects in Africa (Strange et al. 2013). It was then revised
and extended to track Chinese government-financed development projects in Africa, Asia,
Latin America and the Caribbean, the Middle East, Oceania, and Eastern and Central Europe in
September 2015, January 2017, and October 2017 (Muchapondwa et al. 2016; BenYishay et al.
2016; Strange et al. 2017; Bluhm et al. 2018; Dreher et al. 2018, 2019, 2021, 2022). These
revisions were chronicled in a book entitled Banking on Beijing: The Aims and Impacts of
China’s Overseas Development Program (Dreher et al. 2022). AidData then re-engineered the
TUFF methodology to support the creation of AidData’s Global Chinese Development Finance
(GCDF) Dataset, Version 2.0, which was published in September 2021.1 This retooling of the
methodology involved (a) increased reliance on official sources, (b) the collection of more
detailed information on the terms and conditions of the financing agreements issued by
Chinese state-owned entities, and (c) a stronger focus on project implementation to improve
the reporting of commencement dates, completion dates, and precise geographical locations.
To build upon these improvements, capture the changing nature of China’s overseas lending
and grant-giving portfolio during the late BRI period (2018-2021), and improve the usability of
the dataset for analysis, we have introduced new variables and structural changes to the latest
(3.0) version of the TUFF methodology.2

We made six major improvements to the 3.0 version of the TUFF methodology to support new
types of analysis.3

1. Improved financial instrument coverage and categorization: The 3.0 dataset captures
10,291 grant-financed projects/activities and 4,776 loan-financed projects/activities that
were formally approved, active, or completed. Given that China relies on an increasingly
diverse set of debt instruments to finance its overseas development program in
low-income countries (LICs) and middle-income countries (MICs), we have introduced a

1
The 1.0, 1.1, 1.2, 1.3, and 2.0 versions of the TUFF methodology are available on the aiddata.org
website. The 3.0 version of the TUFF methodology was used to create the 3.0 version of AidData’s
GCDF dataset. For analysis of key patterns and trends from the 3.0 version of AidData’s GCDF dataset,
we encourage readers to consult Parks et al. (2023).
2
Appendix F provides an overview of the similarities and differences between the 2.0 and 3.0 versions
of AidData’s GCDF Dataset.
3
For detailed descriptions of each field in the dataset, refer to Section 1.4.

6
 new loan categorization scheme that allows users to isolate 23 specific types of loan
instruments: (1) bilateral loans (“Number of Lenders” field), (2) syndicated/club loans
(“Number of Lenders” field), (3) export buyer's credits, (4) supplier’s credits/export
seller’s credits, (5) interest-free loans, (6) loans for debt refinancing purposes
(“Refinancing” field), (7) investment project loans (“Investment” field), (8) loans that
support a merger or acquisition (“M&A” field), (9) working capital loans (“Working
Capital” field), (10) Engineering, Procurement, and Construction Plus Finance
arrangements (“EPCF” field), (11) lease agreements (“Lease” field), (12) foreign currency
swap borrowings or loans for balance of payments (BOP) support (“FXSL/BOP” field),
(13) cross-currency interest-rate swaps (“CC IRS” field), (14) revolving credit facilities
(“RCF” field), (15) government concessional loans (“GCL” field), (16) preferential buyer’s
credits (“PBC” field), (17) pre-export financing or commodity prepayment financing
arrangements (“PxF/Commodity Prepayment” field) (18) inter-bank loans, (19) overseas
project contracting loans, (20) deferred payment agreements (“DPA” field), (21)
non-recourse or limited-recourse project finance transactions (“Project Finance” field),
(22) short-term loans (“Short-term”), and (23) emergency rescue loans (“Rescue” field).
Additionally, the 3.0 dataset flags loans that involve multilateral institutions as loan
administrators, co-financiers, insurers, and/or advisers (“Involving Multilateral” field),
and loans that involve co-financing agencies which are not of Chinese origin (“Involving
Non-Chinese Financier” field).
2. Improved tracking of debt repayment obligations and liabilities: To help users of the 3.0
dataset better understand how Chinese lenders use credit enhancements, we identify
four distinct types of organizations (“Guarantor," “Insurance Provider," “Collateral
Provider," and “Security Agent/Collateral Agent”) that were previously identified in a
single “Accountable Agencies” field in the 2.0 version of the GCDF dataset.
Additionally, for users who wish to analyze on-lending arrangements, the 3.0 dataset
now includes two different categories of receiving agencies (“Direct Receiving
Agencies” and “Indirect Receiving Agencies") and a flag for loans that involve
on-lending arrangements (“On-Lending” field). To better track debt repayment
obligations in cases when the direct receiving agency is a joint venture or special
purpose vehicle (JV/SPV), we have introduced new variables that track the extent of
host government ownership (“Host Government Ownership” field) and Chinese
government ownership (“Chinese Government Ownership” field) of the JV/SPV. To
support analysis of the actual and potential debt repayment obligations of host
governments, we have introduced a “Level of Public Liability” marker that assigns each
loan to one of six categories: Central government debt, Central
government-guaranteed debt, Other public sector debt, Potential public sector debt,
Private debt, or Unallocable.4 We have also included a “Financial Distress” flag that
identifies whether, for a given loan, there is any indication that the borrower had
difficulty repaying the loan or was financially distressed during the loan’s repayment

4
Users of the 3.0 dataset can identify PPG debt by aggregating lending commitments to government
and majority state-owned institutions as well as other institutions that secured central government
repayment guarantees or repayment guarantees from state-owned entities other than the central
government in the host country (i.e., by using the “Level of Public Liability” variable to identify all loan
commitments assigned to the “Central government debt,” “Central government-guaranteed debt,” and
“Other public sector debt" categories. In addition to PPG debt, the “Level of Public Liability” variable
allows users to identify loans to minority state-owned institutions without public sector repayment
guarantees (“Potential public sector debt”).

7
 period (according to the project/transaction life-cycle information that is identified in
the “description” field)
3. Expanded and enhanced spatial granularity: Another important value addition to the
3.0 dataset is the level of geographical detail regarding where projects/activities take
place. As we describe in greater detail in Goodman et al. (2023), for 9,497
projects/activities that have physical footprints or involve specific locations, the 3.0
dataset extracts point, polygon, and line vector data via OpenStreetMap URLs and
provides a corresponding set of GeoJSON files and geographic precision codes.5 72%
(6,919) of these projects/activities include “precise” or “approximate” geocodes; the
remaining 28% (2,578 projects) are measured at an administrative unit level.6 Measuring
the spatio-temporal rollout of project/activity implementation with a high level of
precision is important because it creates new opportunities to identify cause-and-effect
relationships in rigorous ways.7
4. Expanded and enhanced project implementation and temporal coverage: To better
document the full timeline of the project/activity from official financing commitment to
the completion of implementation, we have introduced new variables and conducted a
systematic review of project/activity records from the 2.0 dataset to record any new
details regarding implementation. While prior versions of the GCDF dataset included
only a “Commitment Year” field, the 3.0 dataset now includes a “Commitment Date”
field that records the precise calendar day on which the official financing agreement was
issued. The 3.0 dataset also provides an unprecedented level of detail on
project/activity commencement (implementation start) dates and completion
(implementation end) dates. It identifies precise, calendar day-level commencement
dates for 11,286 projects/activities, and calendar day-level completion dates for 11,542
projects/activities. By way of comparison, the 2.0 version of the GCDF dataset identified
calendar day-level commencement dates for 5,539 projects/activities, and calendar
day-level completion dates for 6,061 projects/activities. The 3.0 dataset provides data
on the originally scheduled project/activity commencement dates and completion
dates, which has paved the way for the introduction of two new measures (“Deviation
from Planned Implementation Start Date” and “Deviation from Planned Completion
Date”) of the degree to whether and to what degree projects/activities ran (or are
running) ahead of schedule or behind schedule. Additionally, to support users who wish
to study loan repayment schedules, AidData has introduced variables that capture the
first (originally scheduled) loan repayment date and the last (originally scheduled) loan
repayment date. These variables are automatically calculated based on the values
captured in the commitment date field, the grace period field, and the maturity field.

5
Users who wish to conduct analysis at higher levels of spatial aggregation can find the ADM1s
(provinces) and ADM2s (districts) that correspond to these project locations in the 3.0 version of the
GCDF dataset.
6
A project/activity with “precise” geocodes is one for which have highly precise boundaries of the
geofeature(s). A project/activity with “approximate” geocode is one identified within a 5 km radius of
the precise boundaries of the geofeature(s). For more details, see Goodman et al. (2023).
7
To better understand how highly precise data on the spatio-temporal rollout of Chinese grant- and
loan-financed projects/activities make it possible to estimate the causal effects of such projects/activities
on intended and unintended outcomes, see Dreher et al. (2019, 2022), Marty et al. (2019); Blair et al.
(2022); Baehr et al. (2022), Isaksson and Kotsadam (2018a, 2018b), Isaksson (2020), Martorano et al.
(2020), Iacoella et al. (2021), Malik et al. (2021), Bluhm et al. (2020), Anaxagorou et al (2020), Wellner et
al. (forthcoming), and Asmus et al. (forthcoming).

8
 5. Tracking China’s COVID-19 response efforts: The 3.0 dataset includes a marker of
whether it is known that a given project/activity was part of China's global COVID-19
response efforts. The “COVID Flag” field is set to "Yes" if the purpose of the
project/activity was related to COVID-19 control, including providing information,
education, and communication as well as activities or materials enabling testing,
prevention, immunization, treatment, or care. In cases where we identified the type and
amount of the item that was transferred in-kind but found no credible reporting
regarding the monetary value of the donation, we calculated estimated monetary values
based on the per-unit prices reported in the World Health Organization’s Emergency
Global Supplies Catalogues. In cases where we were able to estimate the monetary
value of the donation, we recorded the value in the “Transaction Amount” field and set
the “Amount Estimated” field to “Yes."
6. Infrastructure project flag: With a new infrastructure marker, the 3.0 dataset identifies
4,800 active and completed infrastructure projects, and an additional 94 infrastructure
projects for which grants and loans were issued but which were subsequently
suspended or canceled. We define "infrastructure projects" as those that involve
physical construction activities (e.g. roads, railways, pipelines, transmission lines, fiber
optic networks, etc.). More specifically, we capture projects that involve (1) building a
new physical structure, (2) rehabilitating or adding onto an existing physical structure,
and (3) maintaining an existing physical structure.

In addition, we have introduced a number of supplemental variables and structural changes, as

described below, to enhance the usability of the 3.0 dataset for analysis.

1. AidData Parent ID: To better represent the linkages between project/activity records,
the 3.0 dataset includes an "AidData Parent ID" variable. Project/activity records that
are related to each other are assigned to the same parent identification number.
Examples of cases in which project/activity records may be assigned to the same Parent
ID include (1) the records capture separate loans from Chinese state-owned financiers
which are allocated to the same project/activity, (2) the records capture separate
tranches of financing from the same framework agreement, (3) the records capture
financing to different phases of the same project/activity, (4) the records capture
financing that is derived from a special fund or joint fund.
2. Amount estimated: The 3.0 dataset includes a marker which designates whether
AidData estimated the monetary amount that the funding agency committed or
pledged. There are a number of circumstances in which AidData estimates transaction
(financial commitment) amounts. Examples include (1) If the precise face value of a
Preferential Buyer's Credit (PBC) or Buyer's Credit Loan (BCL) from China Eximbank is
unknown but the total cost of the commercial (EPC) contract is known, AidData assumes
that the face value of the PBC/BCL is equivalent to 85% of the total commercial (EPC)
contract cost; (2) If the face value of a syndicated loan (involving one or more official
sector creditors from China) is known and the total number of participants in the loan
syndicate is known, AidData assumes that each bank provided equal contributions to
the syndicated loan; (3) If material is transferred in-kind and there is no credible
reporting on the monetary value of the in-kind transfer, AidData calculates the monetary
value of the in-kind materials by multiplying the number of units of donated material by
the market value of those materials (in unit cost terms). Whenever a transaction
(financial commitment) amount has been estimated, AidData includes an explanation in
the “Description” or “Staff Comments” field.

9
 3. Grant element calculation and flow class determination: In previous versions of the
GCDF dataset, we calculated the grant element of each loan based on the OECD’s
cash-flow methodology and determined its flow class accordingly. Now, the 3.0 dataset
also provides a grant element variable based on the OECD’s grant-equivalent
methodology, and a grant element variable based on the current (post-2013) World
Bank/IMF methodology. These two additional grant element measures allow users to
select the method of measurement that best supports their research objectives and
ensures comparability across a wider range of datasets. It also ensures adherence to
newly established OECD standards. In order to make flow class determinations for flows
reported between commitment years 2000 and 2017, we follow the OECD's practice of
using the cash-flow methodology to define ODA, which relies on a 25% grant element
threshold and a discount rate of 10% for all loans. For flows reported in commitment
years 2018 and beyond, we use the OECD's grant-equivalent methodology, which
relies on a tiered concessionality threshold system for loans. Under the grant-equivalent
methodology, the threshold level of concessionality for loans to the official sector in the
recipient country is 45% for LDCs and other LICs (using a discount rate of 9%), 15% for
LMICs (using a discount rate of 7%) and 10% for UMICs (using a discount rate of 6%).
4. Flow type simplified: To more easily facilitate the aggregation of data for certain types
of analysis, the 3.0 dataset includes a “Flow Type Simplified” field in addition to the
standard “Flow Type” field that was included in previous versions of the dataset. This
field captures the type of financial or in-kind transfer supporting the project/activity in a
smaller number of categories than the “Flow Type” field. Each flow is assigned to one
of four categories: Grant, Loan, Debt Rescheduling, or Vague. Unlike the “Flow Type”
field, the "Grant" category in the “Flow Type Simplified” field encompasses "Grant,"
"Debt Forgiveness,” "Scholarships/Training in Donor Country," and "Free-standing
technical assistance" flows.
5. Recipient ISO-3 and income group: The 3.0 dataset includes the ISO Alpha-3 Country
Code for the recipient country of each project/activity, and a field that captures the
recipient country income status based on the OECD’s ODA Eligibility lists. The “OECD
ODA Income Group” field records whether a country is low income (LIC), lower middle
income (LMIC), upper middle income (UMIC), or high income (HIC).8 AidData uses this
field for the calculation of the grant element according to the OECD grant-equivalent
methodology (in the “Grant Element (OECD Grant-Equiv)” field), which uses a fixed
discount rate that depends on the recipient country income level (9% for LICs, 7% for
LMICs, and 6% for UMICs). The OECD income brackets also inform AidData’s
classification of projects as ODA-like, whereby loans must have a concessionality level
of at least 45% for LICs, 15% for LMICs, and 10% for UMICs.
6. Organizational presentation: Rather than separating the names, types and origins of
organizations into three separate fields, the 3.0 version of the dataset combines them
into two columns in the dataset: one column that displays the name of the organization
(e.g., Angola Ministry of Finance), and a subsequent column that displays both the
country of origin and the organization type (e.g., Recipient Government Agency).9

8
Note that the “OECD ODA Income Group” field does not align with the World Bank income group
classification. The list of countries eligible to receive ODA, as published by the OECD, is updated every
three years and thus there is a delay in registering income bracket changes.
9
In the 2.0 version of the dataset, the country of origin and organization type were captured in two
separate columns.

10
This paper has four sections. Section 1 provides an overview of the scope parameters and key
features of the 3.0 dataset as well as guidance regarding how the dataset should be used in
different types of applications. Section 2 explains how Chinese ODA and OOF is designed and
delivered to other countries, and the coding rules and procedures that AidData used to
capture and categorize projects financed with Chinese ODA and OOF. Section 3 describes the
sources and methods that AidData used to assemble a comprehensive and detailed picture of
Chinese ODA- and OOF-financed activities around the globe. Section 4 describes the methods
that AidData used to collect precise geographic locational details and geocode the dataset.

Section 1 - AidData’s Global Chinese Development

Finance Dataset, Version 3.0
1.1 - Scope Parameters of the 3.0 Dataset

When we set out to build Version 3.0 of AidData’s GCDF dataset, our goal was to create a
comprehensive and detailed picture of China’s overseas development finance program. The
dataset that we constructed covers all regions, all sectors, and all sources and types of financial
and in-kind transfers from government and state-owned institutions in China. There are other
datasets that capture official financial transfers from China to a single sector (e.g., energy) or
region (e.g., Latin America), or that only track certain types of financial flows (e.g. loans) and
funding sources (e.g. China’s policy banks). However, the 3.0 version of AidData's GCDF
dataset is unique in that it covers every major world region, every low-income and
middle-income country, all sectors, and all types of financial and in-kind transfers from
government and state-owned institutions in China. It is also different from other Chinese
development finance datasets in that it measures financial commitments in constant (i.e.
inflation-adjusted) U.S. dollars (USD), which ensures that reliable comparisons can be made
over time and geographic space.10

Prospective users should be aware of several scope parameters of the dataset:

➔ Financiers: The 3.0 dataset captures the full range of projects/activities that align with
the OECD’s definition of Official Development Assistance (ODA) and Other Official
Flows (OOF). Therefore, any project/activity that benefits from financial or in-kind
support from any official sector institution in China is included, regardless of its purpose,
level of financial concessionality, funding source, and overseas destination. The only
type of Official Financing we do not seek to capture is Official Investment, although we
do capture debt financing that facilitates investment. The 3.0 dataset captures projects
supported by 791 different official sector institutions in China, including central
government agencies (like the Ministry of Commerce, the Ministry of Foreign Affairs,
10
It does so by capturing official financial commitments in their original currencies of denomination,
converting these financial amounts into nominal USD values at the average exchange rates that were in
effect during the commitment years, and subsequently converting the nominal USD values to constant
2021 USD values using the OECD’s deflation methodology (to adjust for inflation and ensure
comparability over time and geographic space). See Appendix D.

11
 and the Ministry of Agriculture), regional and local government agencies (like
Chongqing Municipal Health Commission and Tianjin Municipal Government),
state-owned enterprises (like CNPC, CMEC, CATIC, and CRBC), state-owned policy
banks (like China Development Bank and China Eximbank), state-owned commercial
banks (like ICBC, BoC, and CCB), and state-owned funds (like the Silk Road Fund).
➔ Types of Flows: The 3.0 dataset captures grants, technical assistance, loans (with
categorization of 23 distinct loan instruments), debt forgiveness, debt rescheduling,
debt refinancing, scholarships, and training activities.11 By monetary value, the majority
of the ODA and OOF transfers (“flows”) that are captured in the dataset come from
official sector loans. Yet the majority of the project/activity records in the dataset
represent other types of financial or in-kind support. For many flow types other than
loans, AidData was not able to identify monetary commitment values. However, these
project/activity records still provide valuable information for users who are interested in
capturing the full set of Chinese ODA- and OOF-financed activities in a given world
region, country, or subnational area.
➔ Sectors: We capture all officially-financed projects/activities related to ODA and OOF,
regardless of the sector that they support. We classify each project/activity in the 3.0
dataset according to the OECD’s 3-digit sector classification scheme.12 AidData coders
follow the OECD’s classification guidelines to identify the sector that a given
project/activity is meant to support.
➔ Recipients: AidData used the 3.0 version of the TUFF methodology to systematically
search for projects/activities supported by official financial and in-kind transfers from
China across 165 countries and territories. The resulting dataset covers official financial
and in-kind transfers from China to every low-income, lower-middle income, and
upper-middle income country and territory across every major world region, including
Africa, Asia, Oceania, the Middle East, Latin America and the Caribbean, and Central
and Eastern Europe. It also covers 28 high-income countries (as of their 2021 OECD
ODA income status) that were included to help ensure comprehensive coverage in each
world region to the extent possible. In total, the dataset identifies Chinese
government-financed projects/activities in 146 countries and territories, meaning that
no projects were identified in 19 countries and territories despite systematic searches.
See Appendix B for a full list of the countries systematically included in the 3.0 dataset.
➔ Temporal Coverage: The 3.0 dataset captures the known universe of projects (with
development, commercial, and representational intent) supported by official financial
and in-kind commitments (and pledges) from China over a 22-year period (2000-2021),
with details on the timing of project implementation over a 24-year period (2000-2023).
The dataset also assigns every project/activity to one of six status categories (Pipeline:
Commitment, Implementation, Completed, Suspended, or Canceled) based on sources
that were available as late as August 2023.

11
See Section 2.5.3.3 for a detailed description of the loan instruments covered by the GCDF 3.0
dataset.
12
See
https://www.oecd.org/dac/financing-sustainable-development/development-finance-standards/purposec
odessectorclassification.htm

12
1.2 - Key Features of the Dataset

The 3.0 dataset includes 126 fields (variables), each seeking to capture a different aspect of a
project/activity or provide information about the sources used to compile the record. A
complete list of field names and definitions is provided in Section 1.4. The fields in the dataset
capture the following types of information about each project/activity:

➔ Basic Project/Activity Information: The dataset provides foundational information about

each project/activity, including its title in English, Chinese, and host country languages,
a unique and stable project/activity record identification number, the date of the official
commitment, the monetary value of the official commitment, the currency in which the
official commitment was denominated, the identity of the funder and recipient, the
primary purpose of the project/activity, the current status of the project/activity, and
URLs for all sources that supported the creation of the project/activity record.
➔ Transactional Details: The dataset identifies the nature of the financial or in-kind transfer
(e.g., grant, loan, technical assistance, debt forgiveness, debt rescheduling,
scholarship/training) supporting each project/activity in the dataset. Whenever
applicable, it documents loan pricing details (interest rate, default interest rate, grace
period, maturity, commitment fee, management fee, insurance fee); levels of financial
concessionality, as measured by the grant element calculators of the OECD and the
IMF; the first loan repayment date; the last loan repayment date; the monetary value
and timing of disbursements and repayments; the use of credit enhancements,
including guarantees, insurance, and collateral; the establishment of special purpose
vehicles, subsidiary on-lending arrangements, and escrow/revenue/special accounts;
and the monetary value and timing of underlying commercial contracts. The dataset
also provides stable URLs to unredacted grant, loan, and debt forgiveness/rescheduling
agreements whenever they have been successfully retrieved.
➔ Loan Type Categorization: Given that Chinese state-owned creditors rely on an
increasingly diverse set of loan instruments to finance projects/activities in low- and
middle-income countries, AidData has introduced a new loan categorization scheme in
the 3.0 version of the dataset that allows users to isolate 23 specific types of loan
instruments. See Section 2.5.3.3 for more details. Additionally, the 3.0 version of the
dataset identifies loans that involve an on-lending arrangement (in the “On-Lending”
field); loans that involve multilateral institutions as a loan administrator, co-financier,
insurer, and/or financial technical adviser (“Involving Multilateral” field); loans that
involve co-financing agencies from countries other than China (in the “Involving
Non-Chinese Financier” field); and loans that allow sovereign debtors to (i) service
existing debts, (ii) finance general budgetary expenditures and/or, (iii) shore up foreign
reserves (in the "Rescue" field).
➔ Development Finance Categorization: AidData seeks to designate each project/activity
in the 3.0 dataset as Official Development Assistance (ODA) or Other Official Flows
(OOF) based on measurement of the primary intent of the project/activity and the
concessionality level of the financing provided for the project/activity. AidData adheres
closely to the OECD-DAC reporting directives that outline the financial, structural, and
intent-related eligibility criteria for ODA and OOF. This unique feature of the dataset
allows users to make cross-donor and cross-lender comparisons at global, regional,
national, and subnational scales and over time. The 3.0 version of the dataset is aligned
with the OECD's newly established ODA thresholds (for loans issued in 2018 and

13
 subsequent years), which are based upon a tiered grant-equivalent methodology. See
Section 2 for more details. AidData has also included three different grant element
variables in the 3.0 dataset based on OECD and IMF methods of measuring financial
concessionality. See Section 1.4 for "Grant Element (OECD Cash-Flow)," "Grant
Element (OECD Grant-Equiv)," and "Grant Element (IMF)" field definitions.
➔ Sectoral Categorization: AidData assigns 3-digit OECD sector codes and names to all
projects/activities in the 3.0 dataset using the OECD’s classification criteria. This unique
feature of the dataset enables cross-donor and cross-lender comparisons—at global,
regional, national, and subnational scales—since most official sources of international
development finance (including all of the members of the OECD-DAC and the most
multilateral institutions) use the same criteria. It also facilitates analysis of sectoral
patterns and trends over geographic space and time.
➔ Stakeholder Organizations: Another feature that sets the 3.0 dataset apart is the level of
detail that it provides about the organizations involved in Chinese ODA- and
OOF-financed projects and activities. It provides information about nine potential types
of organizations for each project/activity: (1) the official sector institution in China that is
responsible for providing funding and/or in-kind support for the project/activity; (2) the
co-financing institutions from inside and outside of China that are supporting the same
project/activity; (3) the recipient institutions that are responsible for managing incoming
funds and in-kind transfers; (4) the agency or agencies that receive and manage a
financial transfer (loan) from the entity captured in the 'Direct Receiving Agencies' field
(as part of an on-lending arrangement); (5) the contractors and subcontractors that are
responsible for project/activity implementation; (6) the agency that provided a
repayment guarantee; (7) the third-party (accountable agency) that provided a credit
insurance policy to the borrower; (8) the agency that provided one or more sources of
collateral that can be seized in the event the borrower defaults on its repayment
obligation; and (9) the security agent or collateral agent that was appointed to enforce
rights against the collateral in the event that the borrower defaults on its debt
repayment obligations. The 3.0 dataset also categorizes each of these organizations by
type (i.e., Government Agency, State-Owned Bank, State-Owned Policy Bank,
State-Owned Commercial Bank, State-Owned Company, State-Owned Fund,
Intergovernmental Organization, Special Purpose Vehicle/Joint Venture, Private Sector,
NGO/CSO/Foundation, Other, or No Organization Type Specified) and country of origin
(i.e., China, Recipient Country, or Other). In the latest version of the dataset, AidData
identifies 791 official sector financing institutions from China, 1,225 co-financing
institutions, 5,037 recipient institutions, 4,933 implementing institutions, and 422
institutions that provide guarantees, insurance, or sources of collateral.
➔ Actual and Potential Loan Repayment Obligations: To facilitate more analysis of (actual
and potential) loan repayment obligations in cases when the borrower (direct receiving
agency) is a JV/SPV, the 3.0 version of the dataset has introduced new variables that
identify the extent of host government ownership (in the“Host Government Ownership”
field) and Chinese government ownership (in the “Chinese Government Ownership”
field) of the JV/SPV. The 3.0 version of the dataset also includes a new “Level of Public
Liability” field, which assigns each loan record to one of six categories: Central
government debt, Central government-guaranteed debt, Other public sector debt,
Potential public sector debt, Private debt, or Unallocable. Users who wish to isolate all
loans that qualify as public and publicly-guaranteed debt (PPG)—that is to say, loans to
government and majority state-owned institutions in the host country as well as other
institutions that secured central government repayment guarantees or repayment

14
 guarantees from state-owned entities other than the central government in the host
country—should identify all loan commitments assigned to the “Central government
debt,” “Central government-guaranteed debt,” and “Other public sector debt"
categories. The "Potential public sector debt" category captures loans to special
purpose vehicles (SPV) or joint ventures (JV) that are minority-owned by one or more
public sector institutions in the host country and that do not benefit from a central
government repayment guarantee or a repayment guarantee from a state-owned entity
other than the central government in the host country. The “Level of Public Liability”
field also allows for the identification of official sector loans to privately-owned entities
that do not benefit from repayment guarantees from public sector institutions in the
host country (“Private debt") and loans that cannot be easily categorized based on the
level of public liability (“Unallocable” debt).
➔ Timing of Project/Activity Implementation: The 3.0 dataset now includes a
“Commitment Date” field which records the calendar day on which the official financing
agreement was signed and provides an unprecedented level of detail on project/activity
commencement (implementation start) dates and project completion (implementation
end) dates. AidData identifies precise, calendar day-level commencement dates for
11,286 projects/activities, and calendar day-level completion dates for 11,542
projects/activities. By way of comparison, the 2.0 version of the dataset included
calendar day-level commencement dates for 5,539 projects/activities, and calendar
day-level completion dates for 6,061 projects/activities. The 3.0 dataset also includes
two new fields (“Deviation from Planned Implementation Start Date” and “Deviation
from Planned Completion Date)" so that users can easily determine if projects/activities
have been implemented on schedule, behind schedule, or ahead of schedule.
➔ Location Details: For projects/activities that have physical footprints or involve specific
locations, the 3.0 version of the dataset provides precise locational information that
technical users can use to conduct geospatial analysis and non-technical users can use
to easily view where projects/activities are taking (or have taken) place. Written
descriptions of the geographical locations and features of projects/activities and
OpenStreetMap links are available in the 'Location Narrative' field. The 3.0 dataset
extracts point, polygon, and line vector data via OpenStreetMap URLs and provides a
corresponding set of GeoJSON files and geographic precision codes for 9,497
projects/activities. 72% (6,919) of these projects/activities include “precise” or
“approximate” geocodes; the remaining 28% (2,578 projects/activities) are measured at
an administrative unit level. The separate dataset that provides precise geospatial
features (along with usage tips and related documentation) is accessible via
https://www.aiddata.org/data/aiddatas-geospatial-global-chinese-development-finance-
dataset-version-3-0. For those interested in ADM-level geographic analysis, AidData has
generated two new files included in the GCDF 3.0 zip file download: (1)
GCDF_3.0_ADM1_Locations and (2) GCDF_3.0_ADM2_Locations. See the GCDF 3.0
ADM Files README included in the GCDF 3.0 file download.
➔ Risks, Achievements, Failures, and Setbacks: The 3.0 dataset provides a suite of
variables (e.g., Commitment Year, Implementation Start Year, Completion Year, Status)
that allow users to track projects/activities over their full life-cycles. Whenever possible,
it also provides a detailed overview (in the "Description" field) of project/activity
achievements and failures, contractor performance vis-à-vis deadlines and deliverables,
findings from audits and evaluations, a summary of the various challenges that arose
during project/activity design and implementation (such as strikes, riots, public protests,
wars, corruption scandals, natural disasters, public health restrictions, political

15
 transitions, bankruptcies, loan defaults, contractual disputes, lawsuits, and ruptures in
diplomatic relations), and a description of how funding, receiving, implementing, and
accountable institutions responded to these challenges. The 3.0 dataset also includes a
“Financial Distress” flag that identifies whether, for a given loan, there is an indication
that the borrower had difficulty repaying the loan or was financially distressed during
the loan’s repayment period (according to the project/activity life-cycle information that
is identified in the description field).
➔ Sources: One of the hallmarks of the TUFF 3.0 methodology is source transparency. For
each record in the 3.0 dataset, a complete list of the sources is provided, including
public URLs, the title of the source, the publisher, and the type of source. In total, the
20,985 project/activity records in the dataset rely upon 147,703 sources, including
99,393 unique sources.13 The number of sources attached to each record vary from 1 to
124, with an average of 7 sources per record. The sources used to create the dataset
include both official and non-official sources.14 In constructing the 3.0 dataset, we
sought to identify and integrate as many official sources as possible. These sources are
authoritative in that they provide data and documentation from funding agencies,
recipient agencies, and implementing agencies that are directly involved in the
project/activity or have firsthand knowledge of the financial/in-kind transfer supporting
the project/activity.15 We assigned special priority to the use of these sources during
the construction of the 3.0 dataset. 87% of the project/activity records in the 3.0 dataset
were constructed with at least 1 official source.

1.3 - Guide to Using the Dataset

Given the comprehensive nature of the 3.0 dataset and some of the unique challenges that
arise when data on Chinese ODA- and OOF-financed projects/activities are collected from a
highly decentralized set of open sources, we have created six fields to help users easily identify
the subset of project/activity records that they wish to analyze. These fields include:

● Umbrella: This field is designed to capture hierarchical relationships between

projects/activities and various types of agreements. This field identifies projects and/or
activities that fall within "umbrella" agreements (with a “Yes” designation) in two
circumstances. The first circumstance is when a financial agreement was signed by at
least one party in the donor/creditor country and one party in the receiving country, but
funds were not allocated for a specific purpose (or set of purposes) until a subsequent
date. These types of umbrella agreements include Economic and Technical Cooperation
Agreements (ECTA) issued by China’s Ministry of Commerce (MOFCOM), master facility
agreements issued by China Eximbank, lines of credit issued by China Development

13
In many cases, official sources provide information about multiple projects, which is one of the main
reasons why the total number of unique sources is 99,393 but the total number of sources is 147,703.
14
Official source types include “Donor/Recipient Official Source,” “Implementing/Intermediary
Organization Source," and “Other Official Source (non-Donor, non-Recipient, non-Implementing)."
Non-official source types include “NGO/Civil Society/Advocacy (non-Donor, non-Recipient,
non-Implementing)," “Media Report," “Social Media, including Unofficial Blogs," “Academic Journal
Article," “Other Academic (Working Paper, Dissertation)," and “Other."
15
We also treat intergovernmental organizations—like the World Bank and the International Monetary
Fund—with aid and debt monitoring responsibilities as official sources.

16
 Bank, and Framework Agreements issued by a variety of official sector institutions in
China. Due to the nature of the TUFF data collection process, the subsidiary
transactions and projects/activities approved and financed under these types of
umbrella agreements are likely captured elsewhere in the dataset. The second
circumstance is when a single project/activity is financed by multiple Chinese
government or Chinese state-owned institutions. In these cases, AidData creates one
umbrella record to record the full amount of the financial commitment for the
project/activity and a linked set of subsidiary project/activity records to capture the
respective financial commitments of each Chinese government or Chinese state-owned
institution. Umbrella records are included in the 3.0 dataset to clarify linkages between
projects/activities and to capture relevant activities without double-counting financial
amounts or project counts. As a general rule, no umbrella records should be included in
financial analysis or analysis of project counts as doing so will almost certainly result in
double-counting. All umbrella agreements in the dataset are assigned a designation of
"No" in the "Recommended for Aggregates" field to help users avoid double counting.
● Status: The 3.0 dataset captures the full range of potential, active, completed and
suspended/canceled projects/activities, and it distinguishes among them using the
status field. This field identifies the latest status of a project/activity. Each
project/activity is assigned to one of six categories: Pipeline: Pledge, Pipeline:
Commitment, Implementation, Completed, Suspended, Canceled (see Section 1.4 for a
full description of each status). Projects/activities assigned to the Pipeline: Commitment,
Implementation, and Completed categories represent active or completed
projects/activities that either benefit(ed) from (1) a binding, written agreement that
governs the provision of financial or in-kind support from an official sector donor or
lender in China (especially for loans and large grants), or (2) the provision of financial or
in-kind support that has already taken place (e.g., humanitarian aid or small donations
that were handed over to the recipient). As such, we consider the portfolio of
projects/activities assigned to to the Pipeline: Commitment, Implementation, and
Completed categories to represent the actual portfolio of Chinese ODA and OOF (i.e.,
financial and in-kind transfers that have already happened, are underway, or scheduled
to take place in the future). In contrast, projects/activities assigned to the “Pipeline:
Pledge” category represent projects/activities that official sector institutions in China
have indicated interest in supporting but have no binding legal obligation to do so.
These projects/activities may benefit from financial and in-kind transfers in the future,
but additional steps need to be taken by the official sector institutions in China and/or
recipient country counterparts before the projects/activities can move forward with
support from Chinese ODA or OOF. Similarly, projects/activities assigned to the
Suspended and Canceled categories represent those that were backed by an official
commitment but subsequently suspended or canceled (typically due to project design
or implementation problems or disagreements). For analysis that requires the
aggregation of projects supported by Chinese ODA or OOF commitments, including
analysis of monetary amounts and project counts, only projects assigned to the
Pipeline: Commitment, Implementation, and Completed categories should be included.
However, given that some analysts are interested in better understanding China’s
portfolio of pledged, canceled, and suspended projects, we have included them in the
full dataset to provide flexibility to users.
● Recommended for Aggregates: We recommend using this field for analysis that
requires the aggregation of projects/activities supported by official financial (or in-kind)
commitments from China, including analysis of monetary amounts and project/activity

17
 counts. It is useful for identifying formally approved, active, and completed Chinese
ODA- and OOF-financed projects/activities – and excluding all canceled
projects/activities, suspended projects/activities, and projects/activities that never
reached the formal approval (official commitment) stage. The field is set to "Yes" for all
projects/activities with a status designation of Pipeline: Commitment, Implementation,
and Completion that have not also been designated as umbrella agreements. It is set to
“No” for all canceled projects/activities, suspended projects/activities, and
projects/activities that never reached the official commitment stage (i.e. those
projects/activities with a status designation of Pipeline: Pledge, Suspended, and
Canceled). Additionally, to avoid double-counting, the field is set to “No” for all
umbrella agreements. Also, note that not all projects/activities with a “Recommended
for Aggregates” value of “Yes” identify a financial transaction value (since some
transactions are difficult to monetize, such as in-kind donations, technical assistance,
scholarships, and training activities).
● Flow Type: This field captures the type of financial or in-kind transfer supporting the
project/activity. Each project/activity is assigned to one of seven categories: Loan, Debt
Forgiveness, Debt Rescheduling, Grant, Scholarships/Training in Donor Country,
Free-standing Technical Assistance, and Vague TBD. For projects/activities that are
assigned to the "Loan" category, the dataset includes a host of other variables that
capture the type of loan, the borrowing terms, the use of credit enhancements, and the
involvement of co-financiers, among other things. See Section 2 for a more detailed
description.
○ In cases of debt forgiveness, the Umbrella field is set to "Yes" if the original
contracted loan could be captured elsewhere in the dataset as a loan record.
This is done to avoid double counting. If the original contracted loans occurred
before 2000 (when the dataset begins to track Chinese ODA and OOF), then the
Umbrella field is set to "No." As such, if users are interested in isolating all cases
of debt forgiveness, AidData recommends turning the “Recommended for
Aggregates” filter off and then using the “Flow Type” field to identify all
projects/activities assigned to the “Debt Forgiveness” category (irrespective of
whether they are coded as umbrella records).
○ Also, to help users avoid double-counting, AidData does not populate any fields
related to transaction amounts [Amount (Original Currency), Adjusted Amount
(Original Currency), Amount (Constant USD 2021), Adjusted Amount (Constant
USD 2021), Amount (Nominal USD), and Adjusted Amount (Nominal USD)] for
projects/activities assigned to the “Debt Rescheduling” category. However,
users who wish to undertake analysis of debt reschedulings can find detailed
information about the terms and conditions of these reschedulings in the
“Description” fields of the projects/activities that are assigned to the “Debt
Rescheduling” category.
● Flow Class: Based on OECD-DAC guidelines for Official Development Assistance (ODA)
and Other Official Flows (OOF), this field assigns projects to one of three categories:
ODA-like, OOF-like, Vague (Official Finance). See Section 2 for full descriptions of each
category, including the concessionality thresholds for ODA/OOF and how AidData has
applied them to the dataset. Flow class is a useful distinction for users who wish to (a)
distinguish between “development aid” in the strict/traditional sense of the term (i.e.,
ODA) and official sector loans that are non-concessional or semi-concessional in nature
(i.e., OOF); or (b) compare Chinese development finance to other sources of

18
 development finance that are categorized according to OECD-DAC definition and
measurement criteria.
● Adjusted Amount: This field captures the “adjusted” monetary amount that a funding
agency committed (or pledged) in its original currency of denomination. AidData
recommends using this field to calculate the cumulative stock of official financial flows
(ODA/OOF commitments) from China over multiple years—when one or more of
recipient countries secured "rollover" emergency rescue loans and/or swap borrowings
from the People's Bank of China (PBOC) to refinance their maturing debts. For grants
and non-emergency loans, the amounts that are recorded in this field are identical to
the amounts that are recorded in the Amount field (which is not labeled as “Adjusted”).
See Section 1.4 for a more detailed definition of the “Adjusted Amount” field, and see
Section 2 for further details regarding Chinese lending instruments.

While users of the 3.0 dataset may rely on additional fields to identify the subset of transfers
(flows) they are interested in better understanding, the above-mentioned fields should be
carefully considered before conducting any analysis.

1.4 - Field Definitions

The 3.0 dataset includes 126 separate fields (variables) to document a detailed picture of each
Chinese ODA- and OOF-financed project/activity. Field names and definitions are provided in
the table below.

Field Name Description

This field provides the unique identification number that AidData has
AidData Record ID assigned to every project/activity record in the dataset.
This field identifies projects/activities that AidData recommends including
in analysis that requires the aggregation of projects/activities supported
by official financial (or in-kind) commitments from China, including analysis
of monetary amounts and project/activity counts. It is useful for identifying
formally approved, active, and completed Chinese ODA- and
OOF-financed projects/activities—and excluding all canceled
projects/activities, suspended projects/activities, and projects/activities
that never reached the formal approval (official commitment) stage. The
field is set to "Yes" for all projects/activities with a status designation of
Pipeline: Commitment, Implementation, and Completion that have not
also been designated as umbrella agreements. It is set to “No” for all
canceled projects/activities, suspended projects/activities, and
projects/activities that never reached the official commitment stage (i.e.
those projects/activities with a status designation of Pipeline: Pledge,
Suspended, and Canceled). Additionally, to avoid double-counting, the
field is set to “No” for all umbrella agreements. For more information on
umbrella agreements, see the description of the “Umbrella” field in this
Recommended file. Also, note that not all projects/activities with a “Recommended for
For Aggregates Aggregates” value of “True” identify a financial transaction value (since

19
 some transactions are difficult to monetize, such as in-kind donations,
technical assistance, scholarships, and training activities).
This field captures the linkages between project/activity records, whereby
project/activity records that are related to each other are assigned to the
same linked package. Each linked package is assigned a unique parent
AidData Parent ID identification number.
This field identifies projects and/or activities that fall within "umbrella"
agreements (with a “Yes” designation) in two circumstances. The first
circumstance is when a financial agreement was signed by at least one
party in the donor/creditor country and one party in the receiving country,
but funds were not allocated for a specific purpose (or set of purposes)
until a subsequent date. These types of umbrella agreements include
Economic and Technical Cooperation Agreements (ECTA) issued by
China’s Ministry of Commerce (MOFCOM), master facility agreements
issued by China Eximbank, lines of credit issued by China Development
Bank, and Framework Agreements issued by a variety of official sector
institutions in China. Due to the nature of the TUFF data collection
process, the subsidiary transactions and projects/activities approved and
financed under these types of umbrella agreements are likely captured
elsewhere in the dataset. The second circumstance is when a single
project/activity is financed by multiple Chinese government or Chinese
state-owned institutions. In these cases, AidData creates one umbrella
record to record the full amount of the financial commitment for the
project/activity and a linked set of subsidiary project/activity records to
capture the respective financial commitments of each Chinese
government or Chinese state-owned institution. All umbrella agreements
in the dataset are assigned a designation of "No" in the "Recommended
Umbrella for Aggregates" field to help users avoid double counting.
This field captures the country from which the official financial or in-kind
Financier Country transfer originated.
This field captures the country from which the entity receiving the official
financial or in-kind transfer is located. If multiple entities from multiple
recipient countries are involved, this field records the geographical region
Recipient to which the recipient countries belong.
This field captures the three-letter code for the country identified in the
'Recipient' field, according to the standards set by the International
Organization for Standardization (ISO). In cases where the 'Recipient' field
records the geographical region from which multiple recipient countries
belong (such as "Africa, Regional"), the 'Recipient ISO Alpha-3 Country
Recipient ISO-3 Code' field is left blank.
This field captures the geographical region to which the recipient country
belongs: Africa, Americas, Asia, Europe, Middle East, Oceania, or
Recipient Region Multi-Region.

20
 This field captures the year in which an official financial commitment (or
official commitment to provide in-kind support) was codified through the
signing of a formal agreement by an official donor/lender in China and
one or more entities in a recipient country or a set of recipient countries.
Whenever possible, this field is based on the precise calendar day when
the official commitment was issued, which is captured in the 'Commitment
Date' field. In the event an official commitment was made for a
project/activity that entered implementation, but the official commitment
year is not identifiable, AidData records the first year of project/activity
implementation as a proxy for the official commitment year. In the event
an official commitment was made for a project/activity that has not yet
reached implementation, and the official commitment year is not
identifiable, AidData records the year in which the underlying commercial
contract (supported by the official commitment) was issued. If this
information is unavailable, AidData records the first year in which an
informal pledge was made as a proxy for the official commitment year. For
projects/activities with a status designation of Pipeline: Pledge (i.e. cases
in which an official commitment was not made), AidData records the year
Commitment Year in which the informal pledge was made.
This field captures the year in which a project/activity supported by an
official financial (or in-kind) commitment from China began
implementation. Whenever possible, this field is based on the precise
calendar day when project/activity implementation began, which is
captured in the 'Actual Implementation Start Date' field. For
projects/activities that involve the construction of buildings or
infrastructure, the 'Implementation Start Year' field seeks to capture the
first year of construction. In cases when the first year of construction is
unavailable but a proxy for the first year of construction (e.g., the year in
which a formal groundbreaking ceremony took place, a project/activity
commencement order was issued to the contractor responsible for
implementation, or a project/activity implementation agreement was
signed) can be identified, AidData records the proxy for the first year of
construction. For projects/activities that do not involve construction but
involve the provision of personnel, training, analytical/advisory support,
equipment, supplies, or commodities, the 'Implementation Start Year' field
captures the first year in which some type of support was delivered to an
entity in the recipient country. For projects/activities that only involve
financial transactions (e.g., cash donations, loans issued to shore up a
country’s foreign exchange reserves, forgiveness or rescheduling of
outstanding debts), the 'Implementation Start Year' field captures the year
Implementation in which the first disbursement was made (or the year in which new terms
Start Year and conditions went into effect for a previously signed loan agreement).

21
 This field captures the year in which a project/activity supported by an
official financial (or in-kind) commitment from China was completed.
Whenever possible, this field is based on precise calendar day when a
project/activity was completed, which is captured in the 'Actual
Completion Date' field. For projects/activities that involve the construction
of buildings or infrastructure, the 'Completion Year' field seeks to capture
the last year of construction. In cases when the last year of construction is
unavailable but a proxy for the last year of construction (e.g., a road or
railway is opened for use, a power plant reaches its commercial operation
date and begins selling electricity to customers) can be identified,
AidData records the proxy for the last year of construction. For
projects/activities that do not involve construction but involve the
provision of personnel, training, analytical/advisory support, equipment,
supplies, or commodities, the 'Completion Year' field captures the last
year in which some type of support was delivered to an entity (or set of
entities) in the recipient country. For projects/activities that only involve
financial transactions (cash donations, loans issued to shore up foreign
exchange reserves, forgiveness or rescheduling of outstanding debts), the
'Completion Year' field captures the year in which the last disbursement
was made (or the year in which new terms and conditions went into effect
Completion Year for a previously signed loan agreement).
This field briefly describes the name or nature of the project/activity. The
identification numbers of other transactions that are linked to the
Title project/activity are also recorded in this field.
This field provides a detailed summary of the main purposes and activities
of the project/activity; the funding, receiving, and implementing agencies
involved in the project/activity; the terms and conditions of the financial
transaction(s) supporting the project/activity; the timing of project/activity
implementation and completion; the challenges that arose during
project/activity implementation and how funding, receiving, and
implementing agencies responded to these challenges; and main
achievements and shortcomings of the project/activity. For loan-financed
projects/activities, AidData also records the monetary value and timing of
underlying commercial contracts, disbursements, and repayments in this
Description field.
This field captures comments from AidData staff that clarify the
assumptions, logic, and evidence used to address challenging coding and
categorization determinations. It also provides foreign translations of
project/activity titles (used for source identification purposes), information
about related transactions and projects/activities, and information about
the ownership structures of funding, receiving, and implementing
Staff Comments agencies.

22
 This field identifies the latest status of a project/activity. Each
project/activity is assigned to one of six categories: Pipeline: Pledge,
Pipeline: Commitment, Implementation, Completed, Suspended,
Canceled. A project/activity assigned to the “Pipeline: Pledge” category is
one that an official sector institution in China indicated it was interested in
supporting (or willing to consider supporting) but did not result in an
official commitment. Projects/activities assigned to this category include
those that are identified in letters of intent, term sheets, memoranda of
understanding, and non-binding announcements. All projects/activities
given a status designation of Pipeline: Commitment, Implementation,
Completed, Suspended, or Canceled reached the official commitment
stage (i.e., a binding, written agreement that governs the provision of
financial or in-kind support for a specific purpose was signed by an official
sector donor or lender in China and an entity in a recipient country). A
project/activity assigned to the “Pipeline: Commitment” category is one
that is backed by an official commitment but has not yet entered
implementation. A project/activity assigned to the “Implementation”
category is one that is backed by an official commitment and has begun
implementation with financial or in-kind support from the source of the
commitment. A project/activity assigned to the “Completion” category is
one that is backed by an official commitment and that reached completion
with financial or in-kind support from the sources of the commitment.
Projects/activities assigned to the “Suspended” and “Canceled”
categories are those that were backed by an official commitment but
subsequently suspended or canceled. The coding of the “Status” field in
the dataset is based on sources that were available as late as August
Status 2023.
This field seeks to measure the primary purpose of the project/activity.
Each project/activity is assigned to one of five categories: Development,
Commercial, Representational, Mixed, or Military. Projects/activities
assigned to the “Development” category are those that are primarily
oriented towards the promotion of economic development and welfare in
the recipient country. Projects/activities assigned to the “Commercial”
category are those that primarily seek to promote the commercial
interests of the country from which the financial transfer originated (e.g.,
encouraging the export of Chinese goods and services). Projects/activities
assigned to the “Representational” category are those that primarily seek
to promote a bilateral relationship with another country or promote the
language, culture, or values of the country from which the financial
transfer originated (e.g., the establishment of a Confucius Institute or
Chinese cultural center). If a project/activity is assigned to the “Mixed”
category, this designation indicates that it was not possible for AidData to
identify the primary purpose of the project/activity and the project/activity
has multiple purposes (i.e., some combination of development,
commercial, and/or representational intent). Projects/activities assigned to
the “Military” category are those that seek to promote the security
interests of the country from which the financial transfer originates or
Intent strengthen the capabilities of military institutions in the recipient country.

23
 This field captures the type of financial or in-kind transfer supporting the
project/activity. Each project/activity is assigned to one of seven
categories: Loan, Debt Forgiveness, Debt Rescheduling, Grant,
Scholarships/Training in Donor Country, Free-standing Technical
Assistance, and Vague TBD. For projects/activities that are assigned to the
"Loan" category, the dataset includes a host of other variables that
capture the type of loan, the borrowing terms, the use of credit
enhancements, and the involvement of co-financiers, among other things.
In cases of debt forgiveness, the Umbrella field is set to "Yes" if the
original contracted loan could be captured elsewhere in the dataset as a
loan record. This is done to avoid double counting. If the original
contracted loans occurred before 2000 (when the dataset begins to track
Chinese ODA and OOF), then the Umbrella field is set to "No." As such, if
users are interested in isolating all cases of debt forgiveness, AidData
recommends turning the “Recommended for Aggregates” filter off and
then using the “Flow Type” field to identify all projects/activities assigned
to the “Debt Forgiveness” category (irrespective of whether they are
coded as umbrella records). Also, to help users avoid double-counting,
AidData does not populate any fields related to transaction amounts
[Amount (Original Currency), Adjusted Amount (Original Currency),
Amount (Constant USD 2021), Adjusted Amount (Constant USD 2021),
Amount (Nominal USD), and Adjusted Amount (Nominal USD)] for
projects/activities assigned to the “Debt Rescheduling” category.
However, users who wish to undertake analysis of debt reschedulings can
find detailed information about the terms and conditions of these
reschedulings in the “Description” fields of the projects/activities that are
Flow Type assigned to the “Debt Rescheduling” category.
This field captures the type of financial or in-kind transfer supporting the
project/activity in a smaller number of categories than the 'Flow Type'
field in order to facilitate the aggregation of flows based on certain
criteria. Each flow is assigned to one of four categories: Grant, Loan, Debt
Rescheduling, and Vague. Compared with the 'Flow Type' field, the
"Grant" category in this data field includes the "Grant," "Debt
Flow Type Forgiveness,” "Scholarships/Training in Donor Country," and
Simplified "Free-standing technical assistance" flows.

24
 This field identifies the concessionality threshold applied to any loan
record to determine if it met the concessionality threshold to quality as
ODA. This threshold is based on the year in which the project/activity
secures a Chinese ODA or OOF commitment, the income level of the
borrowing recipient country, and the receiving agency type. For all loans
issued between 2000 and 2017, the threshold is always 25% (using a
unified 10% discount rate). For loans issued between 2018 and 2021, the
threshold is determined based on the information below. For loans to
official sector institutions, the following concessionality thresholds apply:
(1) Least-developed countries and low-income countries: a minimum grant
element of 45% (calculated using a 9% discount rate), (2) Lower-middle
income countries: a minimum grant element of 15% (calculated using a
7% discount rate), and (3) Upper-middle income countries: a minimum
grant element of 10% (calculated using a discount rate of 6%), (4) Loans to
OECD ODA private sector institutions (regardless of the income level of the recipient
Concessionality country): a minimum grant element of 25% (using a unified 10% discount
Threshold rate).
Based on the OECD-DAC methodology between 2000-2021 to measure
Official Development Assistance (ODA) and Other Official Flows (OOF),
this field assigns projects/activities to one of three categories: ODA-like,
OOF-like, or Vague (Official Finance). Projects/activities are assigned to
the ODA-Like category if they meet three criteria. First, the primary
purpose of the project/activity must be the promotion of economic
development and welfare in the recipient country (i.e., have development
intent). Second, the project/activity must take place in a country that
qualifies for ODA based on its income level. Third, the official
commitment supporting the project/activity must be concessional in
nature (i.e., grant, technical assistance, scholarship, debt forgiveness, or
loan with a grant element meeting a specified threshold). For official
commitments issued (flows reported) between 2000 and 2017, we follow
the OECD's practice to use the cash-flow methodology to define ODA,
which included a threshold level of 25% grant element with a discount
rate of 10% for all loans. For official commitments issued (flows reported)
in 2018 and subsequent years, we use the OECD's grant-equivalent
methodology, which relies upon a tiered concessionality threshold system
for loans. Under the grant-equivalent methodology, the threshold
concessionality for loans to the official sector in the recipient country is
45% for LDCs and other LICs (using a discount rate of 9%), 15% for LMICs
(using a discount rate of 7%) and 10% for UMICs (using a discount rate of
6%). Loans to the private sector, however, continue to use the 25%
threshold used in the cash-flow methodology (in alignment with
OECD-DAC practices). Users can refer to the "OECD ODA
Concessionality Threshold" field to identify the threshold used for a
particular loan record in the dataset. Projects/activities that are supported
by an official financial or in-kind transfer but do not meet all three of these
criteria are assigned to the OOF-Like category. Projects/activities that are
backed by an official commitment but cannot be reliably categorized as
Flow Class ODA-like or OOF-like because of insufficiently detailed information are

25
 assigned to the “Vague (Official Finance)” category. Projects/activities in
this residual category primarily consist of (a) those with an unspecified
“Flow Type” (i.e., values of “Vague TBD”); and (b) those financed with
development-intent loans for which AidData lacks the borrowing terms
(interest rates, grace periods, or maturity dates) needed for
concessionality determinations. Users who would like to use one
concessionality threshold for the entire time period (or a subset) can use
the relevant grant element calculator to re-classify ODA/OOF loan records
where necessary.
This field provides a 3-digit sector code based upon the primary sectoral
focus of the project/activity. It is based upon the OECD’s sector
categorization scheme. There are 24, three-digit OECD sector codes:
education (110), health (120), population policies/programs and
reproductive health (130), water supply and sanitation (140), government
and civil society (150), other social infrastructure and services (160),
transport and storage (210), communications (220), energy (230), banking
and financial services (240), business and other services (250), agriculture,
forestry and fishing (310), industry, mining, and construction (320), trade
policies and regulation (330), general environmental protection (410),
other multisector (430), general budget support (510), developmental
food aid/food security assistance (520), other commodity assistance (530),
action relating to debt (600), emergency response (720), reconstruction
relief and rehabilitation (730), disaster prevention and preparedness (740),
Sector Code and unallocated/unspecified (998).
This field provides a sector name based upon the primary sectoral focus of
the project/activity. It is based upon the OECD’s sector categorization
scheme. There are 24, three-digit OECD sector codes: education (110),
health (120), population policies/programs and reproductive health (130),
water supply and sanitation (140), government and civil society (150),
other social infrastructure and services (160), transport and storage (210),
communications (220), energy (230), banking and financial services (240),
business and other services (250), agriculture, forestry and fishing (310),
industry, mining, and construction (320), trade policies and regulation
(330), general environmental protection (410), other multisector (430),
general budget support (510), developmental food aid/food security
assistance (520), other commodity assistance (530), action relating to debt
(600), emergency response (720), reconstruction relief and rehabilitation
(730), disaster prevention and preparedness (740), and
Sector Name unallocated/unspecified (998).
This flag provides a marker of whether a project/activity is an infrastructure
project. In the 3.0 version of the dataset, "infrastructure projects"
generally include those that involve physical construction activities (e.g.
roads, railways, pipelines, transmission lines, fiber optic networks). More
specifically, "infrastructure projects" include those that involve (1) building
a new physical structure, (2) rehabilitating or adding onto an existing
physical structure, and/or (3) maintaining an existing physical structure.
Infrastructure The 3.0 version of the dataset does not include projects/activities that

26
 involve the provision of cash, technical assistance, scholarships,
equipment, or supplies in its definition of "infrastructure projects." The
field is set to "Yes" if a project/activity is classifiable as an infrastructure
project.
This field provides a marker of whether it is known that the project/activity
is part of China's global COVID-19 response efforts. The field is set to
"Yes" if the purpose of the project/activity is related to COVID-19 control,
including providing information, education and communication as well as
activities or materials enabling testing, prevention, immunization,
COVID treatment, or care.
This field captures the name of the agency that issued the official financial
or in-kind commitment. The agency identified in this field must be based
in the country (the People’s Republic of China) from which the official
financial or in-kind commitment originated. For projects/activities
assigned to the Pipeline: Pledge category, this field captures the name of
the official sector agency that issued the pledge. The same “origin rule”
applies to funding agencies that issued pledges rather than commitments.
If multiple Chinese funding agencies are involved, the entries are
Funding Agencies pipe-delimited.
This field captures the type of funding agency that issued the commitment
or pledge. Each project/activity is assigned to one of seven categories:
Government Agency, State-Owned Policy Bank, State-Owned Commercial
Bank, State-Owned Bank, State-Owned Company, and State-Owned
Funding Agencies Fund. If multiple official sector funding agencies from China are involved,
Type the entries recording their types are pipe-delimited.
This marker indicates whether a separate funding agency (belonging to
the financier country or another country) provided funding for the
Cofinanced project/activity.
This field provides the names of the cofinancing agencies providing
Cofinancing funding for the project/activity. If multiple cofinancing agencies are
Agencies involved, the entries are pipe-delimited.
This field captures the type of cofinancing agency that provided funding,
as well as the agency's country of origin. Each cofinancing agency is
assigned to one of twelve categories: Government Agency, State-Owned
Bank, State-Owned Policy Bank, State-Owned Commercial Bank,
State-Owned Company, State-Owned Fund, Intergovernmental
Organization, Special Purpose Vehicle/Joint Venture, Private Sector,
NGO/CSO/Foundation, Other, or No Organization Type Specified. Each
cofinancing agency is also categorized based on whether it is from the
financier country, the recipient country, or another country. The
organization type is preceded by one of three descriptors regarding the
country of origin: Chinese, Recipient, or Other (e.g. Chinese State-Owned
Cofinancing Commercial Bank). If multiple cofinancing agencies are involved, the
Agencies Type entries are pipe-delimited.

27
 This field provides the name of the agency designated to receive and
manage the financial or in-kind transfer. For projects/activities that are
financed with loans, the receiving agency is the entity responsible for debt
repayment. If a receiving agency (borrower) on-lends the proceeds of a
loan to an additional entity or entities, then the borrower is captured in
the 'Direct Receiving Agencies' field and the additional entity or entities
which receive loans from the borrower is captured in the 'Indirect
Receiving Agencies' field. If more than one entity is responsible for
receiving and managing incoming grant funds or an in-kind transfer, all of
Direct Receiving these entities are identified in the 'Direct Receiving Agencies' field (as
Agencies pipe-delimited entries).
This field captures the type of agency designated to receive and manage
the financial or in-kind transfer, as well as the agency's country of origin.
Each direct receiving agency is assigned to one of twelve categories:
Government Agency, State-Owned Bank, State-Owned Policy Bank,
State-Owned Commercial Bank, State-Owned Company, State-Owned
Fund, Intergovernmental Organization, Special Purpose Vehicle/Joint
Venture, Private Sector, NGO/CSO/Foundation, Other, or No Organization
Type Specified. Each direct receiving agency is also categorized based on
whether it is from the financier country, the recipient country, or another
country. The organization type is preceded by one of three descriptors
regarding the country of origin: Chinese, Recipient, or Other (e.g.
Direct Receiving Recipient Government Agency). If multiple direct receiving agencies are
Agencies Type involved, the entries are pipe-delimited.
This field provides the name of the agency or agencies that receive and
manage a financial transfer (loan) from the entity captured in the 'Direct
Receiving Agencies' field (as part of an on-lending arrangement). If
Indirect Receiving multiple indirect receiving agencies are involved, the entries are
Agencies pipe-delimited.
This field captures the type of agency that received a financial transfer
from the entity captured in the 'Direct Receiving Agencies' field, as well as
the indirect receiving agency's country of origin. Each indirect receiving
agency is assigned to one of twelve categories: Government Agency,
State-Owned Bank, State-Owned Policy Bank, State-Owned Commercial
Bank, State-Owned Company, State-Owned Fund, Intergovernmental
Organization, Special Purpose Vehicle/Joint Venture, Private Sector,
NGO/CSO/Foundation, Other, or No Organization Type Specified. Each
indirect receiving agency is also categorized based on whether it is from
the financier country, the recipient country, or another country. The
organization type is preceded by one of three descriptors regarding the
country of origin: Chinese, Recipient, or Other (e.g. Recipient Government
Indirect Receiving Agency). If multiple indirect receiving agencies are involved, the entries
Agencies Type are pipe-delimited.

28
 This field provides a marker of whether it is known that the loan involves
an on-lending arrangement, which is an arrangement in which a receiving
agency (borrower) uses the proceeds of a loan to lend to one or more
additional entities. In an on-lending arrangement, the borrower is
captured in the 'Direct Receiving Agencies' field, and the entity or entities
which receive a loan from the borrower is captured in the 'Indirect
Receiving Agencies' field. The On-Lending field is set to "Yes" if the loan
On-Lending involves an on-lending arrangement.
This field provides the name of the agency responsible for implementing
the project/activity. If more than one agency is responsible for
Implementing implementing the project/activity, all such agencies are identified in the
Agencies 'Implementing Agencies' field (as pipe-delimited entries).
This field captures the type of agency that is responsible for implementing
project/activity, as well as the agency's country of origin. Each
implementing agency is assigned to one of twelve categories:
Government Agency, State-Owned Bank, State-Owned Policy Bank,
State-Owned Commercial Bank, State-Owned Company, State-Owned
Fund, Intergovernmental Organization, Special Purpose Vehicle/Joint
Venture, Private Sector, NGO/CSO/Foundation, Other, or No Organization
Type Specified. Each implementing agency is also categorized based on
whether it is from the financier country, the recipient country, or another
country. The organization type is preceded by one of three descriptors
Implementing regarding the country of origin: Chinese, Recipient, or Other (e.g. Chinese
Agencies Type State-Owned Company).
This field provides a marker of whether it is known that a loan repayment
guarantee was issued by a third-party (accountable agency). It assumes a
Guarantee value of “Yes” whenever a sovereign or corporate guarantee is issued in
Provided support of a loan.
This field provides the name of the agency that provided a repayment
guarantee in the event the borrower (i.e. direct receiving agency) cannot
Guarantor meet its debt repayment obligations.
This field captures the type of agency that issued a guarantee, as well as
the agency's country of origin. Each agency (guarantor) is assigned to one
of twelve categories: Government Agency, State-Owned Bank,
State-Owned Policy Bank, State-Owned Commercial Bank, State-Owned
Company, State-Owned Fund, Intergovernmental Organization, Special
Purpose Vehicle/Joint Venture, Private Sector, NGO/CSO/Foundation,
Other, or No Organization Type Specified. Each agency (guarantor) is also
categorized based on whether it is from the financier country, the
recipient country, or another country. The organization type of the agency
(guarantor) is preceded by one of three descriptors regarding the country
Guarantor Agency of origin: Chinese, Recipient, or Other (e.g. Recipient Government
Type Agency).
This field provides a marker of whether it is known that a third-party
(accountable agency) provided a credit insurance policy to the borrower
Insurance Provided (receiving agency). For example, it assumes a value of “Yes” whenever a

29
 loan is backed by credit insurance policy from China Export & Credit
Insurance Corporation (Sinosure).
This field provides the name of the agency that provided a credit
insurance policy to ensure repayment in the event the borrower (i.e. direct
Insurance Provider receiving agency) cannot meet its debt repayment obligations.
This field captures the type of agency that provided a credit insurance
policy, as well as the agency's country of origin. Each agency (insurance
provider) is assigned to one of twelve categories: Government Agency,
State-Owned Bank, State-Owned Policy Bank, State-Owned Commercial
Bank, State-Owned Company, State-Owned Fund, Intergovernmental
Organization, Special Purpose Vehicle/Joint Venture, Private Sector,
NGO/CSO/Foundation, Other, or No Organization Type Specified. Each
agency (insurance provider) is also categorized based on whether it is from
the financier country, the recipient country, or another country. The
organization type is preceded by one of three descriptors regarding the
Insurance Provider country of origin: Chinese, Recipient, or Other (e.g. Chinese State-Owned
Agency Type Company).
This field provides a marker of whether it is known that one or more
sources of collateral (security) can be seized in the event the borrower
defaults on its loan repayment obligations. It assumes a value of “Yes”
when (i) a loan is collateralized through a formal lien or security interest; (ii)
a borrower is required to deposit project-related revenues or unrelated
revenues in a special account, escrow account, revenue account, proceeds
account, payment account, reserve account, or repayment reserve account
that can be accessed (and debited) by the lender; and/or (iii) a security
agent is appointed (to enforce rights against the collateral in the event
that the borrower defaults on its repayment obligations). AidData also
codes all pre-export finance (PXF) facilities as collateralized since they are
almost always secured by (1) an assignment of rights by the producer
under an ‘offtake contract’ (i.e., a sale and purchase contract between the
producer and a buyer of that producer of goods or commodities), and (2)
a collection account charge over a bank account into which proceeds due
Collateralized/Sec to the producer from the buyer of the goods or commodities under the
uritized offtake contract are credited.
This field provides the name of the agency that provided one or more
sources of collateral (e.g., an assignor, mortgagor, pledgor, transferor,
lienee) that can be seized in the event the borrower defaults on its
Collateral Provider repayment obligations.
This field captures the type of agency that provided collateral, as well as
the agency's country of origin. Each agency that provided collateral is
assigned to one of twelve categories: Government Agency, State-Owned
Bank, State-Owned Policy Bank, State-Owned Commercial Bank,
State-Owned Company, State-Owned Fund, Intergovernmental
Organization, Special Purpose Vehicle/Joint Venture, Private Sector,
Collateral Provider NGO/CSO/Foundation, Other, or No Organization Type Specified. Each
Agency Type agency that provided collateral is also categorized based on whether it is

30
 from the financier country, the recipient country, or another country. The
organization type is preceded by one of three descriptors regarding the
country of origin: Chinese, Recipient, or Other (e.g. Recipient Private
Sector).
Security This field provides the name of the security agent or collateral agent that
Agent/Collateral was appointed to enforce rights against the collateral in the event that the
Agent borrower defaults on its debt repayment obligations.
This field captures the type of security agent or collateral agent, as well as
its country of origin. Each security agent or collateral agent is assigned to
one of twelve categories: Government Agency, State-Owned Bank,
State-Owned Policy Bank, State-Owned Commercial Bank, State-Owned
Company, State-Owned Fund, Intergovernmental Organization, Special
Purpose Vehicle/Joint Venture, Private Sector, NGO/CSO/Foundation,
Other, or No Organization Type Specified. Each security agent or
collateral agent is also categorized based on whether it is from the
financier country, the recipient country, or another country. The
Security organization type is preceded by one of three descriptors regarding the
Agent/Collateral country of origin: Chinese, Recipient, or Other (e.g. Chinese State-Owned
Agent Type Commercial Bank).
This field describes the nature of the collateral (security) that can be seized
Collateral in the event the borrower defaults on its debt repayment obligations.
This field captures the monetary amount that the funding agency
committed (or pledged) in its original currency of denomination. For
projects/activities that were at some point supported by an official
commitment (i.e., projects/activities with status designations of Pipeline:
Commitment, Implementation, Completed, Suspended, Canceled), this
field captures the original commitment amount. For projects/activities with
Amount (Original status designations of Pipeline: Pledge, this field captures the amount of
Currency) funding that was pledged.
This field captures that currency of denomination associated with the
monetary amount that the funding agency committed (or pledged), as
Original Currency recorded in the Amount (Original Currency) field.

31
 This marker designates whether AidData estimated the monetary amount
that the funding agency committed (or pledged), as captured in the
'Amount (Original Currency)' field. The field is set to "Yes" when the
'Amount (Original Currency)' field is estimated by AidData. It is otherwise
set to “No,” which indicates that AidData has reported the actual
monetary amount that the funding agency committed (or pledged) based
on explicit, official source documentation. There are a number of
circumstances under which AidData estimates transaction (financial
commitment) amounts. Examples include: (1) If the precise face value of a
Preferential Buyer's Credit (PBC) or Buyer's Credit Loan (BCL) from China
Eximbank is unknown, but the total cost of the commercial (EPC) contract
is known, AidData assumes that the face value of the PBC/BCL is
equivalent to 85% of the total commercial (EPC) contract cost; (2) If the
face value of a syndicated loan (involving one or more official sector
creditors from China) is known and the total number of participants in the
loan syndicate is known, AidData assumes that each bank provided equal
contributions to the syndicated loan; (3) If material is transferred in-kind
and there is no credible reporting on the monetary value of the in-kind
transfer, AidData calculates the monetary value of the in-kind materials by
multiplying the number of units of donated material by the market value
of those materials (in unit cost terms). Whenever a transaction (financial
commitment) amount has been estimated, AidData includes an
Amount Estimated explanation in the 'Description' and/or 'Staff Comments' field.
This field captures the monetary value of the official commitment (or
pledge) issued by the funding agency in constant 2021 U.S. dollars. To
calculate this value, AidData first converts the financial commitment (or
pledge) amount in its original currency of denomination to nominal U.S.
dollars at the average exchange rate in effect during the commitment (or
pledge) year, and then converts this amount to constant 2021 U.S. dollars
Amount (Constant using the OECD’s deflation methodology to adjust for inflation and ensure
USD 2021) comparability over time and space.
This field captures the monetary value of the official commitment (or
pledge) issued by the funding agency in nominal U.S. dollars. It is one of
the inputs used to calculate financial commitment (and pledge) amounts
Amount (Nominal in constant 2021 U.S. dollars, as recorded in the Amount (Constant USD
USD) 2021) field.

32
 This field captures the “adjusted” monetary amount that a funding agency
committed (or pledged) in its original currency of denomination. AidData
recommends using this field to calculate the cumulative stock of official
financial flows (ODA/OOF commitments) from China over multiple
years—when one or more of recipient countries secured "rollover"
emergency rescue loans and/or swap borrowings from the People's Bank
of China (PBOC) to refinance their maturing debts. For grants and
non-emergency loans, the amounts that are recorded in this field are
identical to the amounts that are recorded in the Amount (Original
Currency) field. However, for emergency rescue loans and swap
borrowings from the PBOC (with de jure maturities of one year or less),
this field excludes so-called “rollover” amounts that refinance maturing
debts. The monetary amounts in the Adjusted Amount (Original Currency)
field are calculated, whenever possible, by taking the difference between
the level of outstanding debt in the current year and the previous year.
This approach is consistent with the one used to derive net (new) PBOC
swap borrowings in the following publication: Horn, S., Parks, B., Reinhart,
C. M., and Trebesch, C. 2023. China as an International Lender of Last
Resort. NBER Working Paper No. 31105. Cambridge, MA: National
Bureau of Economic Research (NBER). In cases when this approach cannot
be applied but there is evidence of the same lender providing a series of
short-term emergency rescue loans (with identical face values and de jure
maturities of 1 year or less) to the same borrower that are repaid on their
original contractual maturity dates and subsequently reissued in
consecutive years, the Adjusted Amount (Original Currency) field records
the face value of the original loan commitment in the first year but not the
face values of the loan commitments in subsequent years. For
projects/activities that were at some point supported by an official
commitment (i.e., projects/activities with status designations of Pipeline:
Commitment, Implementation, Completed, Suspended, Canceled), the
Adjusted Amount (Original Currency) field captures the original
commitment amount. For projects/activities with status designations of
Adjusted Amount Pipeline: Pledge, the Adjusted Amount (Original Currency) captures the
(Original Currency) amount of funding that was pledged.

33
 This field captures the “adjusted” monetary amount of the official
commitment (or pledge) issued by the funding agency in constant 2021
U.S. dollars. To calculate this value, AidData first converts the “adjusted”
financial commitment (or pledge) amount in its original currency of
denomination—as recorded in the Adjusted Amount (Original Currency)
field—to nominal U.S. dollars at the average exchange rate in effect
during the commitment (or pledge) year. AidData then converts this
“adjusted” monetary amount to constant 2021 U.S. dollars using the
OECD’s deflation methodology to account for inflation and ensure
comparability over time and space. AidData recommends using the
Adjusted Amount (Constant USD 2021) field to calculate—in constant
2021 U.S. dollars—the cumulative stock of official financial flows
(ODA/OOF commitments) from China over multiple years—when one or
more of recipient countries secured "rollover" emergency rescue loans
and/or swap borrowings from the People's Bank of China (PBOC) to
refinance their maturing debts. For grants and non-emergency loans, the
amounts that are recorded in this field are identical to the amounts that
are recorded in the Amount (Constant USD 2021) field. However, for
emergency rescue loans and swap borrowings from the PBOC (with de
jure maturities of one year or less), this field excludes so-called “rollover”
amounts that refinance maturing debts. The monetary amounts in the
Adjusted Amount (Original Currency) field are calculated, whenever
possible, by taking the difference between the level of outstanding debt
in the current year and the previous year. This approach is consistent with
the one used to derive net (new) PBOC swap borrowings in the following
publication: Horn, S., Parks, B., Reinhart, C. M., and Trebesch, C. 2023.
China as an International Lender of Last Resort. NBER Working Paper No.
31105. Cambridge, MA: National Bureau of Economic Research (NBER). In
cases when this approach cannot be applied but there is evidence of the
same lender providing a series of short-term emergency rescue loans (with
identical face values and de jure maturities of 1 year or less) to the same
borrower that are repaid on their original contractual maturity dates and
subsequently reissued in consecutive years, the Adjusted Amount
Adjusted Amount (Original Currency) field records the face value of the original loan
(Constant USD commitment in the first year but not the face values of the loan
2021) commitments in subsequent years.

34
 This field captures the “adjusted” monetary amount of the official
commitment (or pledge) issued by the funding agency in nominal U.S.
dollars. It is one of the inputs used to calculate financial commitment (and
pledge) amounts in constant 2021 U.S. dollars, as recorded in the
Adjusted Amount (Constant USD 2021) field. AidData recommends using
the Adjusted Amount (Nominal USD) field to calculate—in nominal U.S.
dollars—the cumulative stock of official financial flows (ODA/OOF
commitments) from China over multiple years—when one or more of
recipient countries secured "rollover" emergency rescue loans and/or
swap borrowings from the People's Bank of China (PBOC) to refinance
their maturing debts. For grants and non-emergency loans, the amounts
that are recorded in this field are identical to the amounts that are
recorded in the Amount (Nominal USD) field. However, for emergency
rescue loans and swap borrowings from the PBOC (with de jure maturities
of one year or less), the Adjusted Amount (Nominal USD) field excludes
so-called “rollover” amounts that refinance maturing debts. The monetary
amounts in the Adjusted Amount (Original Currency) field are calculated,
whenever possible, by taking the difference between the level of
outstanding debt in the current year and the previous year. This approach
is consistent with the one used to derive net (new) PBOC swap borrowings
in the following publication: Horn, S., Parks, B., Reinhart, C. M., and
Trebesch, C. 2023. China as an International Lender of Last Resort. NBER
Working Paper No. 31105. Cambridge, MA: National Bureau of Economic
Research (NBER). In cases when this approach cannot be applied but
there is evidence of the same lender providing a series of short-term
emergency rescue loans (with identical face values and de jure maturities
of 1 year or less) to the same borrower that are repaid on their original
contractual maturity dates and subsequently reissued in consecutive years,
the Adjusted Amount (Original Currency) field records the face value of
Adjusted Amount the original loan commitment in the first year but not the face values of
(Nominal USD) the loan commitments in subsequent years.
This field identifies whether, for a given loan, there is an indication that the
borrower had difficulty repaying the loan or was financially distressed
during the loan’s repayment period (according to the project/transaction
life-cycle information that is identified in the description field). The field is
coded only for loans with status designations of Pipeline: Commitment,
Implementation, Completed, Suspended, and Canceled. Umbrella records
Financial Distress are not coded.

35
 This field seeks to capture the day on which an official financial
commitment (or official commitment to provide in-kind support) was
codified through the signing of a formal agreement by an official
donor/lender in China and one or more entities in a recipient country or a
set of recipient countries. Whenever possible, this field is based on the
precise calendar day on which the official commitment was made.
However, in cases when AidData is only able to identify the month and
year in which the formal agreement signed (e.g. May 2018), the
"Commitment Date" field is set to the first day of the month (01/01/2018).
In cases when AidData is only able to identify the year in which the formal
agreement was signed, the "Commitment Date" field is set to the first day
of the first month (e.g. 01/01/2018). In the event an official commitment
was made for a project/activity that entered implementation, but the
official commitment year is not identifiable, AidData records the first year
of project/activity implementation as a proxy for the official commitment
year. In the event an official commitment was made for a project/activity
that has not yet reached implementation, and the official commitment
year is not identifiable, AidData records the year in which the underlying
commercial contract (supported by the official commitment) was issued. If
this information is unavailable, AidData records the first year in which an
informal pledge was made as a proxy for the official commitment year. For
projects with a status designation of Pipeline Pledge (i.e. cases in which an
Commitment Date official commitment was not made), AidData records the date on which
(MM/DD/YYYY) the informal pledge was made.
For projects with a status designation of Pipeline: Commitment,
Implementation, Completion, Suspended, and Cancelled, this marker
designates whether AidData estimated the commitment date or reported
the actual date on which the official commitment was made. The field is
set to "Yes" when the "Commitment Date" field is estimated by AidData.
It is otherwise set to “No,” which indicates that AidData has reported the
actual commitment date. For projects/activities with status designations of
Pipeline: Pledge, this marker designates whether AidData estimated the
Commitment Date pledge date or reported the actual date on which the informal pledge was
Estimated made.
This field seeks to capture the day on which a project/activity supported
by an official financial (or in-kind) commitment from China was originally
scheduled to begin implementation. Whenever possible, this field is
based on the precise calendar day when the project/activity was originally
scheduled to begin implementation. However, in cases when AidData is
Planned only able to identify the month and year in which project/activity
Implementation implementation was scheduled to begin (e.g., May 2018), the “Planned
Start Date Implementation Start Date” field is set to the first day of the month (e.g.,
(MM/DD/YYYY) 05/01/2018).

36
 This field seeks to capture the day on which a project/activity supported
by an official financial (or in-kind) commitment from China began
implementation. Whenever possible, this field is based on the precise
calendar day when project/activity implementation began. However, in
cases when AidData is only able to identify the month and year in which
project/activity implementation began (e.g., May 2018), the “Actual
Implementation Start Date” field is set to the first day of the month (e.g.
05/01/2018). For projects/activities that involve the construction of
buildings or infrastructure, the “Actual Implementation Start Date” field
seeks to capture the first day of construction. In cases when the first day of
construction is unavailable but a proxy for the first day of construction
(e.g., the date on which a formal groundbreaking ceremony took place, a
project/activity commencement order was issued to the contractor
responsible for implementation, or a project/activity implementation
agreement was signed) can be identified, AidData records the proxy for
the first date of construction. For projects/activities that do not involve
construction but involve the provision of personnel, training,
analytical/advisory support, equipment, supplies, or commodities, the
“Actual Implementation Start Date” field captures the first day in which
some type of support was delivered to an entity (or set of entities) in the
recipient country. For projects/activities that only involve financial
transactions (cash donations, loans issued to shore up a country’s foreign
Actual exchange reserves, forgiveness or rescheduling of outstanding debts), the
Implementation “Actual Implementation Start Date” field captures the day on which the
Start Date first disbursement was made (or the day on which new terms and
(MM/DD/YYYY) conditions went into effect for a previously signed loan agreement).
This marker designates whether AidData estimated the implementation
start date or reported the actual date on which project/activity
Actual implementation began. The field is set to "Yes" when the 'Actual
Implementation Implementation Start Date' field is estimated by AidData. It is otherwise
Start Date set to “No,” which indicates that AidData has reported the actual
Estimated implementation start date.
This field captures the difference between the "Planned Implementation
Start Date" and the "Actual Implementation Start Date" when values are
Deviation from recorded for both variables. It captures the difference as the number of
Planned days, whereby positive values represent cases where the project/activity
Implementation started implementation ahead of schedule and negative values represent
Start Date cases where the project/activity started implementation behind schedule.
This field seeks to capture the day on which a project/activity supported
by an official financial (or in-kind) commitment from China was originally
scheduled to reach completion. Whenever possible, this field is based on
the precise calendar day when the project/activity was originally
scheduled to reach completion. However, in cases when AidData is only
able to identify the month and year in which a project/activity was
Planned scheduled to reach completion (e.g., May 2018), the “Planned
Completion Date Completion Start Date” field is set to the first day of the month (e.g.,
(MM/DD/YYYY) 05/01/2018).

37
 This field seeks to capture the day on which a project/activity supported
by an official financial (or in-kind) commitment from China was completed.
Whenever possible, this field is based on the precise calendar day when a
project/activity was completed. However, in cases when AidData is only
able to identify the month and year in which a project/activity was
completed (e.g., May 2018), the “Actual Completion Date” field is set to
the first day of the month (e.g., 05/01/2018). For projects/activities that
involve the construction of buildings or infrastructure, the “Actual
Completion Date” field seeks to capture the last day of construction. In
cases when the last day of construction is unavailable but a proxy for the
last day of construction (e.g., a road or railway is opened for use, a power
plant reaches its commercial operation date and begins selling electricity
to customers) is available, AidData records the proxy for the last day of
construction. For projects/activities that do not involve construction but
involve the provision of personnel, training, analytical/advisory support,
equipment, supplies, or commodities, the “Actual Completion Date” field
captures the last day on which some type of support was delivered to an
entity (or set of entities) in the recipient country. For projects/activities that
only involve financial transactions (cash donations, loans issued to shore
up foreign exchange reserves, forgiveness or rescheduling of outstanding
Actual Completion debts), the “Actual Completion Date” field captures the day on which the
Date last disbursement was made (or the day on which new terms and
(MM/DD/YYYY) conditions went into effect for a previously signed loan agreement).
This marker designates whether AidData estimated the project/activity
completion date or reported the actual date on which project/activity
implementation was completed. The field is set to "Yes" when the "Actual
Completion Date" field is estimated by AidData. It is otherwise set to
Actual Completion “No,” which indicates that AidData has reported the actual completion
Date Estimated date.
This field captures the difference between the "Planned Completion
Date" and the "Actual Completion Date" when values are recorded for
both variables. It captures the difference as the number of days, whereby
Deviation from positive values represent cases where the project/activity was completed
Planned ahead of schedule and negative values represent cases where the
Completion Date project/activity was completed behind schedule.
This field captures the total number of years it will take the borrower to
repay a loan, as specified in the original loan agreement. These de jure
maturity values are inclusive of grace periods. Users should keep in mind
that the rescheduling of a loan can result in a de facto maturity that is
substantially different from its de jure maturity. In cases when a loan’s
maturity is modified after an official commitment is issued, AidData
captures the maturity modification through a separate record in the
Maturity dataset that is given a flow type designation of “Debt Rescheduling.”

38
 This field captures the rate of interest (in percentage terms) that applies to
a loan, as specified in the original loan agreement. In cases when the
interest rate is tied to a floating rate such as LIBOR or EURIBOR, AidData
calculates the value of the floating rate in the month (or year) when the
official commitment was issued. Users should keep in mind that the
rescheduling of a loan can result in a de facto interest rate that is
substantially different from its de jure interest rate. In cases when a loan's
interest rate is modified after an official commitment is issued, AidData
captures the interest rate modification through a separate record in the
Interest Rate dataset that is given a flow type designation of “Debt Rescheduling.”
This field captures the number of years for which the borrower (receiving
agency) is not expected to make principal repayments to the creditor
(funding agency), as specified in the original loan agreement. Users should
keep in mind that the rescheduling of a loan can result in a de facto grace
period that is substantially different from its de jure grace period
(especially for short-term loans that are rolled over year over year). In
cases when a loan’s grace period is modified after an official commitment
is issued, AidData captures the grace period modification through a
separate record in the dataset that is given a flow type designation of
Grace Period “Debt Rescheduling.”
This field captures the management fee (in percentage terms) that applies
to the loan, as specified in the original loan agreement. A management
fee is a one-time, lump sum fee that is charged as a percentage of the
face value of the loan. In cases when a loan's management fee is modified
after an official commitment is issued, AidData captures the management
fee modification through a separate record in the dataset that is given a
Management Fee flow type designation of “Debt Rescheduling.”
This field captures the commitment fee (in percentage terms) that applies
to the loan, as specified in the original loan agreement. A commitment fee
is a fee that a borrower must pay to compensate the lender for its
commitment to lend; it is usually payable semi-annually and the size of the
fee is usually based on a fixed percentage of the undisbursed loan
amount. In cases when a loan’s commitment fee is modified after an
official commitment is issued, AidData captures the commitment fee
modification through a separate record in the dataset that is given a flow
Commitment Fee type designation of “Debt Rescheduling.”
This field captures the insurance fee (premium) that applies to the loan, as
specified in the original loan agreement. It is measured as a percentage of
total estimated debt service (i.e. the loan's principal plus total estimated
interest payments over the lifetime of the loan). The insurance
fee/premium is typically payable in a single lump sum, but in some cases it
Insurance Fee is payable in installments at different points in time or rolled into the
(Percent) principal of the loan.

39
 This field captures the nominal USD value of the insurance fee (premium)
that applies to the loan, as specified in the original loan agreement. The
nominal USD value of the insurance fee/premium is usually based on a
percentage of total estimated debt service (i.e. the loan's principal plus
total estimated interest payments over the lifetime of the loan). The
insurance fee/premium is typically payable in a single lump sum, but in
Insurance Fee some cases it is payable in installments at different points in time or rolled
(Nominal USD) into the principal of the loan.
This field captures the default (penalty) interest rate applied to the loan in
Default Interest the event of default (i.e., non-payment of principal, interest, or fees on
Rate their scheduled payment dates).
This field captures the date on which the first loan repayment should be
made by the borrower (as specified in the original loan agreement). It is
First Loan automatically calculated by adding the grace period to the commitment
Repayment Date date.
This field captures the date on which the final loan repayment should be
Last Loan made by the borrower (as specified in the original loan agreement). It is
Repayment Date automatically calculated by adding the maturity to the commitment date.
This field captures the grant element of the loan according to the OECD
cash-flow methodology, at the time that the original loan agreement was
signed. To calculate the grant element of a loan, which is a measure that
varies from 0 percent to 100 percent, AidData calculates the discounted
cost (or “net present value”) of the future debt service payments that will
be made by the borrower. This calculation requires information about the
loan’s face value, maturity length, grace period, and interest rate. When
AidData has access to the loan's face value, maturity length, grace period,
and interest rate, it uses the OECD’s grant element calculator from the
cash-flow methodology (assuming a fixed, 10 percent discount rate, two
repayments per year, and equal principal repayments). In theory, a grant
element calculator can generate values above 100% or below 0%.
However, AidData bounds grant element values so that they cannot
assume values that exceed 100% or negative values (since negative values
imply lending terms that are “less favorable than market terms,” which
Grant Element does not logically make sense because market terms are risk-adjusted
(OECD Cash-Flow) prices agreed to by willing buyers and sellers of credit).

40
 This field captures the grant element of the loan according to the OECD
grant-equivalent methodology, at the time that the original loan
agreement was signed. To calculate the grant element of a loan, which is a
measure that varies from 0 percent to 100 percent, AidData calculates the
discounted cost (or “net present value”) of the future debt service
payments that will be made by the borrower. This calculation requires
information about the loan’s face value, maturity length, grace period, and
interest rate. When AidData has access to the loan's face value, maturity
length, grace period, and interest rate, it uses the OECD’s grant element
calculator from the grant-equivalent methodology, assuming (a) a fixed
discount rate which depends on the recipient country income level (9% for
LDCs and other LICs and 6% for UMICs), (b) two repayments per year, and
(c) equal principal repayments. In theory, a grant element calculator can
generate values above 100% or below 0%. However, AidData bounds
grant element values so that they cannot assume values that exceed 100%
or negative values (since negative values imply lending terms that are
Grant Element “less favorable than market terms,” which is does not logically make sense
(OECD because market terms are risk-adjusted prices agreed to by willing buyers
Grant-Equiv) and sellers of credit).
This field captures the grant element of the loan according to the current
(post-2013) World Bank/IMF methodology, at the time that the original
loan agreement was signed. To calculate the grant element of a loan,
which is a measure that varies from 0 percent to 100 percent, AidData
calculates the discounted cost (or “net present value”) of the future debt
service payments that will be made by the borrower. This calculation
requires information about the loan’s face value, maturity length, grace
period, and interest rate. When AidData has access to the loan's face
value, maturity length, grace period, and interest rate, it uses the IMF’s
post-2013 grant element calculator (assuming a fixed, 5 percent discount
rate, two repayments per year, and equal principal repayments). In theory,
a grant element calculator can generate values above 100% or below 0%.
However, AidData bounds grant element values so that they cannot
assume values that exceed 100% or negative values (since negative values
imply lending terms that are “less favorable than market terms,” which
Grant Element does not logically make sense if market terms are risk-adjusted prices
(IMF) agreed to by willing buyers and sellers of credit).
This field captures whether the loan record has one or more lenders by
way of two categories: Bilateral Loan or Syndicated/Club Loan. A bilateral
loan is issued by one lender to a single borrower. A syndicated loan or
Number of club loan is issued by a consortium ('syndicate' or 'club') of lenders to a
Lenders single borrower.
This field provides a marker of whether it is known that the loan record is
classifiable as an Export Buyer's Credit, which is a loan that is issued by
Chinese state-owned policy banks and Chinese state-owned commercial
banks to overseas borrowing institutions to facilitate their acquisition of
Export Buyer's goods/services from a Chinese supplier. The field is set to "Yes" if the loan
Credit record is classifiable as an Export Buyer's Credit.

41
 This field provides a marker of whether it is known that the loan record is
classifiable as either an Export Seller's Credit or a Supplier's Credit. An
Export Seller's Credit is a loan issued by a Chinese state-owned bank to a
Chinese company for the purpose of increasing its exports. The proceeds
of export seller's credits are to be used by borrowers (Chinese exporters)
to finance their foreign sales. Chinese exporters usually secure export
seller's credits when they need liquidity to offer a supplier's credit to an
overseas buyer. If a Chinese company extends a loan to a borrower and
the borrower is expected to use the loan proceeds to purchase goods and
services from that Chinese company, then the loan is a supplier's credit
Supplier’s (also known as a seller's credit or vendor financing). The field is set to
Credit/Export "Yes" if the loan record is classifiable as either Export Seller's Credit or
Seller’s Credit Supplier's Credit.
This field provides a marker of whether it is known that the loan record is
classifiable as an Interest-Free Loan, which is a loan that is issued to a
borrower without any interest accruing. The borrower is only responsible
for repaying the loan's principal amount. The field is set to "Yes" if the
Interest-Free Loan loan record is classifiable as an Interest-Free Loan.
This field provides a marker of whether it is known that the loan record
captures debt refinancing, which is a new loan for the purpose of repaying
one or more existing loans/debts. The field is set to "Yes" if the loan is
Refinancing used for debt refinancing.
This field provides a marker of whether it is known that the loan record is
classifiable as an Investment Project Loan, which is a loan that is provided
to finance the provision of goods, works, or services to support a public or
Investment Project private investment project. The field is set to "Yes" if the loan record is
Loan classifiable as an Investment Project Loan.
This field provides a marker of whether it is known that the loan record is
classifiable as a Mergers and Acquisitions (M&A) Loan, which is a loan that
is issued to a borrower to facilitate its acquisition of an equity stake in a
company and/or to facilitate the consolidation of multiple companies (i.e.,
a merger). The field is set to "Yes" if the loan record is classifiable as a
M&A M&A Loan.
This field provides a marker of whether it is known that the loan record is
classifiable as a Working Capital Loan, which is a loan that provides funds
for a borrower's day-to-day operations but not for making capital
investments or facilitating the acquisition of long-term assets. The field is
Working Capital set to "Yes" if the loan record is classifiable as a Working Capital Loan.
This field provides a marker of whether it is known that the loan record
involves an Engineering, Procurement and Construction Plus Finance
(EPC+F or EPCF) Agreement arrangement. In a typical EPC+F
arrangement, a project owner in the host country has selected a Chinese
company as its engineering, procurement, and construction (EPC)
contractor, and a Chinese bank issues a loan to that EPC contractor but
with a sovereign guarantee from the host government. The field is set to
EPCF "Yes" if the loan record involves an EPC+F arrangement.

42
 This field provides a marker of whether it is known that the loan record
captures a lease agreement. A lease is a contractual arrangement calling
for the lessee (user) to pay the lessor (owner) for use of an asset. The
lessor is the legal owner of the asset, while the lessee obtains the right to
use the asset in return for regular rental payments. Under a capital lease (a
financial arrangement where the lessee/borrower uses an asset and pays
regular installments plus interest to the lender/lessor), rental payments are
usually classified as interest and obligation payments, similarly to a
mortgage (with the interest calculated each rental period on the
outstanding obligation balance). The field is set to "Yes" if the loan record
Lease captures a lease agreement.
This field provides a marker of whether it is known that the loan record
captures a borrowing under a Foreign Currency Swap Line (FXSL) or a
Balance of Payments (BoP) Loan. An FXSL agreement is an agreement
between the central banks of two countries to exchange cash flows in
different currencies at predetermined rates over a specified period of
time. Central banks participate in these agreements to (a) facilitate
bilateral trade settlements using their national currencies (rather than
relying upon a third-party currency such as the U.S. dollar), (b) manage
demands from their local banks, and (c) provide liquidity to support
financial market stability. The party that draws down on the swap line
becomes the borrower and the other party becomes lender. During the
term of the swap, the party that draws down on the swap line makes
either fixed or floating interest payments on the principal amount. If both
parties draw down on the swap line, then both parties exchange fixed or
floating interest payments on the principal amounts. A Balance of
Payments (BoP) Loan, Liquidity Support Facility (LSF), or Foreign Currency
Deposit Loan, is a loan issued by a Chinese state-owned policy bank, a
Chinese state-owned commercial bank, or China’s State Administration of
Foreign Exchange (SAFE) to a central bank or finance ministry in another
country that explicitly authorizes the borrower to use the proceeds of the
loan to (a) shore up foreign exchange reserves, (b) repay existing debts,
and/or (b) finance general budgetary expenditures. The field is set to
FXSL/BOP "Yes" if the loan record captures a FXSL borrowing or a BoP Loan.
This field provides a marker of whether it is known that the loan record
captures a Cross-Currency Interest Rate Swap. A cross-currency interest
rate swap is an off-balance sheet way of hedging against interest rate risk
and foreign exchange risk. In a typical cross-currency interest rate swap
agreement, both parties to the transaction are simultaneously lending to
each other. That is to say, each party is both a lender and a borrower. The
field is set to "Yes" if the loan record captures a Cross-Currency Interest
CC IRS Rate Swap.

43
 This field provides a marker of whether it is known that the loan record
involves a Revolving Credit Facility (RCF) arrangement. In a typical RCF
arrangement, the lender commits funding up to a certain level, but unlike
a "term loan" (that is repaid in regular payments over a set period of
time), the borrower can draw down, repay, and redraw on an
irregular/as-needed basis. It provides liquidity for day-to-day operations,
and the borrower is charged an annual commitment fee on unused
amounts (a "facility fee"). The field is set to "Yes" if the loan record
RCF involves a RCF arrangement.
This field provides a marker of whether it is known that the loan record is
classifiable as a Government Concessional Loan (GCL), which is an
RMB-denominated loan that the Export-Import Bank of China (China
Eximbank) issues to government institutions on below-market terms
(typically 20-year maturities, 5-year grace periods, and 2% interest rates)
to facilitate their acquisition of goods/services from a Chinese supplier.
The proceeds of a GCL can be used by government borrowing institutions
to finance up to 100% of the total cost of a commercial contract with a
Chinese supplier. The field is set to "Yes" if the loan record is classifiable
GCL as a GCL.
This field provides a marker of whether it is known that the loan record is
classifiable as a Preferential (Export) Buyer's Credit (PBC), which is a
USD-denominated or EUR-denominated loan that the Export-Import Bank
of China (China Eximbank) issues to government institutions to facilitate
their acquisition of goods/services from a Chinese supplier. The borrowing
terms of these loans vary, but they are offered with fixed rather than
floating (market) interest rates (such as LIBOR or EURIBOR), which are
usually more generous than prevailing market rates. China Eximbank has a
policy of allowing borrowers to use PBC proceeds to finance 85% of the
total cost of a commercial contract with a Chinese supplier. China
Eximbank usually requires that the remaining 15% of the commercial
contract cost be financed with "counterpart funding" from the borrowing
institution. The field is set to "Yes" if the loan record is classifiable as a
PBC PBC.
This field provides a marker of whether it is known that the loan record
involves a Pre-Export Financing (PxF) or Commodity Prepayment
Financing arrangement, which is an arrangement in which a commodity
(e.g. oil) producer gets up-front cash from a customer in return for a
promise to repay the customer with that commodity (possibly at a
discount) in the future. PxF funds may be advanced by a lender or
syndicate of lenders to a commodity producer to assist the company in
meeting either its working capital needs (for example, to cover the
purchase of raw materials and costs associated with processing, storage
and transport) or its capital investment needs (for example, investment in
plant and machinery and other elements of infrastructure). The field is set
PxF/Commodity to “Yes” if the loan record involves a PxF or Commodity Prepayment
Prepayment Financing arrangement.

44
 This field provides a marker of whether it is known that the loan record is
classifiable as an Inter-Bank Loan, which is a loan issued by one bank
(lender) to another bank (borrower). All inter-bank loans are by nature
on-lending arrangements, and will be captured as such in the
“On-Lending” field. The field is set to "Yes" if the loan record is
Inter-Bank Loan classifiable as an Inter-Bank Loan.
This field provides a marker of whether it is known that the loan record is
classifiable as an Overseas Project Contracting Loan, which is a loan
issued by the Export-Import Bank of China (China Eximbank) to a Chinese
company to help it finance an overseas project contract. This loan can be
denominated in USD or RMB. Per China Eximbank policy, the contract
cost that is financed with the loan should not be lower than 1 million USD,
and goods and services exported from China under the contract should
Overseas Project not be lower than 15% of contract cost. The field is set to "Yes" if the loan
Contracting Loan record is classifiable as an Overseas Project Contracting Loan.
This field provides a marker of whether it is known that the loan record
involves a Deferred Payment Agreement (DPA) arrangement. In a typical
DPA arrangement, the Chinese company that the project owner in the
host country has selected as its engineering, procurement, and
construction (EPC) contractor is also a lender to the project owner. The
Chinese company assigns receivables under its EPC contract with the
project owner to one or more Chinese banks. Upon assignment of
receivables, the Chinese bank or banks will release funds to the Chinese
company so it can discharge its obligations under the DPA as a lender.
DPA The field is set to "Yes" if the loan record involves a DPA arrangement.
This field provides a marker of whether it is known that the loan record
involves a Non-Recourse or Limited-Recourse Project Finance transaction.
When a project is financed with a limited-recourse or non-recourse
structure, the loan that is used to finance the acquisition, construction,
and/or maintenance of an asset—such as a toll road, a seaport, or an
electricity grid—is exclusively repaid with the cash flow generated by the
asset (e.g., toll revenue, container fees, or electricity sales), and the
creditor either has no claim (“recourse”) or a limited claim to any other
assets as a basis for recovering the debt. In a standard, limited-recourse or
non-recourse project finance transaction, a creditor lends to an
independent legal entity that is established for the express purpose of
developing, owning, and operating a specific project. This entity is often
called a special purpose vehicle (SPV) because it is only allowed to
engage in activities that relate to a specific purpose (project), and it is
legally prohibited from incurring debts or obligations that are not related
to that purpose (project). The field is set to "Yes" if the loan record
Project Finance involves a Non-Recourse or Limited-Recourse Project Finance transaction.

45
 This field provides a marker of whether it is known that the loan involves a
multilateral (inter-governmental) organization in one or more of the
following capacities: as (a) a loan administrator (such as the Africa Growing
Together Fund or the China Co-Financing Fund for Latin America and
Caribbean), (b) a co-financier (through a syndicated loan, club loan, or
parallel co-financing arrangement), (c) an insurer (such as the World Bank
Group's Multilateral Investment Guarantee Agency), and/or (d) a financial
Involving or technical adviser. The field is set to "Yes" if the record involves a
Multilateral multilateral organization.
This field provides a marker of whether it is known that the loan involves
Involving co-financing agencies which are not of Chinese origin. The field is set to
Non-Chinese "Yes" if there is at least one co-financing agency from a country other than
Financier China.
This field provides a marker of whether it is known that the loan is
classifiable as a short-term loan. The field is automatically set to "Yes" if
the loan's de jure maturity is 1 year or less. Users should keep in mind that
loans with de jure maturities of 1 year or less may be "rolled over" or
Short-Term otherwise rescheduled, resulting in longer de facto maturities.
This field provides a marker for rescue loan records. In the 3.0 version of
the dataset, rescue lending (also known as bailout lending) is defined as
any loan that allows a sovereign debtor to (i) service existing debts, (ii)
finance general budgetary expenditures and/or (iii) shore up foreign
reserves. Any loan in the dataset that meets at least one of two criteria is
designated as a rescue loan: (1) any loan where the FXSL/BOP marker is
checked, and (2) any loan where AidData's Sector Code and Sector Name
fields are set to 510 and General Budget Support, respectively However,
in the time period covered by the dataset, loans to two sovereign debtors
that met the first criterion (PBOC swap line borrowings by Malaysia’s
central bank and Thailand’s central bank) are not classified as rescue
lending, as evidence shows these debtors did not utilize their foreign
currency swap lines with the PBOC during periods of macroeconomic
distress. The central banks of Malaysia and Thailand likely used the foreign
currency swap lines for trade and investment purposes. See Horn et al.
(2023) at
https://docs.aiddata.org/ad4/pdfs/WPS124_China_as_an_International_Le
Rescue nder_of_Last_Resort.pdf for more details.

46
 This field captures the extent of host government ownership of the Joint
Venture/Special Purpose Vehicle (JV/SPV)—in cases where the JV/SPV is
recorded as a receiving agency (borrowing institution) for the loan. Each
JV/SPV is assigned to one of four categories: Majority Host
Government-Owned, Minority Host Government-Owned, No Host
Government Ownership, or No Ownership Information Available. The
value in this field is set based on the following criteria: (1) It is classified as
"Majority Host Government-Owned" if government agencies,
state-owned companies or state-owned banks from the recipient country
have a combined JV/SPV ownership stake greater than 50%; (2) It is
classified as "Minority Host Government-Owned" if government agencies,
state-owned companies or state-owned banks from the recipient country
have a combined JV/SPV ownership stake that is greater than 0% but less
than or equal to 50%; (3) It is classified as "No Host Government
Ownership" if government agencies, state-owned companies or
state-owned banks from the recipient country hold no stake in the JV/SPV;
JV/SPV Host and (4) Lastly, if information is not available to determine the host
Government government ownership stake, then it is classified as "No Ownership
Ownership Information Available."
This field captures the extent of Chinese government ownership of the
Joint Venture/Special Purpose Vehicle (JV/SPV)—in cases where the
JV/SPV is recorded as a receiving agency (borrowing institution) for a loan.
Each JV/SPV is assigned to one of four categories: Majority Chinese
Government-Owned, Minority Chinese Government-Owned, No Chinese
Government Ownership, and No Ownership Information Available. The
value in this field is set based on the following criteria: (1) It is classified as
"Majority Chinese Government-Owned" if government agencies,
state-owned companies or state-owned banks from China have a
combined JV/SPV ownership stake greater than 50%; (2) It is classified as
"Minority Chinese Government-Owned" if government agencies,
state-owned companies or state-owned banks from China have a
combined JV/SPV ownership stake that is greater than 0% but less than or
equal to 50%; (3) It is classified as "No Chinese Government Ownership"
if government agencies, state-owned companies or state-owned banks
JV/SPV Chinese from China hold no stake in the JV/SPV; and (4) Lastly, if information is not
Government available to determine the Chinese government ownership stake, then it is
Ownership classified as "No Ownership Information Available."

47
 This field captures the extent to which the host government may
eventually be liable for debt repayment. Each loan record is assigned to
one of six categories: Central government debt, Central
government-guaranteed debt, Other public sector debt, Potential public
sector debt, Private debt, or Unallocable. The value in this field is
hierarchically and automatically determined based on the following
criteria: (1) The loan record is classified as "Central government debt" if it
is an official sector loan to a central government institution in the recipient
country, measured by whether there is at least one receiving agency
(direct or indirect) from the recipient country that is classified as a
government agency; (2) If the loan record does not meet the first (1)
criterion, it is classified as "Central government-guaranteed debt" if it is
an official sector loan to a state-owned entity (e.g. state-owned enterprise
and state-owned bank) or privately-owned entity in the recipient country
that benefits from a sovereign (central government) repayment
guaranteei; (3) If the loan record does not meet the first (1) criterion or the
second (2) criterion, it is classified as "Other public sector debt" if (a) it is
an official sector loan to a state-owned entity (such as a city/municipal
government, a state-owned bank, or a state-owned enterprise) in the
recipient country that does not benefit from a sovereign (central
government) repayment guarantee; (b) it is an official sector loan to a
private entity or state-owned entity in the recipient country that is backed
by a repayment guarantee from a state-owned entity other than the
central government in the recipient country (such as a city/municipal
government, a state-owned bank, or a state-owned enterprise), OR (c) it is
an official sector loan to a special purpose vehicle (SPV) or joint venture
(JV) that is majority-owned by one or more public sector institutions in the
recipient country and that does not benefit from a sovereign (central
government) repayment guarantee or a repayment guarantee from a
state-owned entity other than the central government in the recipient
country (such as a city/municipal government, a state-owned bank, or a
state-owned enterprise). (4) If the loan record does not meet the first (1)
criterion, the second (2) criterion, or the third (3) criterion, it is classified as
"Potential public sector debt" if it is an official sector loan to a special
purpose vehicle (SPV) or joint venture (JV) borrower that is minority-owned
by one or more public sector institutions in the recipient country and that
does not benefit from a sovereign (central government) repayment
guarantee or a repayment guarantee from a state-owned entity other than
the central government in the recipient country (such as a city/municipal
government, a state-owned bank, or a state-owned enterprise). (5) If the
loan record does not meet the first (1) criterion, the second (2) criterion,
the third (3) criterion, and the fourth (4) criterion, it is classified as "Private
debt" if it is an official sector loan to a privately-owned entity that does
not benefit from a repayment guarantee from a public sector institution in
the recipient country (this includes lending to a private entity, or lending
to a Joint Venture or Special Purpose Vehicle with no level of host
Level of Public government ownership (i.e. the "JV/SPV Host Government Ownership"
Liability variable is set to "No Host Government Ownership"; (6) If the loan record

48
 does not meet the first (1) criterion, the second (2) criterion, the third (3)
criterion, the fourth (4) criterion, or the fifth (5) criterion, then it is classified
as "Unallocable" due to a lack of information.
This field provides a count of the total number of sources used to create
Total Source Count the project/activity record (including official and other source types).
This field provides a count of the total number of official sources used to
create the project/activity record. Official source types include
Donor/Recipient Official Source, Implementing/Intermediary Organization
Official Source Source, and Other Official Source (non-Donor, non-Recipient,
Count non-Implementing).
This field provides URLs to the sources that were used to create the
Source URLs project/activity record. The entries are pipe-delimited
This field provides the titles of the source articles, reports, and websites
Source Titles used to create the project/activity record. The entries are pipe-delimited.
This field provides the names of the publishers of the source articles,
reports, and websites used to create the project/activity record. The
Source Publishers entries are pipe-delimited.
This field identifies the type of sources that were used to create the
Source Type project/activity record. The entries are pipe-delimited.
This field records the names of the people who were involved in the
project/activity and/or the financial (or in-kind) transfer for the
Contact Name project/activity whenever this information is available.
This field records the position titles of the people who were involved in
the project/activity and/or the financial (or in-kind) transfer for the
Contact Position project/activity whenever this information is available.
This field designates whether the recipient country was eligible for ODA
(based on income level and OECD DAC categorizations) in the year that
the official commitment was issued. For projects/activities with a status
designation of "Pipeline: Pledge," this field designates whether the
ODA Eligible recipient country was eligible for ODA (based on income level and OECD
Recipient DAC categorizations) in the year that the informal pledge was issued.
This field provides the income status of the recipient country for each
project/activity at the time that it secured a Chinese ODA or OOF
commitment (or pledge). The classification is based on the OECD's ODA
Eligibility lists. It records whether a country is low income (LIC), lower
middle income (LMIC), upper middle income (UMIC), or high income
OECD ODA (HIC). High-income countries are not eligible for ODA flows in the OECD's
Income Group classification scheme.

49
 This field provides a description of the locations of project activities.
Whenever possible, AidData captures geographical information that
makes it possible to identify (i) the precise physical boundaries and exact
locations of buildings and facilities (e.g. schools, hospitals, stadiums,
government buildings, power plants, and factories); (ii) the precise
geographical scope of special economic zones, industrial parks, mining
concessions, protected areas, and plots of land under cultivation; and (iii)
the exact routes of linear infrastructure (e.g., roads, bridges, tunnels,
railways, power lines, canals, and pipelines). Whenever possible, AidData
also records OpenStreetMap and GoogleMaps URLs that capture the
geographical locations and features of projects. An important caveat is
that AidData is only able to provide provide precise details for the subset
of projects in the dataset that have physical footprints (e.g. roads,
railways, transmission lines) or involve activities at specific locations (e.g.
medical teams stationed at a given hospital, equipment given to park
rangers to patrol a protected area). Less precise location information may
also be recorded for projects, such as the general area or administrative
Location Narrative zone associated with a project.
This field indicates the finest level of geographical precision AidData was
able to pinpoint for each project feature (or set of features). When
identifying OSM features captured in the 'Location Narrative' field, coders
seek to identify features at the highest level of precision as possible.
Given data limitations though, precise locations are not always available.
This field indicates the finest level of precision the geographical
information achieved. If multiple features were captured for one project,
Geographic Level this reports the highest level of precision achieved across the project's
of Precision features. The level of precision values includes the following categories:
Available precise, approximate (5km buffer), or ADM-level.
This field is marked "Yes" for projects where there are associated
sub-national locations at the ADM1 level (a first-order administrative
division such as a province, state, or governorate) available in the
associated GCDF_3.0_ADM1_Locations.csv file. ADM1 locations are only
made available when the project was geocoded at the ADM1 level or
finer. These ADM1 locations are constructed by AidData by using the
OpenStreetMap URLs that are recorded in the "Location Narrative" field
and seek to represent the geographical locations of each project where
available (see the GCDF3.0 ADM Files README included in the GCDF 3.0
file download). AidData has made the full dataset of geospatial features,
as well as usage tips and related documentation accessible via
ADM1 Level https://www.aiddata.org/data/aiddatas-geospatial-global-chinese-develop
Available ment-finance-dataset-version-3-0.

50
 This field is marked "Yes" for projects where there are associated
sub-national locations at the ADM2 level (a second-order administration
division such as a district, municipality, or commune) available in the
associated GCDF_3.0_ADM2_Locations.csv. ADM2 locations are only
made available when the project was geocoded at the ADM2 level or
finer. These ADM2 locations are constructed by AidData by using the
OpenStreetMap URLs that are recorded in the “Location Narrative” field
and seek to represent the geographical locations of each project where
available (see the GCDF3.0 ADM Files README included in the GCDF 3.0
file download). AidData has made the full dataset of geospatial features,
as well as usage tips and related documentation accessible via
ADM2 Level https://www.aiddata.org/data/aiddatas-geospatial-global-chinese-develop
Available ment-finance-dataset-version-3-0.
This field is marked "Yes" for projects where there is a geospatial feature
available in Version 3.0 of AidData's Geospatial Global Chinese
Development Finance Dataset (Goodman et al., 2023). These geospatial
features were constructed by AidData by using the OpenStreetMap URLs
that are recorded in the "Location Narrative" field and seek to represent
the geographical locations of each project where available. AidData has
made the complete set of geospatial features along with usage tips and
related documentation accessible via
Geospatial Feature https://www.aiddata.org/data/aiddatas-geospatial-global-chinese-develop
Available ment-finance-dataset-version-3-0.

This metric varies on a scale of 1 to 5, with 1 indicating that the

Source Quality project/activity record is exclusively underpinned by unofficial sources and
Score 5 indicating reliance upon multiple, official sources.
This metric varies on a scale of 0 to 5, with 5 indicating that the basic
fields of the project/activity record are complete. The "threshold" for a
score of 5 is similar to the key fields in the OECD-DAC’s Creditor
Reporting System: an actual rather than estimated commitment year, a
Data non-missing transaction amount, a flow type/flow class that is not defined
Completeness as "Vague," and identifiable funding, implementing, and receiving
Score agencies.
This metric varies on a scale of 0 to 5, with higher scores indicating that
more implementation details have been captured in the project/activity
record. The following implementation details are considered: whether the
implementing agency (or agencies), implementation start and completion
dates (actual or planned), and geographical locations of the
project/activity are specified; and whether the project/activity has a
specified sector allocation. Project Implementation Scores are only
calculated for project/activity records with a “Recommended for
Implementation Aggregates” value of "True" and a “Status” value of "Implementation" or
Detail Score “Completion.

51
 This metric varies on a scale of 0-5, with higher values indicating that more
financial transaction details are captured in the project/activity record.
Loan Detail Scores are only calculated for project/activity records with a
"Recommended for Aggregates" value of "True" and a "Flow Type"
designation of “Loan." A score of 5 indicates that a loan’s interest rate,
maturity, transaction value, loan type, funding agencies, and receiving
Loan Detail Score agencies are all specified (i.e., not missing).

Section 2 - Capturing Chinese ODA and OOF

With the 3.0 version of the TUFF methodology, we seek to identify overseas projects/activities
supported by financial or in-kind transfers from official sector institutions in China (i.e., Chinese
government and state-owned institutions). With respect to temporal coverage, we aim to
identify all ​projects/activities backed by official commitments that took place between 2000
and 2021, with details on the timing of implementation over a 24-year period (2000-2023).
With respect to spatial coverage, we aim to capture projects/activities in every low-income,
lower-middle income, and upper-middle income country and territory across every major world
region, including Africa, Asia, Oceania, the Middle East, Latin America and the Caribbean, and
Central and Eastern Europe. In total, AidData’s GCDF Dataset, Version 3.0 covers 165
countries: 147 countries where systematic searches were undertaken and Chinese
government-financed projects/activities were identified and 18 countries where systematic
searches were undertaken but no Chinese government-financed projects/activities were
identified.

Our goal is to capture a comprehensive and detailed picture of projects/activities backed by

Chinese ODA and OOF but not Chinese Official Investment (see Appendix A). Consistent with
OECD definitions, we use the terms “Chinese Official Finance” or “Official Financial Flows from
China” to refer to financial or in-kind transfers (or “flows”) from official sector institutions in
China (i.e., Chinese government and state-owned institutions) to another country or territory.
Official Finance (Official Financial Flows) consists of Official Development Assistance (ODA),
Other Official Flows (OOF), and Official Investment. However, AidData only systematically
tracks and publishes data on Chinese ODA and OOF. It does not systematically track and
publish data on Chinese Official Investment. In the remainder of this paper, we document our
efforts to define and measure Chinese ODA and OOF in support of the construction of
AidData’s GCDF Dataset, Version 3.0.

The challenge of measuring ODA and OOF from China is not only capturing the full range of
flows, but also classifying these flows accurately and in ways that make comparisons between
different financiers valid. To make the study of financial and in-kind transfers from China more
comparable with those from OECD-DAC donors and creditors, the 3.0 version of the TUFF
methodology uses OECD definitions and measurement criteria, as outlined in the OECD-DAC
Directives. We use these standards to classify each project according to its source of financing,
type of financing, intent, and level of concessionality.

52
2.1 - Measuring Concessionality and Intent

As part of its data collection and classification system, AidData designates each financial and
in-kind transfer (“flow”) from an official sector institution as Official Development Assistance
(ODA) or Other Official Flows (OOF). The OECD’s Development Assistance Committee (DAC)
has used these designations since 1972 to distinguish between flows from official sector
institutions that (a) are provided on concessional terms and that promote and specifically target
the economic development and welfare of developing countries (ODA), and (b) are provided
on non-concessional terms or do not specifically target the economic development and welfare
of developing countries (OOF). The sum of ODA and OOF is sometimes referred to as Official
Financial Flows, Official Financing, or Overseas Development Finance. Many DAC countries,
non-DAC countries, and multilateral institutions report the volume and composition of their
official sector flows according to these categories and criteria. In alignment with the
OECD-DAC’s own definitions, AidData classifies each project/activity record in the 3.0 dataset
as either “ODA-like” or “OOF-like.” This unique feature of the 3.0 dataset sets it apart from
other publicly available datasets that measure Chinese development finance in that it allows
analysts to make “apples-to-apples” comparisons of Chinese development finance and other
international sources of development finance (that report their ODA and OOF flow data to the
OECD-DAC).

The criteria for whether a flow qualifies as ODA or OOF is determined by the OECD-DAC. It is
based on (1) the intent of the flow (whether its primary intent was development or not), (2) the
income classification of the receiving country, and (3) the concessionality level of the flow.16 All
grants and in-kind transfers are treated as concessional. However, a “grant element” measure is
used to calculate the concessionality level of all loans. This measure, which varies from 0
percent to 100 percent, seeks to capture the generosity of a loan—or the extent to which it is
priced below market rates. In principle, any loan provided on entirely non-concessional terms
should have a grant element of 0 percent.

While the first two criteria have remained consistent since the concept of ODA was introduced
more than five decades ago, the OECD-DAC recently made changes to the third
(concessionality) criterion. Until 2017, a loan from an official sector institution to a low-income
or middle-income country had to meet a concessionality (grant element) threshold of 25% to
qualify as ODA. However, in 2018, the OECD-DAC introduced a tiered system of discount rates
and concessionality thresholds based on the income classifications of borrower countries and
whether borrowing institutions are official sector or private sector institutions. The 2018
definition of concessionality is based on the following criteria:

● For loans to official sector institutions, the following concessionality thresholds apply:
○ Least-developed countries and low-income countries: a minimum grant element
of 45% (calculated using a 9% discount rate).

16
An additional criteria is that the flow must be provided by official agencies, including state and local
governments or their executive agencies. AidData’s GCDF 3.0 only tracks official Chinese agencies, so
this criteria is always met.

53
 ○ Lower-middle income countries: a minimum grant element of 15% (calculated
using a 7% discount rate).
○ Upper-middle income countries: a minimum grant element of 10% (calculated
using a discount rate of 6%).
● For loans to private sector institutions, the OECD-DAC maintains the pre-2018
definition of concessionality and requires a grant element of at least 25% (that is
calculated using a 10% discount rate).17

To ensure comparability between the flows documented in the 3.0 version of the GCDF dataset
and the flow data published by the OECD-DAC, AidData has applied these definitions in the
following manner:

Intent: AidData codes the intent of each financial and in-kind transfer (“flow”). Flows with
“development intent” are those that are primarily oriented toward the promotion of economic
development and welfare in the recipient country. Flows with “commercial intent” are those
that primarily seek to promote the commercial interests of the country from which the financial
transfer has originated (e.g., encouraging the export of Chinese goods and services). Flows
with “representational intent” are those that primarily seek to promote a bilateral relationship
with another country or otherwise promote the language, culture, or values of the country from
which the financial transfer has originated (e.g., the establishment of a Confucius Institute or
Chinese cultural center). Flows with “military intent” are those that seek to promote the security
interests of the country from which the financial transfer originates or strengthen the lethal
force capabilities of military institutions in the recipient country.

ODA Income Classification: AidData reports the income classification group of the borrowing
country. Flows to countries not eligible for ODA are automatically assigned to the “OOF-like”
category.

Concessionality:

● For flows committed between 2000 and 2017, a flow is classified as “ODA-like” when it
(1) has development intent, (2) has a grant element of at least 25% (using a 10%
discount rate), and (3) supports a country that is ODA-eligible according to the
OECD-DAC’s ODA income classification list.
● For flows committed between 2018 and 2021, a flow is classified as “ODA-like” when it
(1) has development intent, (2) has a concessionality level that meets the new criteria
(established in 2018 definition), and (3) supports country that is ODA-eligible according
to the OECD-DAC’s ODA income classification list.

By definition, any international official sector flows not classified as ODA-like are classified as
OOF-like. The OOF-like flows in the 3.0 version of AidData’s GCDF dataset largely consist of
export credits and non-concessional loans.

17
According to the OECD, the method for calculating the ODA grant equivalent for loans to private
sector institutions has not yet been formalized, and discussions to do so are currently ongoing at the
OECD-DAC. Until an agreement has been formalized, the pre-2018 concessionality definition still
applies.

54
In some cases, we are not able to determine if an international official sector flow would qualify
as ODA or OOF because of insufficiently detailed information in source documentation. In such
cases, the flow in question is categorized as Vague (Official Finance).

2.2 - Measuring Emergency Rescue Loans and the Cumulative

Stock of Official Financial Flows from China to LICs and MICs
As explained at greater length in AidData’s Belt and Road Reboot report, emergency rescue
loans represent an increasingly important part of China’s overseas portfolio of loans to LICs and
MICs (Parks et al. 2023). Nearly all of these borrowings, which are typically used to refinance
maturing debts, carry de jure maturities of one year or less (i.e., they are initially scheduled for
repayment in 12 months or less). However, it is not unusual for financially-distressed LICs and
MICs to receive short-term emergency rescue loans from the same Chinese creditor in a series
of consecutive years (Horn et al. 2023). So-called “rollover” emergency rescue loans come in
two varieties: (1) those that reach their original contractual maturity dates and secure final
maturity date extensions; and (2) those that are repaid on their original contractual maturity
dates and reissued (with similar or different face values and borrowing terms) and assigned new
maturity dates.18 However, among serial recipients of short-term emergency rescue loans, it is
seldom possible—with publicly available sources of information—to differentiate between
those who had their final maturity dates extended and those who fully repaid on their original
contractual maturity dates but were reissued new loans.

This relatively new feature of China’s overseas lending program raises an important question
about how to accurately estimate the cumulative stock of official financial flows—or lending
commitments—from China to LICs and MICs. Neither the OECD's Creditor Reporting System
(CRS) nor the World Bank’s Debtor Reporting System (DRS) ask lenders or borrowers to disclose
loans with maturities of one year or less.19 However, most of China’s short-term emergency
rescue loans have de facto maturities that substantially exceed one year (Horn et al. 2023),
which makes it difficult to justify the exclusion of all emergency rescue loans from stock- or
flow-based measures of official financial commitments (or lending commitments) from China to
LICs and MICs.20

At the same time, rollover debt presents an overcounting risk because it straddles a fine line
between new lending commitments and maturity extensions of existing lending commitments.
This risk is particularly relevant to estimations of the cumulative stock of official financial flows
18
The 3.0 version of AidData’s GCDF dataset captures the full range of China’s international rescue
lending operations. Parks et al. (2023) demonstrate that an increasing proportion of China's official sector
lending to LICs and MICs consisted of “rollover” emergency rescue loans during the early BRI period of
2014-2017 (8%) and the late BRI period of 2018-2021 (34%).
19
The reporting directives of the OECD's Creditor Reporting System (CRS) specify that "[l]oans with a
maturity of one year or less are not reportable in DAC statistics" (OECD 2021: 51). Similarly,
governments that participate in the World Bank’s Debtor Reporting System (DRS) are asked to report
their long-term debt repayment obligations to external creditors on an annual basis. Long-term debt is
defined in the DRS reporting manual as debt “with an original contractual or extended maturity of more
than one year […]” (World Bank 2000: 4).
20
Central banks that borrow from the PBOC frequently see their final maturity dates extended—or they
repeatedly receive short-term loans to refinance maturing debts. Horn et al. (2023) provide evidence that
the de facto maturity of the average PBOC swap line borrowing is 3.5 years.

55
(or lending commitments) from China. In order to address this challenge, the 3.0 version of
AidData’s GCDF dataset includes three new variables (fields) that measure transaction amounts
without including any rollover amounts from PBOC swap line borrowings or emergency rescue
loans from other creditors (with maturities of one year or less).21 These amounts are reported in
their original currencies of denomination, nominal USD, and constant 2021 USD via the
"Adjusted Amount (Original Currency),” "Adjusted Amount (Constant USD 2021)," and
"Adjusted Amount (Nominal USD)" variables.

Users of the 3.0 version of AidData’s GCDF dataset can estimate “rollover” loan amounts (in
their original currencies of denomination) by subtracting the values in the Adjusted Amount
(Original Currency) field from the values in the Amount (Original Currency) field. Nominal USD
“rollover” loan amounts can be estimated by subtracting the values in the Adjusted Amount
(Nominal USD) field from the values in the Amount (Nominal USD) field. Constant 2021 USD
“rollover” loan amounts can be estimated by subtracting the values in the Adjusted Amount
(Constant USD 2021) field from the values in the Amount (Constant USD 2021) field.

Additionally, the 3.0 version of the GCDF dataset includes a new variable (the “Rescue” field)
that identifies emergency rescue loans from official sector lenders in China. Consistent with the
method of measurement that was first introduced in Horn et al. (2023), this variable captures
any loan that allows a sovereign debtor to (1) service existing debts, (2) finance general
budgetary expenditures and/or (3) shore up foreign reserves. Any loan in the dataset that
meets at least one of two criteria is designated as a rescue loan: (1) any loan where the
FXSL/BOP marker is checked, and (2) any loan where AidData's Sector Code and Sector Name
fields are set to 510 and General Budget Support, respectively. However, during the time
period covered by the 3.0 version of the GCDF dataset (commitment years 2000-2021), loans
to two sovereign debtors that met the first criterion (PBOC swap line borrowings by Malaysia’s
central bank and Thailand’s central bank) are not classified as rescue lending, as evidence
shows these debtors did not utilize their foreign currency swap lines with the PBOC during
periods of macroeconomic distress (Horn et al. 2023). The central banks of Malaysia and
Thailand likely used the foreign currency swap lines for trade and investment purposes.

2.3 - Categorizing Chinese Lending to Different Types of

Borrowers
In the 3.0 version of the GCDF dataset, we have introduced new variables that identify the
extent of host government ownership (in the “JV/SPV Host Government Ownership” field) and
Chinese government ownership (in the “JV/SPV Chinese Government Ownership” field) of the

21
Whenever possible, for each emergency rescue loan (PBOC swap borrowing) of the rollover variety, we
calculate a transaction amount that excludes the rollover amount by taking the difference between the
level of outstanding debt in the current year and the previous year. This approach is consistent with the
one taken by Horn et al. (2023) to derive net (new) PBOC swap borrowings. In cases when this approach
cannot be applied but there is evidence of the same lender providing a series of short-term emergency
rescue loans (with identical face values and de jure maturities of 1 year or less) to the same borrower that
are repaid on their original contractual maturity dates and subsequently reissued in consecutive years,
we record the face value of the original loan commitment in the first year but not the face values of the
loan commitments in subsequent years.

56
JV/SPV in order to facilitate more analysis of (actual and potential) loan repayment obligations
in cases when the borrower (direct receiving agency) is a JV/SPV. This field is automatically
generated at the organization-level with four options:

1. Majority Host Government-Owned: The sum of all owners for the JV/SPV where origin
= recipient and type = government agency, state-owned company, and/or state-owned
bank is more than 50%.
2. Minority Host Government-Owned: The sum of all owners for the JV/SPV where origin
= recipient and type = government agency, state-owned company, and/or state-owned
bank is between .01-49.99%.
3. No Host Government Ownership: If the owner organization type for any owners of the
SPVs does not involve a host government type (e.g origin = recipient and type =
government agency, state-owned company, and/or state-owned bank).
4. No Ownership Information Available: If the ownership information = “No Information
Available.”

At the project/activity-level, one value is shown for this variable. If one JV/SPV is involved, then
the category assigned to that JV/SPV is presented. When multiple JV/SPVs are identified as
receiving agencies, the ownership category is assigned at the project/activity-level that is
highest among the JV/SPVs -- e.g., if one is “Majority Host Government-Owned” and one is
“No host government ownership," then the “Majority Host Government-Owned” designation
would be assigned for that specific project/activity.

The “Level of Public Liability” field in the 3.0 version of AidData’s GCDF dataset captures the
extent to which the host government may eventually be liable for debt repayment. It is
hierarchically and automatically determined based on the following criteria:

1. The loan record is classified as "Central government debt" if it is an official sector loan
to a central government institution in the recipient country, measured by whether there
is at least one receiving agency (direct or indirect) from the recipient country that is
classified as a government agency;
2. If the loan record does not meet the first (1) criterion, it is classified as "Central
government-guaranteed debt" if it is an official sector loan to a state-owned entity (e.g.,
state-owned enterprise and state-owned bank) or privately-owned entity in the recipient
country that benefits from a sovereign (central government) repayment guarantee;
3. If the loan record does not meet the first (1) criterion or the second (2) criterion, it is
classified as "Other public sector debt" if (a) it is an official sector loan to a state-owned
entity (such as a city/municipal government, a state-owned bank, or a state-owned
enterprise) in the recipient country that does not benefit from a sovereign (central
government) repayment guarantee; (b) it is an official sector loan to a private entity or
state-owned entity in the recipient country that is backed by a repayment guarantee
from a state-owned entity other than the central government in the recipient country
(such as a city/municipal government, a state-owned bank, or a state-owned enterprise),
OR (c) it is an official sector loan to a special purpose vehicle (SPV) or joint venture (JV)
that is majority-owned by one or more public sector institutions in the recipient country
and that does not benefit from a sovereign (central government) repayment guarantee
or a repayment guarantee from a state-owned entity other than the central government
in the recipient country (such as a city/municipal government, a state-owned bank, or a
state-owned enterprise).

57
 4. If the loan record does not meet the first (1) criterion, the second (2) criterion, or the
third (3) criterion, it is classified as "Potential public sector debt" if it is an official sector
loan to a special purpose vehicle (SPV) or joint venture (JV) borrower that is
minority-owned by one or more public sector institutions in the recipient country and
that does not benefit from a sovereign (central government) repayment guarantee or a
repayment guarantee from a state-owned entity other than the central government in
the recipient country (such as a city/municipal government, a state-owned bank, or a
state-owned enterprise).
5. If the loan record does not meet the first (1) criterion, the second (2) criterion, the third
(3) criterion, and the fourth (4) criterion, it is classified as "Private debt" if it is an official
sector loan to a privately-owned entity that does not benefit from a repayment
guarantee from a public sector institution in the recipient country (this includes lending
to a private entity, or lending to a Joint Venture or Special Purpose Vehicle with no level
of host government ownership (i.e., the "JV/SPV Host Government Ownership" variable
is set to "No Host Government Ownership";
6. If the loan record does not meet the first (1) criterion, the second (2) criterion, the third
(3) criterion, the fourth (4) criterion, or the fifth (5) criterion, then it is classified as
"Unallocable" due to a lack of information.

2.4 - Identifying when China’s Borrowers are Experiencing

Financial Distress

The 3.0 version of AidData’s GCDF dataset includes a “Financial Distress” flag that identifies
whether, for a given loan, there is an indication that the borrower had difficulty repaying the
loan or was financially distressed during the loan’s originally scheduled repayment period
(according to the project/transaction life-cycle information that is identified in the description
field). This field is set to “Yes” for loans that showed signs of distress (within their originally
scheduled repayment periods). Examples of distress include the borrower accruing principal or
interest arrears, defaulting on its repayment obligations, experiencing bankruptcy, or
seeking/securing a rescheduling of the loan’s repayment terms. Other examples include
Sinosure making indemnity payments under the loan’s insurance policy or lower-than-expected
levels of revenue generation from the project/activity funded by the loan.22

22
For some types of analysis in the Belt and Road Reboot report, Parks et al. (2023) modifies the financial
distress measure to more clearly differentiate between repayment risks and repayment risk mitigation
efforts. Instead of using all loan records where the “Financial Distress” variable is set to “Yes,” they
exclude all observations for which the only source of evidence of the borrower having difficulty making
repayments or experiencing financial distress is an attempted or actual debt rescheduling.

58
2.5 - Chinese ODA and OOF Agencies and Instruments

2.5.1 - Sources of Chinese ODA and OOF

2.5.1.1 - Overview
The OECD defines “Official Financing” as “transactions undertaken by the official sector (i.e.
Government) at their own risk and responsibility, regardless of the source of funds (taxation of
or borrowing from the private sector). Official agencies include federal, state and local
departments and agencies.”23 The OECD also considers autonomous and semi-autonomous
state-owned entities—like KfW, the German state-owned investment and development
bank—to be official sector institutions. Therefore, the 3.0 version of the TUFF methodology
seeks to capture ODA and OOF from all Chinese government and state-owned entities,
including central government agencies (like the Ministry of Commerce, the Ministry of Foreign
Affairs, and the Ministry of Agriculture), regional and local government agencies (like
Chongqing Municipal Health Commission and Tianjin Municipal Government), state-owned
enterprises (like CNPC, CMEC, CATIC, and CRBC), state-owned policy banks (like China
Development Bank and China Eximbank), state-owned commercial banks (like ICBC, BoC, and
CCB),24 and state-owned funds (like the Silk Road Fund).25 The specific agency or set of
agencies that provide financial or in-kind support is captured in the “Funding Agency” field.
Each funding agency is also assigned to one of six “Funding Agency Type” categories:
Government Agency, State-Owned Policy Bank, State-Owned Commercial Bank, State-Owned
Bank, State-Owned Company, and State-Owned Fund.

23
See OECD factsheet at ​https://www.sheffield.ac.uk/polopoly_fs/1.659262!/file/Is_it_ODA.pdf.
24
AidData classifies the following institutions as Chinese state-owned commercial banks: China
Construction Bank Corporation (CCB), Industrial and Commercial Bank of China (ICBC), Bank of China
(BOC), China Bank of Communications (BoCom or BoComm), Agricultural Bank of China, Postal Savings
Bank of China (PSBC), China Bohai Bank, Bank of Shanghai, China CITIC Bank, China Merchants Bank,
Huaxia Bank Co., Ltd., and China Everbright Bank Co., Ltd. This group of banks includes so-called
shareholding commercial banks that are subsidiaries of state-owned enterprises (e.g., China CITIC Bank)
and city commercial banks (i.e., Bank of Shanghai).
25
We consider institutions to be “state-owned” if the government has the largest ownership stake
compared to all other owners. Due to a lack of agreement about whether Huawei Technologies Co., Ltd.
(“Huawei”) should be treated as an official sector institution, we do not include projects financed with aid
or debt from Huawei or any of its subsidiaries in the 3.0 version of AidData’s Chinese Global
Development Finance Dataset. AidData published a separate Global Huawei Finance Dataset in
September 2021 that captures 153 Huawei-financed projects worth $1.4 billion in 64 countries from
commitment years 2000-2017.

59
Figure 1: Chinese Official Sector Agencies

2.5.1.2 - China’s Ministry of Commerce (MOFCOM)

MOFCOM is the lead administrator of the country’s interest-free (or “zero-interest”) loan and
grant program for developing countries.26 While there are many different Chinese government
institutions that provide small-scale grants and donations, MOFCOM is the primary Chinese
government institution responsible for providing large-scale, RMB-denominated grants to host
government institutions that support the construction, maintenance, upgrading, or expansion
of infrastructure and other physical assets (like schools, hospitals, convention centers, and
government buildings).27

As part of its outreach to other countries, MOFCOM officials often meet with government
counterparts in developing countries and sign Economic and Technical Cooperation
Agreements (ETCAs, in Chinese: 经济技术合作协议). When MOFCOM signs an ETCA with a
foreign government, it is issuing an official grant or interest-free loan commitment. The
interest-free loan commitments that are issued via ETCAs are typically denominated in RMB
with the following borrowing terms: 20 year maturities, 10 year grace periods, and 0% interest

26
In August 2021, China International Development Agency (CIDCA), MOFCOM, and the Ministry of
Foreign Affairs (MOFA) reviewed and approved a new set of foreign aid administration measures. These
measures specify that, as of October 1, 2021, CIDCA will be responsible for all planning, policymaking,
regulatory, and supervisory functions that support the country’s foreign aid program. MOFCOM will
continue to implement foreign aid projects, among other line ministries (including MOFA). See
http://www.cidca.gov.cn/2021-08/31/c_1211351312.htm
27
An illustrative MOFCOM grant agreement can be accessed here:
​https://www.dropbox.com/s/fv965ko40q88pp7/01.12.2020.%20ENG.pdf?dl=0. MOFCOM grants and
interest-free loans usually support projects with development intent, although there are some cases
when it finances projects with representational, commercial, or military intent.

60
rates.28 Counterpart funding is not required, and when borrowers have difficulty repaying their
debts to the Chinese government, these are often the first loans to be forgiven or rescheduled
(Morris et al. 2020).

ETCAs are often signed with recipient governments for unspecified purposes (or generically
worded purposes, such as economic development or disaster relief and reconstruction). Then,
a bank account is set up between a Chinese state-owned bank (usually China Development
Bank or Bank of China) and the recipient country’s central bank or finance ministry to facilitate
the transfer of funds. In many cases, a bank/loan agreement and/or “letters of exchange” are
subsequently signed (sometimes multiple years after the original ETCA was signed). However,
not every ETCA is implemented in this way. Sometimes, the ETCA, the bank account
agreement, and letters of exchange are simultaneously approved. The vast majority of Chinese
government grants and interest-free loans are funded through ETCAs, so AidData coders are
informed that if a project is financed by a Chinese government grant or interest-free loan, there
is a high likelihood that the project was funded through an ETCA (and supplemental online
searches will most likely be necessary to determine if that project was funded through an ETCA
and the specific date when the ETCA was signed).29

Although ETCAs represent official financial commitments, AidData coders are instructed to
treat them as “umbrella” agreements—meaning that ETCAs are effectively framework
agreements that govern follow-on grants and zero-interest loans for specific projects, but those
projects are not specified at the time of ETCA signing. Also, recipient governments often sign
multiple ETCAs over time, which means that one of the most challenging tasks associated with
ETCAs is accurately tracking all of the subsidiary projects that they have funded.30 AidData
coders are instructed to track down the exact ETCA that is financing a project if it is reported
that the project was financed through an ETCA.

AidData has several coding guidelines that are specific to ETCAs:

● The funding agency for an ETCA that has committed a grant or interest-free loan should
always be coded as “China Ministry of Commerce.” When a funding agency is not
explicitly identified in the source materials, AidData coders are instructed to assign
“China Ministry of Commerce” as the funding agency if the following criteria are met:
(1) The grant or zero-interest loan is denominated in RMB, (2) the receiving agency is a
government institution from the developing country, (3) the project has development
intent, (4) the project involves the provision of large-scale funding for infrastructure or a
physical asset. Chinese government grant- and loan-financed project activities/events
28
An illustrative MOFCOM loan agreement can be accessed here:
https://www.documentcloud.org/documents/20485643-cmr_2011_518. Based on the OECD-DAC
concessionality calculator, these loans usually have a grant element of approximately 75% (Malik 2021).
29
In many cases, MOFCOM’s support for a specific project that is financed with the grant or loan
proceeds from previously signed ETCAs will be codified in “letters of exchange,” an “exchange of
notes,” or an “implementation agreement.” For illustrative financing agreements that followed the
signing of ETCAs, see ​https://www.dropbox.com/s/fv965ko40q88pp7/01.12.2020.%20ENG.pdf?dl=0
and
https://www.dropbox.com/s/t86s480xiqyvzih/Supplemental%20Implementation%20Agreement%20for%
20China%20Aided%20Ministry%20of%20Foreign%20Affairs%20Construction%20Project.pdf?dl=0
30
This is especially true because (a) ETCAs and subsidiary agreements are often agreed upon in different
years, and (b) it is not always immediately obvious that a project was funded by a particular ETCA (which
itself may have been signed years ago).

61
 that involve a MOFCOM representative are treated by AidData coders as evidence that
MOFCOM is the funding agency. Additionally, when the underlying source materials
used to construct project descriptions refer to MOFCOM as issuing project design or
implementation contracts to Chinese firms or MOFCOM deploying personnel to
conduct on-site project inspections or post-project evaluations, AidData coders are
instructed to treat this as a evidence that MOFCOM is the funding agency.
● The ETCA should be explicitly identified by its full name in both the title and the project
description. An ETCA should not be referred to as just “an agreement.” AidData coders
are instructed to differentiate ETCAs by specifying the year of signing of an individual
ETCA in the project title. They are also instructed to specify the exact date of the ETCA
signing in the description field if that information is available (as sometimes multiple
ETCAs are signed in a given year).
● The commitment date of an ETCA should be the date on which the ETCA was signed
and countersigned, and not the date on which the ETCA was ratified by the recipient
government’s legislature.
● When an ETCA is issued for an unspecified purpose at the time of its signing, AidData
coders are instructed to adhere to the following guidelines:
○ It should always be coded as an umbrella record.
○ The status of the ETCA should be coded as an official commitment and not as a
pledge.
○ If it is found that the funds committed through the ETCA supported multiple,
subsidiary projects, those projects should be created as separate records, and
linked back to the ETCA umbrella record.
○ If it is found that the funds committed through the ETCA supported a single
project, then the ETCA record should not be marked as an umbrella record, and
all of the other project fields should be populated/updated.
○ When specific projects are identified at the time of the ETCA signing, AidData coders
are instructed to adhere to the following guidelines:
○ If multiple projects are specified, then AidData coders are instructed to create
one umbrella record for the ETCA, and one record for each of the specified
projects. Umbrella records should be status-coded as official commitments.
○ If there is only one specific project receiving the full amount of funding that was
committed through the ETCA, then AidData coders are instructed to create only
one record (non-umbrella), with a description that specifies that the financial
commitment for the project came from an ETCA.

2.5.1.3 - Export-Import Bank of China

The Export-Import Bank of China (or “China Eximbank”) is one of two state-owned policy banks
in China that provide overseas financing. It is designated as China’s official export credit agency
and recognized as an official bilateral creditor. China Eximbank is also unique in that it is
responsible for the implementation of the “two preferential loans” program, which consists of
Government Concessional Loans (GCLs) and the Preferential Buyer’s Credits (PBCs). These are
sometimes referred to in Chinese as 两优贷款. In addition to the loans provided through this
program, AidData captures Buyer’s Credit Loans (BCLs), Overseas Investment Loans, and
Overseas Project Contracting Loans from China Eximbank. These lending instruments are
described in greater detail below with the specific guidelines that AidData coders use to
classify projects that are financed with these instruments.

62
Government Concessional Loans (GCLs): The GCL (in Chinese: 优惠贷款) is a loan that China
Eximbank issues to foreign governments maintaining diplomatic ties with China.31 These
RMB-denominated loans are granted on below-market terms (typically 20-year maturities,
5-year grace periods, and 2% interest rates).32 China’s Ministry of Finance calculates the
difference between the interest rates attached to these loans and the central bank’s benchmark
rate and reimburses China Eximbank accordingly.33 GCL proceeds can be used by borrowing
institutions to finance up to 100% of the total cost of a commercial contract with a Chinese
supplier. China Eximbank does not expect the borrowing institution to provide any
“counterpart funding.” The Chinese government characterizes the GCL as a form of ODA.
Similar to preferential buyer’s credits, GCLs will usually be explicitly identified as such in official
sources.34

AidData has several coding guidelines that are specific to GCLs:

● Currency: The currency of denomination should be represented in RMB, unless the only
information about the transaction amount is denominated in USD (or another currency).
● Flow Type: All GCLs should be coded as loans.
● GCL Flag: With the new loan categorization scheme introduced in TUFF 3.0, all GCLs
are flagged as such in the corresponding “GCL” field.
● Intent: AidData coders are instructed to categorize the intent variable according to the
primary purpose of the project being financed with a GCL. Even though the Chinese
government refers to GCLs as foreign aid or ODA, GCL-financed projects may be
coded as having development, commercial, representational, military, or mixed intent.
In cases where GCL-financed projects are coded as having commercial,
representational, military, or mixed intent, they are not given a flow class designation of
ODA-like.
● Title and Description: If a project is financed with a GCL, AidData coders are instructed
to make this clear in the project title and description. That is to say, if a GCL agreement
was signed, it should be described as such and not simply referred to as a “loan
agreement.”

Preferential Buyer’s Credit (PBC) and Nonpreferential Buyer’s Credit Loan (BCL) Program: PBCs
(In Chinese: 优惠出口买方信贷) are USD-denominated loans that are granted to foreign
government institutions.35 When China Eximbank issues a PBC, it provides a loan to a foreign
government institution (in a country that maintains diplomatic ties with China) and that
government institution uses the loan proceeds to buy goods or services from a Chinese
supplier. The borrowing terms of these loans vary, but they are offered with fixed rather than
floating interest rates that are usually more generous than prevailing market rates.36 China
Eximbank has a policy of allowing borrowers to use PBC proceeds to finance 85% of the total
cost of a commercial contract (often an Engineering, Procurement, and Construction contract)
31
An illustrative GCL can be found accessed here:
https://www.documentcloud.org/documents/20485597-cmr_2011_172
32
See Morris et al. (2020) and Export-Import Bank of China (n.d.).
33
See
​https://www.dropbox.com/s/ctvqu1jmopny6vv/392125599-Key-Points-of-Evaluation-pptx.pdf?dl=0
34
For example, the loan agreement ID number may contain the abbreviation “GCL” (referring to
Government Concessional Loan)—e.g., CHINA EXIMBANK GCL NO.1 (2011) TOTAL NO. (351).
35
An illustrative PBC agreement can be found accessed here:
https://www.documentcloud.org/documents/20488747-phl_2018_422
36
See Morris et al. (2020) and Export-Import Bank of China (n.d.).

63
with a Chinese supplier. China Eximbank usually requires that the remaining 15% of the
commercial contract cost be financed with “counterpart funding” from the borrowing
institution. There are cases when China Eximbank deviates from this norm (e.g., by allowing a
borrower to use up to 95% of the proceeds of a PBC to finance a commercial contract), but
most PBCs adhere to this policy. Similar to GCLs, PBCs will usually be explicitly identified as
such in official sources.37

China Eximbank also has a (non-preferential) buyer’s credit loan (BCL) program that shares
many of the same features as the PBC program.38 However, BCLs can be denominated in USD
or EUR; they are usually priced at a floating market interest rate (LIBOR or EURIBOR) plus a
margin; they often have shorter maturity lengths and grace periods than PBCs; and the
borrowers need not be government institutions.39

AidData has several coding guidelines that are specific to PBCs and BCLs:
● Transaction amount: If the precise face value of the PBC or BCL is unknown but the total
cost of the commercial (EPC) contract is known, AidData coders are instructed to
assume that the face value of the PBC/BCL is equivalent to 85% of the total EPC cost.
AidData coders are also instructed to note any such assumptions were made in the
“Description” field or “Staff Comments” field.
● Flow Type: All PBCs and BCLs should be coded as loans.
● Export Buyer’s Credit Flag: PBCs and BCLs are both flagged as export buyer’s credit in
the corresponding field.
● PBC Flag: With the new loan categorization scheme introduced in TUFF 3.0, all
preferential buyer’s credit records are flagged as such in the corresponding “PBC” field.
For all PBC records, the “Export Buyer’s Credit” field will automatically be set to “Yes.”
● Currency: Eximbank PBCs are exclusively denominated in USD, so AidData coders are
instructed to populate the amount field and the currency field accordingly.
● Intent: PBCs and BCLs are trade promotion instruments. As such, projects financed with
PBCs and BCLs should never be given a development intent designation. The possible
intent values for PBCs and BCLs (as well as export seller’s credits) are mixed or
commercial. If a PBC/BCL is sufficiently concessional and it is financing a project that
seeks to improve economic development or welfare in the recipient country, AidData
coders are instructed to code it as having mixed intent (i.e., both commercial intent and
development intent). However, in cases when a PBC or BCL supports a project that only
has commercial intent (for example, a loan to help a shipping company acquire vessels
that will allow it to move ocean containers from country to country, a loan to help a
company finance its general operations, or a loan to help a company service its existing
debts), AidData coders are instructed to designate the project as having commercial
intent.

37
For example, a PBC agreement ID number may contain the abbreviation “PBC” (referring to
Preferential Buyer’s Credit)—e.g., CHINA EXIMBANK PBC NO. (2016) 33 TOTAL NO. (421). Buyer’s
credit loans are usually referred to with the abbreviation “BCL” in official sources. However, among
buyer’s credit loans that were issued by China Eximbank in the early 2000s, it is not uncommon to see
the abbreviation “BLA” in official sources.
38
See https://www.dropbox.com/s/mlg5lz5fqnh8aef/China%20Eximbank%20Pitch%20Deck.pdf?dl=0
39
An illustrative BCL agreement can be found accessed here:
https://www.documentcloud.org/documents/20488172-ecu_2010_444

64
 ● Flow Class: All export credits, PBCs and BCLs issued by China Eximbank, should be
assigned to the flow class category of OOF-like.
● Title and Description: If a project is financed with a PBC or BCL, AidData coders are
instructed to make this clear in the project title and description. That is to say, if a PBC
or BCL agreement was signed, it should be described as such and not simply referred
to as a “loan agreement” or an “export credit agreement.”

Overseas Investment Loans Program: Overseas Investment Loans (In Chinese: 境外投资贷款)
are RMB and foreign-currency denominated loans issued by China Eximbank to support
Chinese enterprises’ overseas investments.40 The proceeds of these loans can be used to fund
acquisitions, fixed asset investments, and overseas equity investments approved by Chinese
authorities.41 These loans can also be used for working capital needs and be used to cover fees
associated with overseas investments. The face value of an overseas investment loan can cover
up to 70% of total contract value.

AidData has several coding guidelines that are specific to to Overseas Investment Loans:
● Flow Type: The flow type field should be coded based on the nature of financing (debt),
not how loan proceeds are used (equity/investment). As such, AidData coders are
instructed to assign these loans to the flow type category of “loan” (rather than “FDI”).
These loans enable equity investments (FDI), but they are not themselves FDI.
● M&A Flag: All Overseas Investment Loans should be flagged as Mergers and
Acquisitions (M&A) Loans in the corresponding “M&A” field.
● Intent: The intent of Overseas Investment Loans should always be coded commercial
since they enable commercial investments.
● Title and Description: If a project is financed with an Overseas Investment Loan,
AidData coders are instructed to make this clear in the project title and description.
That is to say, if an Overseas Investment Loan agreement was signed, it should be
described as such and not simply referred to as a “loan agreement.”
● Flow Class: All Overseas Investment Loans should be assigned to the flow class
category of OOF-like.

Overseas Project Contracting Loans (In Chinese: 对外承包工程贷款): These loans are provided
by China Eximbank to help Chinese companies finance overseas project contracts.42 They can
be denominated in USD or RMB. Per China Eximbank policy, the contract cost that is financed
by the loan should not be lower than 1 million USD. Goods and services exported from China
under the contract should not be lower than 15% of contract cost.

AidData has several coding guidelines that are specific to Overseas Project Contracting Loans:
● Flow Type: The flow type field should always be set to loan.
● Overseas Project Contracting Loan Flag: With the new loan categorization scheme
introduced in the 3.0 version of the TUFF methodology, all Overseas Project
Contracting Loans are flagged as such in the corresponding field.
● Intent: The intent of Overseas Project Contracting Loans should be coded as mixed
when they seek to facilitate the export of Chinese goods and services and promote

40
See the following website for more information about Overseas Investment Loans:
http://english.eximbank.gov.cn/Business/CreditB/SupportingCrossBI/201810/t20181016_6967.html
41
See ​https://www.dropbox.com/s/r5xhu7zdqiebbn3/2.EXIM-Bank.pptx?dl=0
42
See https://www.dropbox.com/s/r5xhu7zdqiebbn3/2.EXIM-Bank.pptx?dl=0

65
 economic development or welfare in the recipient country through the project that is
being financed. The intent of Overseas Project Contracting loans should be coded as
commercial if the loans only seek to facilitate the export of Chinese goods and services.
● Title and Description: If a project is financed with an Overseas Project Contracting Loan,
AidData coders are instructed to make this clear in the title field and description field.
That is to say, if an Overseas Project Contracting Loan agreement was signed, it should
be described as such and not simply referred to as a “loan agreement.”
● Flow Class: All Overseas Project Contracting Loans should be assigned to the flow class
category of OOF-like.

In addition to the types of loans that we have described, China Eximbank provides a range of
other loans for various purposes (some concessional, some not). These include export seller’s
credits, which are loans to a Chinese company that the Chinese company may on-lend to a
buyer/borrower who wishes to buy goods or services from that company. These types of loans
from China Eximbank can be denominated in local or foreign currency.43

2.5.1.4 - China Development Bank (CDB)

CDB is one of two state-owned policy banks in China that provides overseas financing. It has a
wide array of lending instruments, including but not limited to term loans, bridge loans,
revolving credit facilities, working capital loans, commodity-backed loans, club loans,
syndicated loans, and buyer’s credits. Its RMB-denominated and foreign-currency denominated
loans are generally provided on less concessional terms than China Eximbank loans because,
unlike China Eximbank, CDB must maintain its own balance sheets and lend without receiving
official subsidies from the state.44 Typically, the base interest rate on a CDB loan is tethered to
the (floating) London Interbank Offered Rate (LIBOR) or Euro Interbank Offered Rate (Euribor),
with an additional margin incorporated to account for borrower-specific risk and repayment
capacity (Morris et al. 2020). While “all-in” interest rates on CDB loans usually fall somewhere
in the 4.5% to 6% range, maturities and grace periods can vary widely. Loans from CDB are
granted to both government agencies and companies. Debt collateralization is often required
by the bank as a way to limit repayment risk. Whereas 29% of China Eximbank’s overseas
lending portfolio is collateralized, 70% of CDB’s overseas lending portfolio is collateralized
(Malik et al. 2021). China Development Bank also engages in inter-bank lending far more
frequently than China Eximbank.45

CDB is characterized by the Chinese government as a bank that follows commercial lending
practices. However, using the OECD’s grant element calculator, AidData has found that some
of its loans do qualify as concessional loans.46 CDB’s lending portfolio has also become
significantly concessional over time (see Malik et al. 2021).

AidData has several coding guidelines that are specific to China Development Bank loans:

43
This means that export seller’s credits may be denominated in RMB. See Section 2.5.3.3 for more
guidance regarding how export seller’s credits are coded.
44
An illustrative CDB loan agreement can be found at
https://www.documentcloud.org/documents/20488181-ecu_2010_462_1_of_2
45
Inter-bank lending involves one bank providing a loan to another, typically with higher interest rates
and short maturities.
46
See Section 2.1 for a detailed description on how AidData determines loan concessionality.

66
 ● Intent: AidData codes the intent of all loans according to the purpose of the project
that is being financed. Even though the Chinese government claims that CDB follows
commercial lending practices, loans from CDB can be coded as having development,
commercial, representational, military, or mixed intent, depending on the primary
purpose of the project that is being financed. In cases where loans from CDB are
identified as having development intent and a grant element which meets the
concessionality threshold described in the OECD’s guidelines, AidData coders are
instructed to make a flow class designation of ODA-like.
● ETCAs: As the bank accounts that are used to disburse ETCA funds are often opened
with CDB, there is a risk of CDB being identified as the funding agency responsible for
providing grants or interest-free loans through ETCAs. This is incorrect. AidData coders
are therefore instructed to consistently identify MOFCOM as the funding agency
responsible for providing grants or interest-free loans through ETCAs.

2.5.1.5 - State-Owned Commercial Banks

AidData defines Chinese state-owned commercial banks as Chinese banks that are
majority-owned by the Chinese government or one of its subsidiaries. The three largest
state-owned commercial banks are the Industrial and Commercial Bank of China (ICBC), the
Bank of China (BOC), and the China Construction Bank Corporation (CCB). There are some
banks that are not officially designated as Chinese state-owned commercial banks but that are
consistent with AidData’s definition.47 AidData classifies the following institutions as Chinese
state-owned commercial banks: China Construction Bank Corporation (CCB), Industrial and
Commercial Bank of China (ICBC), Bank of China (BOC), China Bank of Communications
(BoCom or BoComm), Agricultural Bank of China, Postal Savings Bank of China (PSBC), China
Bohai Bank, Bank of Shanghai, China CITIC Bank, China Merchants Bank, Huaxia Bank Co.,
Ltd., and China Everbright Bank Co., Ltd. This group of banks includes so-called shareholding
commercial banks that are subsidiaries of state-owned enterprises (e.g., China CITIC Bank) and
city commercial banks (i.e., Bank of Shanghai).

Loans from Chinese state-owned commercial banks include term loans, bridge loans, revolving
credit facilities, working capital loans, commodity-backed loans, club loans, syndicated loans,
and buyer’s credits.48 They are typically denominated in USD or EUR. They are generally
provided on less concessional terms than China Eximbank loans because, unlike China
Eximbank, these institutions must maintain their own balance sheets and lend without receiving
official subsidies from the state. The base interest rate on a loan from a Chinese state-owned
commercial bank is usually tethered to the (floating) London Interbank Offered Rate (LIBOR) or
Euro Interbank Offered Rate (Euribor), with an additional margin incorporated to account for
borrower-specific risk and repayment capacity. Maturities and grace periods vary widely. These
loans are granted to both government agencies and companies.49

47
For example, China Merchants Bank is officially classified as a joint-stock commercial bank, but its
parent company, China Merchants Group, is under the direct supervision of State-owned Assets
Supervision and Administration Commission of the State Council (SASAC) and is classified as a
state-owned enterprise. As such, China Merchants Bank fits AidData’s definition of a state-owned
enterprise, but has been coded as a state-owned commercial bank for the purposes of analysis.
48
See, for example, https://www.dropbox.com/s/ekijqc9ubp1bdb7/ICBC%20Pitch%20Deck.pdf?dl=0
49
See Morris et al. (2020).

67
2.5.1.6 - People’s Bank of China (PBOC)
In recent years, China has ramped down its lending for infrastructure projects and ramped up
emergency rescue lending operations in LICs and MICs (Horn et al. 2023). The People’s Bank of
China (PBOC)—China’s central bank—is by far the most important financier of international
emergency rescue lending operations, which it provides in the form of drawdowns under
foreign currency swap line (FXSL) agreements.50

An FXSL agreement—also known as a bilateral currency swap (BCS) agreement or a central

bank liquidity swap agreement—is an agreement between the central banks of two countries to
exchange cash flows in different currencies at predetermined rates over a specified period of
time. Central banks participate in these agreements to facilitate bilateral trade settlements
using their national currencies (rather than relying on a third-party currency such as the U.S.
dollar), manage demands from their local banks, and provide liquidity support to financial
markets. The party that draws down on the swap line becomes the borrower and the other
party becomes lender. During the term of the swap, the party that draws down on the swap
line makes either fixed or floating interest payments on the principal amount. If both parties
draw down on the swap line, then both parties exchange fixed or floating interest payments on
the principal amounts.51

AidData has several coding guidelines that are specific to FXSL agreements:
● Drawdowns under FXSL Agreements: AidData should not capture the signing of FXSL
agreements as records in the dataset. Rather, it should only capture drawdowns
(borrowings) under FXSL agreements.
● Funding Agency: An FXSL agreement is by definition an agreement between the central
banks of two countries. As such, PBOC should be the funding agency for all FXSL
agreements.
● Receiving Agency: The receiving agency of all borrowings via FXSL agreements should
be the central bank of the recipient country.
● Flow Type: The flow type field should be set to loan.
● FXSL/BOP Flag: With the new loan categorization scheme introduced in the 3.0 version
of the TUFF methodology, all drawdowns under FXSL agreements are flagged as such
in the corresponding “FXSL/BOP” field.
● Collateralization: AidData should treat drawdowns under FXSL agreements with the
PBOC as collateralized loans because, in a FXSL arrangement, the currency of the
borrower is held as collateral while the lender receives interest on the amount drawn
down by the borrower until repayment is made. The “Collateralized” field should be set
to “Yes” for all such records.
● Collateral Provider: The receiving agency should be coded as the “Collateral Provider”.
● Collateral: The source of collateral should be recorded as the receiving agency’s deposit
in a bank account accessible to PBOC (e.g., “SBE deposit of Pakistani rupees equivalent
to RMB 5 billion in a bank account accessible to the PBOC”).

50
In 2013, the PBOC and SAFE (its subsidiary) were responsible for only 6% of China’s official sector
lending commitments to LICs and MICs. By 2021, that figure reached 54% (Parks et al. 2023).
51
More detailed information about currency swaps with the PBOC can be found at
https://www.imf.org/-/media/Files/Publications/WP/2021/English/wpiea2021210-print-pdf.ashx and
https://thechinaguys.com/the-rise-of-the-renminbi-the-reality-of-bilateral-swap-agreements/ and
https://www.imf.org/external/pubs/ft/bop/2017/pdf/17-25a.pdf.

68
 ● Intent: The intent for all drawdowns under FXSL agreements should be coded as
“Mixed.”
● Flow Class: The flow class for all drawdowns under FXSL agreements should be coded
as “OOF-like,” because the intent is “Mixed” and the loans are non-concessional (with
short maturities).
● Sector: The flow class for all drawdowns under FXSL agreements should be coded as
“Banking and Financial Services.”
● Title: AidData coders are instructed to make it clear in the title field if the record
captures a drawdown under an FXSL agreement. The title should adhere to a format in
line with the following example: “SBP makes RMB 5 billion drawdown under currency
swap agreement with PBOC in Fiscal Year 2013.”

In addition to the described lending under FXSL arrangements, PBOC has established funds
through intergovernmental organizations to provide lending for projects/activities, such as the
China Co-Financing Fund for Latin America and the Caribbean (CHC) which is administered by
the Inter-American Development Bank.

AidData has several coding guidelines that are specific to PBOC funds administered by
intergovernmental organizations:
● Funding Agency: The funding agency for the record should be the People’s Bank of
China.
● Co-Financing Agencies: If the fund supports co-financed projects/activities, then the
intergovernmental organization which administers the fund and/or any other
organizations providing financing in support of the project/activity should be included in
the “Co-Financing Agencies” field.
● Implementing Agencies: The fund (e.g. CHC) should be included in the “Implementing
Agencies” field with the Organization Type set to “Intergovernmental Organization,”
along with any other organizations involved in the implementation of the
project/activity.
● Flow Type: The flow type field should be set to loan because PBOC exclusively supports
lending activities.
● Umbrella Record: Coders should create an umbrella record which captures the total
allocation made by PBOC to the fund. The specific projects/activities financed through
the fund should be captured as separate non-umbrella records.
● Parent ID: Coders should create a Parent ID for the fund, and all projects/activities
supported by the fund should be assigned to this Parent ID.52
● Title and Description: If a project/activity is financed through one of these funds,
AidData coders are instructed to refer to the name of the fund (or an abbreviated
version thereof) in brackets at the start of the title (e.g. “[China Co-Financing Fund]”).

2.5.2 - Recipients of Chinese ODA and OOF

In tracking Chinese ODA and OOF, we seek to capture the provision of funding and in-kind
transfers of goods and services from official sector institutions in China to any entity overseas
(excluding transfers for investment purposes). Therefore, consistent with OECD-DAC
guidelines, we seek to capture Chinese ODA and OOF flows to public sector and private

52
For example, all projects/activities financed from the China Co-Financing Fund for Latin America and
the Caribbean are assigned to Parent ID 8.

69
sector agencies in recipient countries including but not limited to government agencies,
state-owned banks, state-owned companies, private companies, and special purpose vehicles
(including joint ventures). To capture the specific entities responsible for receiving and
managing these incoming financial and in-kind transfers, the 3.0 dataset includes a “Direct
Receiving Agencies” and “Indirect Receiving Agencies” fields. The “Direct Receiving
Agencies” field provides the name of the agency designated to receive and manage the
financial or in-kind transfer. For projects/activities that are financed with loans, the receiving
agency is the entity responsible for debt repayment. If a receiving agency (borrower) on-lends
the proceeds of a loan to an additional entity or entities, then the borrower is captured in the
“Direct Receiving Agencies” field and the additional entity or entities which receive loans from
the borrower is captured in the “Indirect Receiving Agencies” field. If more than one entity is
responsible for receiving and managing incoming grant funds or an in-kind transfer, all of these
entities are identified in the “Direct Receiving Agencies” field (as pipe-delimited entries).

Each receiving agency is also assigned to one of ten organization type categories in the
"Direct Receiving Agencies Type" and “Indirect Receiving Agencies Type” fields: Government
Agency, State-Owned Bank, State-Owned Company, State-Owned Fund, Intergovernmental
Organization, Special Purpose Vehicle/Joint Venture, Private Sector, NGO/CSO/Foundation,
Miscellaneous, or Unspecified. The organization type is preceded by one of three descriptors
regarding the country of origin: Chinese, Recipient, or Other (e.g. Recipient Government
Agency).

AidData codes the recipient country based on where the project/activity took place (or is
scheduled to take place) and not where the borrowing institution is legally
incorporated/domiciled. However, in cases where the project/activity is not taking place in a
particular jurisdiction (e.g. when a shipping vessel or an airplane is being purchased to facilitate
trade/travel between multiple countries), we code the recipient country by identifying the
owner of the borrowing institution (receiving agency) and its country of origin. If the borrowing
institution is wholly-owned or majority-owned by an entity in Country A, we code Country A as
the recipient country. If the borrowing institution is legally incorporated in Country B and is a
publicly traded company (not majority-owned by a company or resident from one country), we
code Country B as the recipient. AidData Record ID #62763 is one example of this type of
record.

2.5.3 - Chinese ODA and OOF Financing Mechanisms

2.5.3.1 - AidData’s “Flow Type” Categorization

In an effort to capture all Chinese ODA and OOF, we record a wide range of projects and
activities that benefit from transfers (“flows”) of goods, services, or funding from official sector
institutions in China. These flows include loans, grants, technical assistance, scholarships or
training provided in China to citizens of other countries, debt forgiveness, and debt
rescheduling. In the 3.0 dataset, these types of activities and flows are captured as a record’s
“Flow Type.” Below is an overview of each flow type category:
● Grant: The donation of money or an in-kind donation of goods from an official sector
institution in China. Chinese grant assistance commonly includes donations of supplies
or equipment, the provision of humanitarian aid or disaster relief, the establishment of a

70
 Confucius Institute within the host country, or financing for the construction of a
government building, school, hospital, or sports stadium.
● Technical Assistance: The formal provision of skills training, instruction, consulting
services, and information sharing by official sector entities and experts from China.
● Scholarships: Funding from an official sector institution in China that allows a citizen
from the host country to study at a Chinese university or other educational institution.
● Training in Donor Country: Training programs and activities that are sponsored by an
official sector institution in China held for host country citizens in China. Training
provided by Chinese entities outside of China is classified as technical assistance.
● Loan: A financial transfer from an official sector institution in China to an overseas entity
under certain terms and conditions of repayment. See Section 2.5.3.3 for a detailed
description of the types of loans included in the 3.0 dataset.
● Debt Forgiveness: The total or partial cancellation of debt owed by a borrowing
institution in the host country to a Chinese government or state-owned entity.
● Debt Rescheduling: Changes to the terms of a loan issued by an official sector
institution in China, such as interest rate, grace period, or maturity date. This is usually
meant to ease the repayment burden of a borrower institution in the host country.

To facilitate the aggregation of flow types based on certain criteria, we have introduced a
“Simplified Flow Type” field in the 3.0 version of the dataset. The table below displays a
side-by-side comparison of the two flow type fields.

Flow Type Simplified Flow

Type

Free-standing Technical
Assistance

Grant Grant

Scholarships/Training

Debt Forgiveness

Loan Loan

Debt Rescheduling Debt Rescheduling

Vague Vague

2.5.3.2 - Financing Agreements

Official sector institutions in China have developed several different types of “standard”
financing agreements. These agreement types are described in greater detail below, with
coding guidance for each type of agreement.

71
Framework Agreement (In Chinese: 框架协议): This is a non-binding agreement that
memorializes the intent of the lender and the prospective borrowing institution. These
agreements typically precede the signing of an actual loan agreement and sometimes identify
the expected face value of the loan and its borrowing terms. Framework agreements often
correspond to a single project that will be financed with a single loan, but they can also serve
as umbrella agreements through which multiple projects will be financed. Under a framework
agreement, the borrowing institution (receiving agency) typically must request and secure
approval for a specific loan/project from the official sector financing institution in China before
the transfer of funds can occur.53

AidData coders are instructed to follow several coding guidelines that are specific to framework
agreements (in 3 different scenarios):
● Scenario 1: Only a framework agreement is signed. If there is no evidence of a
subsequent financing agreement being signed or a subsidiary project being approved
under the framework agreement, the framework agreement should be coded as an
umbrella project. The financing amount that is referenced in the framework agreement
should be coded as the transaction amount. If the financing amount is unknown, the
transaction amount should be set to missing. If only a framework agreement was
signed, the status field should be coded as Pipeline: Pledge. If a framework agreement
was signed for a single project, it should not be coded as an umbrella project.
● Scenario 2: A single financing agreement for the full financial amount referenced in the
framework agreement is signed. The status field should be coded as Pipeline:
Commitment. The commitment year field should be set to the year in which the
financing agreement was signed and not the year in which the framework agreement
was signed.
● Scenario 3: Multiple projects are financed under the framework agreement. The
framework agreement should be retained as an umbrella record, and separate (linked)
records should be created for any subsequent financing agreements and subsidiary
projects.

Economic and Technical Cooperation Agreements (In Chinese: 济技术合作协定): See Section
2.5.1.2 for coding guidance.54

Letters of Intent: Letters of Intent (and “term sheets”) are usually pre-commitment documents
that are issued unilaterally by official sector institutions in China to indicate interest in financing
a project or an intention to finance a project.55 Sometimes letters of intent from official sector
institutions in China are characterized as a memorandum of understanding (MOU).56

53
By way of illustration, a concessional loan framework agreement can be accessed via
​https://www.dropbox.com/s/oc69pos782xj3sx/China_Framework%20agreement%20on%20concessional
%20loan%20for%20Malekula%20and%20Tanna%20roads_09112018.pdf?dl=0
54
An illustrative ETCA can be found here:
https://www.dropbox.com/s/pfuan0lxsmxosgz/Law%20for%2010%20million%20yuan%20loan%20in%20
2009.pdf?dl=0
55
An illustrative pre-commitment document is accessible via
https://www.dropbox.com/s/r608h03z7pjy4hy/2015.06.08%20ICBC%20_%20Amu%20Power%20Term%2
0Sheet.pdf?dl=0
56
An illustrative MOU from China Eximbank can be accessed here:
https://www.dropbox.com/s/l9hnqq9bzwxb61w/2015%20China%20Eximbank%20Loan%20for%20Centr
al%20Termoelectica%20Manuel%20Belgrano%20Project%20in%20Argentina.pdf?dl=0

72
AidData coders are instructed to follow one guideline that is specific to letters of intent (and
term sheets):
● The status field should be set to Pipeline: Pledge.

Letters of Exchange (In Chinese: 项目换文 or 换文): Letters of Exchange constitute an

agreement between the Chinese government and a recipient government institution regarding
one or more specific projects. These documents provide explanations, detailed elaborations, or
amendments to projects financed with Chinese government grants or interest-free loans
(usually from MOFCOM) that have been agreed upon by both governments. Letters of
Exchange are sometimes referred to as an Exchange of Notes. Letters of Exchange can
supplement an existing agreement, or serve as a stand-alone agreement.57

AidData coders are instructed to follow several guidelines that are specific to Letters of
Exchange:
● Letters of exchange provide formal documentation that codifies and elaborates the
Chinese government’s official commitment to support a specific project. While the
projects described in Letters of Exchange are sometimes funded with the grant or loan
proceeds from an ETCA, they can also be supported through separate, stand-alone
funding (usually from MOFCOM).
● Letters of Exchange provide evidence that an official commitment has taken place.
Therefore, the status field should be set to Pipeline: Commitment (unless there is
evidence that the project in question has progressed beyond the official commitment
stage).
● If an ETCA was signed prior to the signing of Letters of Exchange, the project described
in the Letters of Exchange should be linked to the ETCA record.

2.5.3.3 - Loan Flow Type

Number of Lenders

● Bilateral loan
○ A loan issued by one lender to a borrower. A bilateral loan can coexist with
co-financiers, but these co-financing institutions must provide financing via
legally separate mechanisms, i.e. an entirely separate loan agreement.
○ The full loan amount can be offered in several tranches that need not have
identical terms (e.g., Tranche A is a term loan with a 2% interest rate and a 10
year maturity, Tranche B is a working capital loan with a 4% interest rate and a 5
year maturity; Tranche C is a bridge loan with a 7% interest rate and a 1 year
maturity).

57
For more information about different types of Letters of Exchange, see Subsection 3 (国际发展援助换
文) of
https://books.google.com/books?id=vFP1e9HQXNcC&pg=PA230&lpg=PA230&dq=%E6%94%BF%E5%
BA%9C%E9%A1%B9%E7%9B%AE%E6%8D%A2%E6%96%87&source=bl&ots=TR03fKDb_y&sig=ACfU3
U2dSs4DLDFLynnQBoMQDGG1LyGoJw&hl=en&sa=X&ved=2ahUKEwjX-aD3vqToAhUDmHIEHfAJA-AQ
6AEwAHoECAkQAQ#v=onepage&q=%E6%94%BF%E5%BA%9C%E9%A1%B9%E7%9B%AE%E6%8D%
A2%E6%96%87&f=false

73
 ○ Nearly all of the official sector Chinese lenders—Chinese state-owned policy
banks and commercial banks, as well as other lenders including China’s Ministry
of Commerce and other state-owned enterprises—participate in bilateral loans.
○ If there is only one funding agency, no co-financing agencies, and the flow type
= loan, then the “Number of Lenders” field auto-populates as "Bilateral Loan,"
because the record is likely a bilateral loan. However, AidData coders can
manually change it to “Syndicated/Club Loan” (such as in cases where it is
known a Chinese bank participated in a syndicated loan, but the specific lenders
are unknown).
● Syndicated Loan or Club Loan
○ A loan issued by a consortium (‘syndicate’ or ‘club’) of lenders to a borrower.
○ Syndicated loans are offered by a group of lenders and are an attractive option
when one lender does not have the capacity to finance a large project on its
own and/or wishes to share credit risk. The full loan amount can be offered in
several tranches that need not have identical terms (e.g., Tranche A is a term
loan with a 2% interest rate and a 10 year maturity, Tranche B is a working
capital loan with a 4% interest rate and a 5 year maturity; Tranche C is a bridge
loan with a 7% interest rate and a 1 year maturity); nor does each member of the
syndicate need to participate in each tranche. Lenders can also sell their shares
of a syndicated loan in the secondary market. China’s state-owned policy banks
and commercial banks are the main official sector financing institutions in China
that participate in syndicated loans.
○ Known in Chinese as 银团贷款.
○ AidData coders are instructed to follow several coding guidelines that are
specific to syndicated loans:
i. Scenario 1: Each official sector financing institution from China and its
financial commitment amount to the syndicate is known. In this scenario,
separate records should be created for each official sector financing
institution from China. The transaction amounts in these records should
be populated with the financial amount that each official sector financing
institution from China committed to the syndicated loan. All of the
records should be linked through the title and project description fields
(referencing each AidData Record ID number), and they should be
assigned to the same Parent ID. For each record, the official sector
financing institution from China whose loan is being captured should be
coded as the funding agency, and other lenders that participate in the
syndicate should be coded as co-financing agencies.
ii. Scenario 2: Every official sector financing institution from China (or the
total number of official sector financing institutions from China in the
syndicate) is known but the individual financial contributions
(commitments) of each participant in the syndicate are not known. In this
scenario, AidData coders are instructed to (a) assume that each official
sector financing institution from China contributed (committed) an equal
amount to the syndicated loan and (b) estimate the contributions
(commitments) of each financier by dividing the total face value of the
loan with the total number of financiers in the syndicate. AidData coders
are additionally instructed to create a single record that captures the
total amount contributed (committed) by all of the official sector
financing institutions from China that participated in the syndicate. In this

74
 scenario, all official sector financing institutions from China are coded as
funding agencies. Participants in the loan syndicate that are not official
sector financing institutions from China are coded as co-financing
agencies.
iii. Scenario 3: Every official sector financing institution from China (or the
total number of official sector financing institutions from China in the
syndicate) is known and the individual financial contributions
(commitments) of each participant in the syndicate are also not known.
AidData coders are instructed to create a single record with no
transaction amount. The official sector financing institutions from China
of the syndicated loan are coded as funding agencies. Participants in the
loan syndicate that are not official sector financing institutions from China
are coded as co-financing agencies.

Loan Type Definitions

● Export Buyer’s Credit

○ A loan that is issued by Chinese state-owned policy banks and Chinese
state-owned commercial banks to overseas borrowing institutions to facilitate
their acquisition of goods and services from a Chinese supplier (i.e. export
promotion).
○ These loans are typically denominated in USD or EUR; they are usually issued
with a floating market interest rate (such as LIBOR or EURIBOR) plus a margin;
they often have shorter maturity lengths and grace periods than preferential
buyer’s credits (PBCs); and the borrowers need not be government institutions.
○ The lender usually authorizes the borrower to use the proceeds from the export
buyer’s credit to finance 85% of the total cost of a commercial contract with a
Chinese supplier.
○ In a typical export buyer’s credit (loan) agreement, there are four parties (see
Figure 2):
1. Chinese State-Owned Bank (Lender)
2. Chinese Company (Supplier)
3. Foreign Government (Borrower)
4. Foreign Government Ministry or SOE (Importer)

75
Figure 2: The Structure of a Typical Export Buyer’s Credit from China

○ Between these four parties, two agreements are concluded: (1) a commercial
contract between the Chinese supplier and the foreign importer, and (2) a loan
agreement between the Chinese state-owned bank and foreign borrower to
partially finance the commercial contract. Typically, the borrower can use the
proceeds of the loan to finance up to 85% of the total cost of the commercial
contract between the Chinese supplier (exporter) and the foreign importer.
However, this percentage can be higher or lower than 85%, depending on the
policies and practices of the official sector financing institution in China. The
borrower is expected to provide counterpart funding to cover the percentage of
the total cost of the commercial contract that is not covered by the loan.
Counterpart funding is sometimes used to provide an advance payment to the
Chinese supplier so that project implementation can commence before the loan
agreement is finalized.
○ The signing of the commercial contract usually predates the loan agreement, but
discussions/negotiations with the Chinese state-owned bank are often underway
at the time that the commercial contract is signed (or being negotiated).
Consequently, it is not unusual for the foreign importer or Chinese supplier to
publicly reference a loan agreement before it is finalized. AidData coders are
therefore instructed not to automatically treat the signing of a commercial
contract (that a foreign importer or Chinese supplier says will be financed by a
Chinese state-owned bank) as evidence that an official financial commitment has
taken place (i.e., a loan agreement has been signed).
○ China Eximbank provides two types of export buyer’s credit facilities: preferential
buyer’s credits (PBC) and buyer’s credit loans (BCLs). Sinosure provides credit
insurance for both loan types.58 China Development Bank and multiple Chinese
state-owned banks also provide export buyer’s credits.
○ Export buyer’s credits are frequently referred to as “buyer’s credits,“ buyer’s
credit loans,” and BCLs. In Chinese, export buyer’s credits are referred to as 优惠
出口买方信贷.

58
The full name of Sinosure (中国出口信用保险公司) is China Export & Credit Insurance Corporation.

76
 ○ Determining if a loan record is an export buyer’s credit. Export buyer’s credits are
usually identified as such in source materials, but this is not always the case.59 If
a loan is not explicitly identified as an export buyer’s credit, AidData coders are
instructed to mark a loan as an export buyer’s credit only if it meets four criteria:
i. The loan is denominated in USD or EUR.
ii. The borrower is a foreign company or foreign government.
iii. The face value of the loan is explicitly identified in an official source, and
it is not estimated or assumed.
iv. The reported face value of the loan is worth less than 100% of the
commercial contract cost.
○ AidData coders are also instructed that the following conditions provide
evidence that a loan may be an export buyer’s credit, but none of these
conditions should be considered to be sufficient to assume so:
i. The loan is covered by buyer’s credit insurance.
ii. The loan is insured by Sinosure, and the insurance appears to be credit
insurance rather than investment insurance.60
iii. The proceeds of the loan are to be used by the borrower to procure
goods, equipment, or services from a Chinese company.
○ AidData coders are instructed to follow several coding guidelines when working
with export buyer’s credits:
i. Loan Categorization: The “Export Buyer’s Credit” field should be coded
as “Yes” for any loan that is an export buyer’s credit.
ii. Intent: Export buyer’s credits can either have commercial intent or mixed
intent.
● Commercial intent should be identified when there is no evidence
that the project is seeking to improve economic development or
welfare in the recipient country.
● Mixed intent should be identified when there is at least some
evidence that the project is seeking to improve economic
development or welfare in the recipient country.
● Export buyer’s credits should never be coded as having only
development intent as they are explicitly designed to promote
the export of Chinese goods and services.
○ Title and Description: If an export buyer’s credit (loan) agreement was signed, it
should be explicitly identified as such in the project title and description and not
referred to as simply a “loan agreement.”
○ Flow class: All loans that are export buyer’s credits should be assigned to the
“OOF-like” flow class category.

● Supplier’s Credit/Export Seller’s Credit

○ An export seller’s credit (In Chinese: 出口卖方信贷) is a loan issued by a Chinese
state-owned bank (usually China Eximbank) to a Chinese company for the

59
They are usually identified in official source materials as “buyer’s credits” or “buyer’s credit loans”
rather than “export buyer’s credits.”
60
If there is evidence of Chinese investment, then the provision of investment insurance from Sinosure
cannot be ruled out. See
https://www.dropbox.com/s/uziiuht4wtuyzsg/Sinosure%20pitch%20deck.pdf?dl=0 and
https://www.dropbox.com/s/g4wvemp0txonztm/CDB%20and%20Sinosure%20Pitch%20Deck.pdf?dl=0

77
 purpose of increasing its exports. The proceeds of export seller’s credits are to
be used by borrowers (Chinese exporters) to finance their foreign sales. Chinese
exporters usually secure export seller’s credits when they need liquidity to offer a
supplier’s credit to an overseas buyer. Export seller’s credits can be denominated
in both local and foreign currency.
○ If a Chinese company extends a loan to a borrower and the borrower is
expected to use the loan proceeds to purchase goods and services from that
Chinese company, then the loan is a supplier’s credit. Supplier’s credits are also
known as seller’s credits or vendor financing.
○ Supplier’s credits from Chinese state-owned enterprises (e.g. ZTE, CATIC,
NORINCO, AVIC International, and Poly Technologies) are granted to both
public and private sector customers. Their terms vary widely, but they usually
have shorter maturities and grace periods, and interest rates are typically
tethered to LIBOR or EURIBOR plus a margin.61
○ There are typically 2 scenarios involving supplier’s credits (issued by Chinese
state-owned enterprises) and export seller’s credits (issued by Chinese
state-owned banks):
i. Scenario 1. A Chinese state-owned enterprise (exporter) provides
supplier’s credit to a buyer (borrower) in the recipient country for the
purchase of its goods and/or services. This scenario involves 1
agreement (i.e., the supplier’s credit agreement).
ii. Scenario 2. A Chinese state-owned bank provides an export seller’s credit
to a Chinese state-owned enterprise (exporter) to finance its sales to a
buyer in the recipient country. The Chinese state-owned enterprise, in
turn, uses the proceeds of the export seller’s credit to issue a supplier’s
credit to the buyer (borrower) in the recipient country. This scenario
involves 2 agreements (i.e., the supplier’s credit agreement between the
Chinese state-owned enterprise and the buyer/borrower in the recipient
country and the export seller’s credit agreement between the Chinese
state-owned enterprise and the Chinese state-owned bank).
○ AidData coders are instructed to follow several coding guidelines that are
specific to supplier’s credits and export seller’s credits:
i. Loan Categorization: The “Export Seller’s Credit/Supplier’s Credit” field
should be coded as “Yes” for any loan that is a supplier’s credit or export
seller’s credit.
ii. When a Chinese supplier receives an export seller’s credit from a Chinese
state-owned policy bank (or any other Chinese state-owned bank) and
uses the proceeds of the export seller’s credit to on-lend to its foreign
customers, it is ultimately funding from the Chinese government that is
being transferred to the recipient country. The funding agency in this
case should only be coded as the bank that provides the export seller’s
credit. The Chinese supplier that receives the export seller’s credit should
be coded as the direct receiving agency and the foreign customer in the
recipient country that in turn receives the supplier’s credit from the
Chinese supplier should be coded as the indirect receiving agency.

61
An illustrative supplier’s credit agreement can be accessed at
https://www.documentcloud.org/documents/20488282-gha_2019_485

78
 iii. In the event that a supplier’s credit is provided by a Chinese state-owned
enterprise (supplier) and there is no evidence of the provision of an
export seller’s credit from a Chinese state-owned bank to that Chinese
supplier, then the Chinese state-owned enterprise should be coded as
the funding agency and the foreign customer (borrower) that received
the supplier’s credit should be coded as the direct receiving agency.
iv. Intent: Supplier’s credits and export seller’s credits can either have
commercial intent or mixed intent.
● Commercial intent should be coded when there is no evidence
that the project is seeking to improve economic development or
welfare in the recipient country.
● Mixed intent should be coded when there is evidence that the
project is seeking to improve economic development or welfare
in the recipient country.
● Supplier’s credits and export seller’s credits should never be
coded as having only development intent since they explicitly
promote the export of Chinese goods and services.
■ Title and Description: When a supplier’s credit/export seller’s credit
agreement is issued, it should be explicitly identified as such in the
project title and description and not referred to as simply a “loan
agreement.”
■ Flow class: All supplier’s credits and export seller’s credits should be
assigned to the “OOF-like” flow class category.

● Cross-Currency Interest Rate Swap (CC IRS)

○ A cross-currency interest rate swap (CC IRS) is an off-balance sheet way of
hedging against interest rate risk and foreign exchange risk. In a typical
cross-currency interest rate swap agreement, both parties to the transaction are
simultaneously lending to each other. That is to say, each party is both a lender
and a borrower, because they are lending to each other. The parties to the
transaction can lend to each other at different interest rates.
○ Cross-currency interest rate swaps are sometimes referred to simply as “hedging
arrangements,” “interest rate swaps,” or “cross-currency swaps.”
○ AidData coders are instructed to follow several coding guidelines that are
specific to cross-currency interest rate swaps:
i. Flow Type: The flow type of cross-currency interest rate swaps is loan.
● For purposes of capturing outbound official financial flows from
China, AidData codes cross-currency interest rate swaps as
one-way loans from China, and the project title, maturity period,
interest rate, currency, etc. should be reflective of this.
ii. Loan Categorization: The “CC IRS” field should be coded as “Yes” for
any loan that is a CC IRS.
iii. Transaction Amount: The transaction amount for cross-currency interest
rate swap records is the lending amount that the Chinese creditor
extends to the recipient as part of the swap.
iv. Interest Rate: The “hedge interest rate payable” of the CC IRS should be
coded as the interest rate, as it represents the interest rate that the
borrower has to pay the Chinese bank.

79
 ● The “hedge interest rate receivable” is the interest rate the
Chinese bank has to pay to the recipient lender.
v. Intent: CC IRS loans should be coded as having commercial intent.
● Cross-currency interest rate swaps are a form of corporate
financing and such they represent transactions with commercial
intent, such as helping local companies hedge against foreign
exchange risk and interest rate risks associated with foreign
currency-denominated borrowings).
vi. Flow class: All CC IRS loans should be assigned to the “OOF-like” flow
class category.
vii. Funding agency: The Chinese state-owned bank involved in the
cross-currency interest rate should be coded as the funding agency.
viii. Staff Comments: Coders are to use the staff comment field to describe
the inbound financial flow to China (i.e. the amount that the recipient lent
to the Chinese lender, the borrowing terms of that loan, and the currency
of denomination).

● Deferred Payment Agreement (DPA)

○ In a typical DPA, the Chinese company that the project owner in the host
country has selected as its engineering, procurement, and construction (EPC)
contractor is furthermore a lender to the project owner. The Chinese company
assigns receivables under its EPC contract with the project owner to one or more
Chinese banks. Upon assignment of receivables, the Chinese bank(s) will release
funds to the Chinese company so it can discharge its obligations under the DPA
as a lender.
○ DPAs are sometimes referred to as receivables loans, receivables financing
(finance), accounts receivable financing (finance), or A/R financing (finance). In
Chinese, they are known as 应收账款融资. These other terms are used because
the accounts receivable of a company (i.e., unpaid invoices) are being used as
collateral to unlock working capital—typically in the form of a bank loan
(“receivables loan”). Sellers often face cash flow problems when their buyers do
not make full payment at the due date of the invoice. A DPA—or receivables
financing arrangement—addresses this problem by allowing them to sell their
outstanding invoices to a bank at a discounted rate. This approach allows the
seller to receive the remaining invoice amount before the due date of the
invoice. The bank either gets its money back at invoice maturity through the
seller (acting as a collecting agent) or directly from the debtor.
○ AidData coders are instructed to follow several coding guidelines that are
specific to deferred payment agreements:
i. Flow Type: The flow type of DPAs is loan.
ii. Loan Categorization: The “DPA” field should be coded as “Yes” for any
loan that is a DPA.
iii. Intent: DPAs can either have commercial intent or mixed intent.
● Commercial intent should be coded when there is no evidence
that the project is seeking to improve economic development or
welfare in the recipient country.
● Mixed intent should be coded when there is evidence that the
project is seeking to improve economic development or welfare
in the recipient country.

80
 ● DPAs should never be coded as having only development intent
since they explicitly promote a Chinese company’s business
interests.
iv. Flow Class: All DPAs should be assigned to the “OOF-like” flow class
category.

● Engineering, Procurement and Construction (EPC) Plus Finance (EPC+F or EPCF)

Agreement
○ In a typical EPC+F (EPCF) arrangement (in Chinese: 融资+EPC), a project owner
in the host country has selected a Chinese company as its engineering,
procurement, and construction (EPC) contractor, and a Chinese bank issues a
loan to that EPC contractor but with a sovereign guarantee from the host
government.
○ AidData coders are instructed to follow several coding guidelines that are
specific to EPCFs:
i. Flow Type: The flow type of EPCFs is loan.
ii. Loan Categorization: The “EPCF” field should be coded as “Yes” for any
loan that is an EPCF.
iii. Intent: EPCFs can either have commercial intent or mixed intent.
● Commercial intent should be coded when there is no evidence
that the project is seeking to improve economic development or
welfare in the recipient country.
● Mixed intent should be coded when there is evidence that the
project is seeking to improve economic development or welfare
in the recipient country.
● EPCFs should never be coded as having only development intent
since they explicitly promote a Chinese company’s business
interests.
iv. Flow Class: All EPCFs should be assigned to the “OOF-like” flow class
category.

● Foreign Currency Swap Line (FXSL)

○ An FXSL agreement—also known as a bilateral currency swap (BCS) agreement
or a central bank liquidity swap agreement (双边货币互换协议 in Chinese)—is an
agreement between the central banks of two countries to exchange cash flows
in different currencies at predetermined rates over a specified period of time.
Central banks participate in these agreements to (a) facilitate bilateral trade
settlements using their national currencies (rather than relying upon a third-party
currency such as the U.S. dollar), (b) manage demands from their local banks,
and (c) provide liquidity to support financial market stability. The party that
draws down on the swap line becomes the borrower and the other party
becomes lender. During the term of the swap, the party that draws down on the
swap line makes either fixed or floating interest payments on the principal
amount. If both parties draw down on the swap line, then both parties exchange
fixed or floating interest payments on the principal amounts.
○ China’s central bank—the People’s Bank of China (PBOC)—is the only official
sector lending institution in China that issues FXSL agreements.
○ See Section 2.5.1.6 for a more detailed description of FXSL agreements,
including the guidelines AidData coders should follow.

81
● Government Concessional Loan (GCL) [Only from China Eximbank]
○ An RMB-denominated loan that China Eximbank issues to government
institutions on below-market terms (typically 20-year maturities, 5-year grace
periods, and 2% interest rates) to facilitate their acquisition of goods/services
from a Chinese supplier. The proceeds of a GCL can be used by government
borrowing institutions to finance up to 100% of the total cost of a commercial
contract with a Chinese supplier. China Eximbank does not expect the borrowing
institution to provide any “counterpart funding.”
○ In Chinese, GCLs are referred to as 优惠贷款.
○ See Section 2.5.1.3 for a more detailed description of GCLs, including the
guidelines AidData coders should follow.

● Inter-Bank Loan
○ A loan issued by one bank (lender) to another bank (borrower). All inter-bank
loans should also be coded as on-lending arrangements, but the opposite is not
true (i.e., not all on-lending arrangements should be treated as inter-bank loans).
○ In Chinese, inter-bank lending is known as 同业拆借.
○ AidData coders are instructed to code the “Inter-Bank Loan” field in the Loan
Categorization section as “Yes” for any loan that is an inter-bank loan.

● Balance of Payments (BoP) Loan, Liquidity Support Facility (LSF), or Foreign Currency
Deposit Loan
○ A loan issued by a Chinese state-owned policy bank, a Chinese state-owned
commercial bank, or China’s State Administration of Foreign Exchange (SAFE) to
a central bank or finance ministry in another country that explicitly authorizes the
borrower to use the proceeds of the loan to (a) shore up foreign exchange
reserves, (b) repay existing debts, and/or (b) finance general budgetary
expenditures.
○ BOP loans, LSFs, and foreign currency deposit loans usually have short
maturities and high interest rates. They are often referred to as “rescue” or
“emergency” loans (see Horn et al. 2023). In Chinese, they are sometimes
referred to as “sovereign loans” (主权贷). They do not support individual
projects or investments.
○ AidData coders are instructed to follow several coding guidelines that are
specific to BoP Loans, LSFs, and Foreign Currency Deposit Loans.
i. Loan Categorization: The “FXSL/BOP” field should be coded as “Yes”
for any loan that is a BoP Loan, LSF, or Foreign Currency Deposit Loan.
ii. Receiving Agency: The receiving agency should always be either the
central bank or finance ministry of a given recipient country.

● Interest-Free Loan [Automatically Populated]

○ A loan that is issued to a borrower without any interest accruing. The borrower is
only responsible for repaying the loan's principal amount.
○ China's Ministry of Commerce (MOFCOM) is the lead administrator of the
country’s interest-free (or “zero-interest”) loan program, often through Economic
and Technical Cooperation Agreements (ETCAs). However, MOFCOM is not the
only official sector institution in China that offers interest-free loans. For a full

82
 discussion of interest-free loans from the Ministry of Commerce, see Section
2.5.1.2 MOFCOM section.
○ In Chinese, interest-free loans are referred to as 无息贷款、零息贷款、or 免息贷
款.
○ The interest rate field for interest-free loans should be set to zero, and the
“Interest-free Loan” field should be coded as “Yes”.

● Investment Project Loan

○ A loan that is provided to finance the provision of goods, works, or services to
support a public or private investment project. These types of loans usually
involve building, rehabilitating, or upgrading physical assets and infrastructure.
○ These types of loans are sometimes referred to as capital expenditure (capex)
loans, project loans, or investment loans.
○ AidData coders are instructed to code the “Investment Project Loan” field in the
Loan Categorization section as “Yes” for any loan determined to be an
Investment Project Loan.

● Lease
○ A lease is a contractual arrangement calling for the lessee (user) to pay the lessor
(owner) for use of an asset. The lessor is the legal owner of the asset, while the
lessee obtains the right to use the asset in return for regular rental payments.
Under a capital lease (a financial arrangement where the lessee/borrower uses
an asset and pays regular installments plus interest to the lender/lessor), rental
payments are usually classified as interest and obligation payments, similarly to a
mortgage (with the interest calculated each rental period on the outstanding
obligation balance).
○ Many Chinese state-owned commercial and policy banks (i.e. ICBC’s wholly
owned subsidiary ICBC Financial Leasing Co., Ltd.) have their own leasing arms
and subsidiaries that provide financial leasing services.
○ AidData coders are instructed to follow several coding guidelines that are
specific to leases:
i. Flow Type: The flow type of leases is loan.
ii. Loan Categorization: The “Lease” field should be coded as “Yes” for any
lease project.
iii. Funding Agency: The specific leasing company involved in the
transaction (lessor) and not its banking parent (e.g. China Development
Bank Leasing Co., Ltd. instead of its parent CDB).
iv. Receiving Agency: The lessee should be coded as the receiving agency
of a lease agreement.
v. Title: The title should explicitly state the transaction is a lease (e.g. “ICBC
Leasing leases 6 Airbus A321 passenger jets to Transaero”).
vi. Maturity: The maturity period for leases is the length of the lease
agreement.
vii. Staff Comments: The following should be added to the Staff Comments
field: “AidData treats this lease as a loan. A lease is a contractual
arrangement calling for the lessee (user) to pay the lessor (owner) for use
of an asset. The lessor is the legal owner of the asset, while the lessee
obtains the right to use the asset in return for regular rental payments.
Under a capital lease (a financial arrangement where the lessee/borrower

83
 uses an asset and pays regular installments plus interest to the
lender/lessor), rental payments are usually classified as interest and
obligation payments, similarly to a mortgage (with the interest calculated
each rental period on the outstanding obligation balance).”

● Sale-and-leaseback agreements
○ Sale-and-leaseback agreements (SLBs)—also known as purchase-and-leaseback
agreements (PLBs), sale-leaseback agreements, and leasebacks— are specific
types of lease agreements in which the original owner sells an asset to a leasing
entity, which, in term, leases the asset back to the original owner, allowing the
original owner to raise upfront capital for the asset, without taking on a loan,
while still continuing to operate it, using the rental payments to repay the value
of the asset. At the end of a SLB, the original owner typically has an option or
obligation to repurchase the asset from the leasing entity. SLBs are generally
considered to be off-balance-sheet hybrid debt products.62
○ SLBs can be conducted for a variety of assets including real estate (i.e. a
freehold property), equipment (e.g. a drilling rig), and vehicles (e.g. an aircraft or
container ship).
○ AidData coders are instructed to follow several coding guidelines that are
specific to SLBs:
i. Flow Type: The flow type of SLBs is loan.
ii. Funding Agency: The specific leasing company involved in the
transaction and not its banking parent (e.g. China Development Bank
Leasing Co., Ltd. instead of its parent CDB).
iii. Loan Categorization: The “Lease” field should be coded as “Yes” for any
SLB agreement.
iv. Transaction Amount: The value of the SLB agreement, i.e. the purchase
consideration paid by the leasing company to the original owner for the
asset.
v. Receiving Agency: The lessee should be coded as the receiving agency
of a lease agreement.
vi. Maturity: Length of the SLB agreement.
vii. Title: The title should explicitly state the transaction is a SLB, (e.g. “ICBC
Financial Leasing enters into three sale and leaseback agreements worth
$93.15 million USD with Star Bulk Carriers Corp”).
viii. Description: If the SLB includes details on the existence of a bareboat
charter or security, these should be included in the description.
ix. Collateral and Accountable Agencies: Should the lessor or another entity
provide security for the SLB, that entity should be coded as the
“Collateral Provider”, and a description of the collateral should be
entered into the “Collateral” field.
x. Start Implementation and End Implementation Dates: The start and end
date of a sale-leaseback agreement is similar to the loan commitment
date and end date, and so does not correspond to the database’s Actual
Implementation Start Date and Actual Completion Date fields (which

62
See https://www.nreionline.com/mag/sale-leaseback-king-balance-sheet-financing and
https://www.investopedia.com/terms/l/leaseback.asp

84
 should capture actual implementation start and end date). As such, start
and end dates for sale-leaseback agreements should not be included.
xi. Staff Comments: The following should be added to the Staff Comments
field: “Sale and leaseback (or sale-leaseback) agreements are generally
considered to be off-balance-sheet hybrid debt products.”

● Mergers and Acquisitions (M&A) Loan

○ A loan that is issued to a borrower to facilitate its acquisition of an equity stake
in a company and/or to facilitate the consolidation of multiple companies (i.e., a
merger).
○ M&A loans—also known as “acquisition loans” or “acquisition financing” or 并
购贷款 in Chinese—are typically issued to Chinese companies with short
maturities and high interest rates. They are sometimes referred to as “bridge
loans” because they provide short-term cash flow until a borrower can secure
permanent financing or remove an existing obligation (although not all bridge
loans are M&A loans).
○ One particular type of M&A loan is China Eximbank’s Overseas Investment Loan
(In Chinese: 境外投资贷款), which is used to support Chinese enterprises’
overseas investments. These loans can be used to fund acquisitions, fixed asset
investments, and overseas equity investments approved by the Chinese
authorities. For further information and coding instructions on the Overseas
Investment Loan, see Section 2.5.1.3.
○ AidData coders are instructed to follow several coding guidelines that are
specific to M&A loans.
i. Loan Categorization: The “M&A” field should be coded as “Yes” for any
loan that is an M&A.
ii. Intent: M&A loans have commercial intent, as they support purely
commercial interests.
iii. Flow Class: All M&A loans should be assigned to the “OOF-like” flow
class category.

● Preferential (Export) Buyer’s Credit (PBC)

○ A USD-denominated loan that China Eximbank issues to government institutions
to facilitate their acquisition of goods/services from a Chinese supplier. The
borrowing terms of these loans vary, but they are offered with fixed rather than
floating (market) interest rates (such as LIBOR or EURIBOR), which are usually
more generous than prevailing market rates. China Eximbank has a policy of
allowing borrowers to use PBOC proceeds to finance 85% of the total cost of a
commercial contract with a Chinese supplier. China Eximbank usually requires
that the remaining 15% of the commercial contract cost be financed with
“counterpart funding” from the borrowing institution.
○ In Chinese, PBCs are referred to as 优惠出口买方信贷.
○ For further information and coding instructions on the PBC, see Section 2.5.1.3.

● Pre-Export Financing (PxF) or Commodity Prepayment Financing

○ A PxF facility is an arrangement in which a commodity (e.g. oil) producer gets
up-front cash from a customer in return for a promise to repay the customer with
that commodity (possibly at a discount) in the future. PXF funds may be
advanced by a lender or syndicate of lenders to a commodity producer to assist

85
 the company in meeting either its working capital needs (for example, to cover
the purchase of raw materials and costs associated with processing, storage and
transport) or its capital investment needs (for example, investment in plant and
machinery and other elements of infrastructure).
○ PxF facilities are also known as commodity prepayment financing arrangements.
In Chinese, PxF facilities are known as 出口前融资 or 预出口融资.
○ AidData coders are instructed to follow several coding guidelines that are
specific to PxFs:
i. Flow Type: The flow type of PxFs is loan.
ii. Loan Categorization: The “PxF/Commodity Prepayment” field should be
coded as “Yes” for any loan that is a PXF or commodity prepayment
financing.
iii. Intent: PxFs can either have commercial intent or mixed intent.
● Commercial intent should be coded when there is no evidence
that the project is seeking to improve economic development or
welfare in the recipient country.
● Mixed intent should be coded when there is evidence that the
project is seeking to improve economic development or welfare
in the recipient country.
● EPCFs should never be coded as having only development intent
since they promote commercial interests, often imports to China.
iv. Flow Class: All PxFs should be assigned to the “OOF-like” flow class
category.

● On-Lending Arrangement
○ ​An arrangement in which a borrower uses the proceeds of a loan to lend to one
or more additional entities.
○ Also known as a “re-lending” arrangement.
○ AidData coders are instructed to follow several coding guidelines that are
specific to loans with on-lending.
i. Loan Categorization: The “On-Lending” field should be coded as “Yes”
for any loan that includes on-lending.
ii. Receiving agencies: The direct borrower of the loan (i.e. the institution
that signed the loan contract with a Chinese bank) should be coded as a
direct receiving agency. Any entities that the direct borrower on-lends to
should be added coded as indirect receiving agencies.

● Revolving Credit Facility (RCF)

○ In a typical RCF arrangement, the lender commits funding up to a certain level,
but unlike a “term loan” (that is repaid in regular payments over a set period of
time), the borrower can draw down, repay, and redraw on an
irregular/as-needed basis; an RCF provides liquidity for day-to-day operations; it
functions like a a credit card, except that the borrower is charged an annual
commitment fee on unused amounts (a “facility fee”).
○ RCFs are commonly known as “revolvers” and “revolving loans.”
○ AidData coders are instructed to code the “RCF” field in the Loan
Categorization section as “Yes” for any loan that is a revolving credit facility.

86
 ● Debt Refinancing
○ A new loan for the purpose of repaying one or more existing loans/debts.
○ Debt refinancing can occur when a borrower decides to change the source of its
financing for a project or activity (e.g., to a different bank, or from a loan to a
bond).
○ AidData coders are instructed to code the “Refinancing” field in the Loan
Categorization section as “Yes” for any loan intended to refinance existing debt.

● Working Capital Loan

○ A loan that provides funds for a borrower’s day-to-day operations (including
general corporate purposes) but not for making capital investments or
facilitating the acquisition of long-term assets.
○ AidData coders are instructed to code the “Working Capital” field in the Loan
Categorization section as “Yes” for any loan intended for working capital
purposes.

● Non-Recourse or Limited-Recourse Project Finance Transaction

○ When a project is financed with a limited-recourse or non-recourse structure, the
loan that is used to finance the acquisition, construction, and/or maintenance of
an asset—such as a toll road, a seaport, or an electricity grid—is exclusively
repaid with the cash flow generated by the asset (e.g., toll revenue, container
fees, or electricity sales), and the creditor either has no claim (“recourse”) or a
limited claim to any other assets as a basis for recovering the debt. In a
standard, limited-recourse or non-recourse project finance transaction, a creditor
lends to an independent legal entity that is established for the express purpose
of developing, owning, and operating a specific project. This entity is often
called a special purpose vehicle (SPV) because it is only allowed to engage in
activities that relate to a specific purpose (project), and it is legally prohibited
from incurring debts or obligations that are not related to that purpose (project).
○ In a non-recourse or limited recourse project finance arrangement, the
borrowing institution is always a special purpose vehicle (SPV) or joint venture
(JV).
○ Unlike non-recourse or limited recourse project finance arrangements, the
repayment of a full-recourse sovereign loan (i.e. a loan directly to a government
agency) does not depend upon the financial viability of a project or the cash
flow generated by any particular asset. A sovereign government borrower
guarantees the repayment of the loan, regardless of whether the asset
supported by the loan generates enough revenue to facilitate repayment, and
the creditor has a legal right to seize any and all assets of the borrower until the
full amount of the debt is covered (i.e. it has “full recourse” to the assets of the
borrowing government).
○ Limited-recourse and no-recourse project finance transactions usually involve a
mix of equity from the project sponsor63 (also known as equity investors or

63
In a public-private partnership (PPP) context, the terms project sponsor, project owner and
concessionaire are often used interchangeably. That being said, a concessionaire is slightly different
(from a typical project sponsor/owner) because its ownership of the project is time-limited (as
determined by the concession agreement). So, ownership of the project can return to the host
government at the end of the concession agreement.

87
 project founders) and debt from banks/financial institutions. Common
debt-to-equity ratios are 90:10, 80: 20, and 70:30. More often than not, project
sponsors have limited financial means and cannot on their own provide the total
capital required for the construction, development and operation of the
project/asset.
○ An example of a limited-recourse project finance transaction that is supported
by an official sector financing institution in China is the China-Laos Railway
Project (captured via AidData Record IDs #33726 and #85304).64 To implement
this project, three Chinese state-owned companies and a Lao state-owned
enterprise established a joint venture (SPV) called the Laos-China Railway
Company Limited (LCRC). The LCRC was established as a limited liability
corporation (LLC) to finance, design, construct, and manage a 418 kilometer
railway between the Chinese city of Kunming and the Laotian capital of
Vientiane on a public-private partnership (PPP) basis. The total cost of the
China-Laos Railway Project is $5.9 billion—equivalent to roughly one-third of
Laos’ GDP—and it is being financed according to a 60:40 debt-to-equity ratio
($3.54 billion of debt and $2.36 billion of equity). LCRC directly secured $3.54
billion of debt financing from China Eximbank, and the Government of Laos and
the Chinese Government contributed $730 million and $1.63 billion of equity
financing, respectively. In order to make its $730 million equity contribution to
the project, the Government of Laos secured a $480 million loan from China
Eximbank and it agreed to provide $250 million of its own funding (in annual
installments). The $3.54 billion debt secured by LCRC, which is jointly owned by
three Chinese state-owned companies that collectively hold a 70% equity stake
and one Lao state-owned enterprise that owns a 30% equity stake, is not backed
by a sovereign guarantee.
○ In Chinese, no-recourse or limited-recourse project finance transactions are
known as 项目融资.
○ AidData coders are instructed to follow several coding guidelines that are
specific to limited-recourse and no-recourse project finance transactions:
i. Loan Categorization: The “Project Finance” field should be coded as
“Yes” for any loan that is a limited-recourse and no-recourse project
finance transaction.
ii. Receiving Agency: The receiving agency should have an organization
type of “Joint Venture/Special Purpose Vehicle”.
iii. Description: The ownership of the JV/SPV involved in the project finance
must be explicitly stated in the description.
iv. Estimation of Transaction Amount: If the debt component of the project
was exclusively financed by an official sector financing institution from
China and no information on the exact amount of debt financing is
identified, coders should use the debt-to-equity ratio and the total
project cost to estimate the transaction (debt financing) amount. For
example, if the total project cost is 100 million USD and the project is
financed through a debt-equity ratio of 80:20, and the debt financing is

64
For an illustrative loan agreement issued by official sector financing institutions in China in support of a
limited-recourse project finance transaction, see
https://www.documentcloud.org/documents/20488803-sle_2017_468

88
 exclusively provided by an official sector financing institution from China.
The transaction amount should be coded as 80 million USD.

● Overseas Project Contracting Loan

○ A loan issued by China Eximbank to a Chinese company to help it finance an
overseas project contract. This loan can be denominated in USD or RMB. Per
China Eximbank policy, the contract cost that is financed with the loan should
not be lower than 1 million USD, and goods and services exported from China
under the contract should not be lower than 15% of contract cost.
○ In Chinese, overseas project contracting loans are referred to as 对外承包工程贷
款.
○ For further information and coding instructions on the Overseas Project
Contracting Loan, see Section 2.5.1.3.

● Collateralized Lending
○ Collateral is a right to an asset or a revenue stream that a creditor can rely upon
to secure repayment in the event that a borrower defaults on its payment
obligations. Collateral can come in many forms including “cash, stocks, and
negotiable bonds; irrevocable letters of credit; certificates of deposit;
assignment of receivables such as export earnings, electricity generation
charges, road tolls, and telecoms receipts; as well as physical assets such as
buildings, ports, and industrial plants. [...] Collateral can be (i) an existing or
future asset (stock) or (ii) a future flow or stream. The latter case, also called
future receipts or future receivables, can be defined as a financial amount (e.g.
USD) or a physical amount of goods to be delivered (e.g. barrels of crude oil)”
(IMF and World Bank 2020: 4, 6).
○ In a legal sense, collateralized debt “entails a borrower granting liens over
specific existing assets or future receivables to a lender as security against
repayment of the loan” (IMF and World Bank 2020: 4). However, collateralized
debt “also includes arrangements that do not constitute granting of a formal
security interest, but that have an equivalent effect.” For example, regardless of
whether a formal security interest is granted over an escrow account (or a
so-called “revenue account,” “special account,” or “proceeds account”) with a
minimum cash balance requirement, a loan is de facto (rather than de jure)
collateralized if the account is controlled by the lender (Gelpern et al. 2022).
○ According to the IMF and the World Bank (2020: 6), “[c]ollateral can be related
or unrelated to the purpose of the loan. Collateralized debt is considered to
have ‘related collateral’ if the loan is used to purchase or construct a new asset
(e.g. an airplane, an oil platform), and the asset or the future receipts it is
expected to generate (e.g., airline ticket sales, the revenues from the sale of oil)
serve as collateral to secure the debt. An example of ‘unrelated collateral’ is a
budget loan collateralized with oil receivables.”
○ An example of a collateralized loan is the buyer’s credit loan (BCL) that China
Eximbank issued to the Government Ghana for the Bui Dam Construction
Project (ID#183). It was collateralized with (a) net revenue from a Power Purchase
Agreement (PPA) between Bui Power Authority, an organization with a mandate
to plan, execute and manage the Bui Hydroelectric Project, and the Electricity
Company of Ghana (ECG) for the purchase of the energy to be generated from

89
 the Bui hydroelectric power plant, and (b) receivables from the Ghana Cocoa
Board’s sale of cocoa beans to Genertec International Corporation of Beijing.
○ Sometimes official sector lenders in China will demand an escrow account (or a
so-called “revenue account,” “special account,” or “proceeds account”) as a
form of collateral (Gelpern et al. 2022). These are often offshore bank accounts
(located in the lending country) into which project revenues or the proceeds
from export sales are deposited. The funds held in the account can then be used
to service the loans and/or serve as collateral (sometimes called a “security
package”).
i. For example, as a guarantee for a $1 billion China Eximbank loan, the
Republic of Congo is required to keep a minimum deposit balance
equivalent to 20% of total outstanding China Eximbank loans in an
escrow account (that China Eximbank controls) from the proceeds of its
oil sales to China.
○ Collateral requirements in loan agreements do not affect the way that a loan’s
grant element is calculated. However, they do influence the favorability of
lending terms in a broader sense. If two loans have identical interest rates,
maturities, grace periods, and fees, but one requires the borrower to provide a
source of collateral that China can seize in the event of default (e.g., foreign
currency earnings in an escrow account, a revenue-generating infrastructure
asset) and the other does not, the borrower would almost certainly consider the
loan with the collateral requirement to be less favorable than the one without
such a requirement (Morris et al. 2020).
○ Determining whether a loan involves collateralization. When source materials do
not specify if a loan is collateralized, AidData coders are instructed to follow
several guidelines to determine if the loan is collateralized in a de facto or de
jure sense:
■ If a source indicates that the borrower granted a formal lien or security
interest to the lender, the loan should be coded as collateralized.
■ If a source mentions a source of “security” for the loan or characterizes
the loan as “securitized,” this should be treated as evidence that the
borrower granted security interest to the lender and the loan should be
coded as collateralized (e.g., “The escrow account will provide a source
of security for the loan.”).
■ If a source indicates that a security agent was appointed (to enforce
rights against the collateral in the event that the borrower defaults on its
repayment obligations), the loan should be coded as collateralized.
■ If a special account, escrow account, revenue account, or proceeds
account (into which the borrower is required to deposit project-related
revenues or unrelated revenues) is mentioned, and the account is either
(a) controlled by the Chinese lender or (b) located in China, then the loan
should be coded as (de facto) collateralized. Under these conditions, it is
not difficult for the Chinese lender to seize or debit the account without
the consent of the borrower. However, when the account is controlled by
the borrower or a third party (or the account is located in the borrower
country or a third country), the Chinese lender does not have a de facto
source of collateral.
■ All pre-export finance (PXF) facilities should be coded as collateralized
since they are almost always secured by (1) an assignment of rights by

90
 the producer under an ‘offtake contract’ (i.e., a sale and purchase
contract between the producer and a buyer of that producer of goods or
commodities), and (2) a collection account charge over a bank account
into which proceeds due to the producer from the buyer of the goods or
commodities under the offtake contract are credited.
■ If the word ‘guarantee’ is mentioned in relation to the repayment of the
loan by a non-English language source, this may indicate
collateralization. In languages other than English, collateralized debt
arrangements are sometimes referred to as ‘guarantee’ or ‘guaranteed.’
However, in English, a (third-party) guarantee is a concept that is distinct
from collateralization. Whenever a loan guarantee is issued, an entity
other than the primary borrower (or ‘the primary obliger’) agrees to repay
the loan if the primary obliger goes into default. So if the primary obliger
is a government line ministry, subnational government, or state-owned
enterprise and a sovereign guarantee is issued in support of the loan,
that means that a central government entity (usually the Ministry of
Finance) has agreed to serve as the loan guarantor. As such, whenever a
guarantee is issued, there has to be an additional entity responsible for
repayment in the event that the primary obliger goes into default. It is
possible for a loan to be both guaranteed (in the English-language sense)
and collateralized. China Eximbank’s 2004 MLFA with the Government
Angola was guaranteed by Sonangol (the country’s state-owned oil
company) and collateralized with future revenues from the sale of oil
exports.
■ If the word ‘guarantee’ is mentioned in relation to a specific
asset/revenue stream, it is most likely referring to collateral. A third-party
repayment guarantee (e.g., a sovereign guarantee or corporate
guarantee) is not related to a specific asset or revenue stream. The
issuance of a third-party repayment guarantee allows the creditor to
secure repayment by pursuing any assets or revenue streams controlled
by the guarantor in the event of default (assuming the assets/revenue
streams in question are not protected by sovereign immunity). E.g., “The
Government of Togo agreed to establish an airport fee (RDIA) to
guarantee the repayment of the PBC and the GCL.” In this case the
airport fee reported is a form of loan collateral, instead of a third-party
guarantee.
■ If a source refers to a lending arrangement that is following the ‘Angola
model’ or the ‘resources-for-loan’ model, this should be treated as
evidence of collateralization. The resources-for-loan model pioneered by
China Eximbank in Angola, or the ‘Angola model,’ involves
collateralization. In Chinese, the resources-for-loan model is sometimes
called “互惠贷款”（reciprocal loan) or “资源与贷款合作框架协议” or “‘资
源、信贷、项目’一揽子合作模式” or “‘石油、信贷、工程’一揽子合作模式”
(oil-backed loan) or "安哥拉模式.”65
○ AidData coders are also instructed to follow several additional guidelines that
are specific to collateralized loans:

65
See https://www.dropbox.com/s/r5xhu7zdqiebbn3/2.EXIM-Bank.pptx?dl=0

91
 ■ The “Collateralized/Securitized” field should be set to “Yes” and the
specific source(s) of collateral should be identified in the Collateral
Description field.
■ The organization(s) responsible for providing the collateral should be
coded as the “Collateral Provider.”
■ The organization(s) responsible for acting as a Security/Collateral Agent,
if identified, should be coded as the “Security agent/Collateral Agent.”
■ Collateralization should not be used to determine a project’s intent
designation, concessionality designation, or flow class designation.

● Master Loan Facility Agreements and Credit Lines

○ Master loan facilities and lines of credit are not loan agreements and are not
treated as such in the 3.0 methodology. They are agreements designed to
support multiple subsidiary projects with multiple loans that must be
individually approved by the lender. In a typical master loan facility agreement or
credit line agreement, the lender and borrower first agree on the types of
allowable projects and the terms and conditions of lending. Then, the borrower
prepares individual project loan applications to give to the lender for approval
under the parameters established in the master facility agreement. Master loan
facilities and lines of credit typically identify the total amount of funding that can
be accessed, the types of projects that can be supported, and the borrowing
terms o(e.g., interest rate, maturity, grace period) for subsidiary projects under
the agreement. Master loan facilities are sometimes called master facilities,
master loan agreements, or framework agreements.66
○ A single loan can have multiple purposes or components. The key distinction
between a single loan with multiple components and a master loan facility is that
the latter involves subsidiary loan agreements that must be individually
approved by the lender.
○ AidData coders are instructed to follow several guidelines that are specific to
master loan facilities and lines of credit:
■ Scenario 1: If the master loan facility (or line of credit) is for unspecified
purposes, and there is information about how much has been drawn
down from the facility. Create one umbrella record for the master loan
facility, and one subsidiary record for the drawn down amount.
■ Scenario 2: If the master loan facility (or line of credit) is for specified
purposes, and coders know the exact number of subsidiary
projects/activities financed through the facility. Create one umbrella
record for the master loan facility, and subsidiary records for each
project/activity that was financed through the facility with separate
transaction amounts.
■ Scenario 3: If the master loan facility (or line of credit) is for specified
purposes, and coders are unable to identify all subsidiary
projects/activities financed through the facility (i.e., coders only find
information on one or a few of the subsidiary projects/activities financed
through the facility). Create one record to capture the entire transaction

66
These should not be confused with “facility agreements,” which is a shorthand term that is often used
by (Chinese state-owned) banks to refer to loan (or export credit) agreements. See, for example,
https://www.documentcloud.org/documents/20488803-sle_2017_468

92
 amount and set the umbrella field to “No.” For any subsidiary
projects/activities financed by the facility that coders are able to find,
create subsidiary records for each project/activity that was financed
through the facility but do not record their specific transaction amounts
to avoid duplication of transaction amounts.
■ Scenario 4: If the master loan facility (or line of credit) is for unspecified
purposes, and there is evidence that the facility has been completely
(100%) drawn down. Create one record to capture the entire transaction
amount and set the umbrella field to “No.”
■ Scenario 5: If the master loan facility (or line of credit) is for unspecified
purposes, and we are confident that (1) there is no risk of duplication with
existing flows, (2) the financing is going to one specific recipient entity,
(3) there is little to no chance of coders identifying the specific, subsidiary
projects/activities funded through the facility, and (4) there is no evidence
that the financing was provided through a cooperation agreement.
Create one record to capture the entire transaction amount and set the
umbrella field to “No.”
■ Scenario 6: If the master loan facility (or line of credit) is provided to the
Central Bank of the recipient country. Create one record to capture the
entire transaction amount and set the umbrella field to “No.”

Section 3 - TUFF 3.0 Data Collection Process

The TUFF 3.0 data collection process is undertaken in three stages: (1) project identification, (2)
project verification and enhancement, (3a) project-level data quality assurance, and (3b) quality
assurance of the dataset as a whole. In this section, we document each of the stages, which
were followed to construct the 3.0 version of AidData’s GCDF dataset.

3.1 - Stage 1: Identifying New Projects and Sources

The objective of Stage 1 is to first identify the universe of Chinese ODA- and OOF-financed
projects and/or activities.67 This is done on a recipient country-by-recipient country basis, so
each step of Stage 1 is repeated for each of the 165 countries included in the 3.0 dataset.

67
In early iterations of the TUFF methodology, AidData relied on global databases of news reports
(Factiva, DNA) to identify Chinese ODA- and OOF-financed activities during stage 1. Then, during the
project verification and enhancement stage (Stage 2), AidData would use additional sources to find and
verify project details. These additional sources in Stage 2 often included “official” sources—such as
official publication from Chinese government agencies or data and documentation from recipient
government agencies. However, the 3.0 version of the TUFF methodology takes a different approach. In
Stage 1, we use a large catalogue of official sources (that covers 165 countries) to identify projects, and
then supplement the list of Chinese ODA- and OOF-financed projects that are identified via official
sources with media-based sources to identify any remaining missing projects.

93
Stage 1 is completed in three steps:
1. Coders review a catalogue of official sources that AidData faculty and staff have
assembled in order to (a) identify projects/activities that are supported by official sector
institutions in China and consistent with OECD-DAC definitions of ODA and OOF; and
(b) document any basic/foundational information about these projects/activities (e.g.,
funding agency, receiving agency, commitment year, transaction amount) that is
specified by these official sources. Coders then create a unique record in AidData’s
data management platform for each project/activity with a unique identification number
(‘AidData Record ID’) and populate as many fields as possible for those records with the
information that is provided by the official sources.
2. Coders review a fixed and pre-processed set of media articles from Factiva/DNA in
order to identify (a) any additional projects/activities that are supported by official sector
institutions in China and consistent with OECD-DAC definitions of ODA and OOF; and
(b) any additional details about the projects/activities exclusively identified via
Factiva/DNA and the projects/activities jointly identified by official sources and
Factiva/DNA.
3. AidData faculty and staff conduct supplemental searches to identify any major sources
of Chinese ODA or OOF—and specific Chinese ODA- and OOF-financed projects— that
were not identified during the first two steps of Stage 1.

A description of each of these steps is detailed in the next section.

3.1.1 - Official Country Profiles

Before Stage 1 is initiated, AidData faculty and staff create or update an Official Country Profile
(OCP) for each recipient country. The OCP is a catalogue of all known official sources that may
provide information about Chinese ODA- or OOF-financed projects and/or activities in a given
recipient country. Each OCP identifies websites, documents, and datasets from official sources,
such as the Chinese government, the recipient government, and official sector entities with
international aid and debt monitoring responsibilities (e.g., the World Bank and the IMF). On
average, each recipient country’s OCP includes around 100 sources.
● Each OCP includes key Chinese government sources, such as the Chinese Embassy and
Economic and Commercial Counselor (ECCO) websites in the given recipient country,
MOFCOM investment guides, and the annual reports of Chinese state-owned banks
and state-owned companies. These sources typically demonstrate that a project/activity
exists; provide precise official commitment dates and project implementation start and
end dates (e.g., the calendar days on which the loan agreement was signed and
construction started/ended); identify a official project title (in Mandarin Chinese); and
provide information about the funding agency, the receiving agency, and/or the nature
of the flow type (e.g., a preferential buyer’s credit from China Eximbank was issued to
the Ministry of Finance in the recipient country to support the project).
● Every country’s OCP identifies key recipient government sources, such as the Ministry of
Foreign Affairs website, the debt registry and budget documents of the Ministry of
Finance, government registers and gazettes that publish information about foreign loan
and grant agreements, the government’s aid and debt information management
system, and the websites of legislative and executive oversight institutions in host
countries (e.g. Public Accounts Committee, Office of the Auditor General). These
sources often identify official commitment dates, funding agencies, receiving agencies,
transaction amounts, borrowing terms, information about the timing and value of

94
 disbursements, information about project implementation progress (including but not
limited to construction start and end dates), and official project titles in local languages.
● Official sources that are not from the Chinese government or recipient
government—like International Monetary Fund’s (IMF) Article IV Staff Reports—also
appear in each country’s OCP.
● Additional sources include major implementing organizations and Chinese state-owned
enterprises operating in the recipient country, as well as government-affiliated media
and/or major media outlets in the recipient country. These sources typically provide
supplemental information about a project’s implementation progress.

Coders are instructed to take a source-specific approach to data collection, which means that
they retrieve information from one official source at a time, compiling an initial project list that
is de-duplicated as they review additional official sources

Within each OCP, AidData faculty and staff provide coders with specific descriptions of each
official source and source-specific instructions, which is important because the sources that are
identified in the OCP often contain a great deal of information that is not related to Chinese
ODA- or OOF-financed projects/activities.68 Therefore, AidData faculty and staff review each
source in advance and specify which particular sections require the attention of coders.
● Source-specific instructions also include guidelines for navigating websites, documents,
and datasets, as well as tips for conducting searches in foreign languages:
○ E.g. “肯尼亚” is the Chinese name for Kenya
○ E.g. Google this search term: “http://dj.china-embassy.org/chn/” 贷款 (Loan in
chinese)
○ E.g. Use Ctrl+F in this French-language document to search for “Chine” and
“Chinois”
● Coders are instructed to download especially useful sources and add them to OCPs for
future rounds of data collection (e.g., a time-stamped export of a recipient country’s Aid
Management Platform).

3.1.2 - DNA/Factiva Articles

After AidData coders conduct a systematic review of official sources that provide information
about Chinese ODA- and OOF-projects/activities in each recipient country, they search for
additional projects/activities and project-level information through targeted searches in Factiva
and Dow Jones DNA69. Factiva—a Dow Jones-owned media database that draws on
approximately 33,000 media sources worldwide in 28 languages, including newspapers and
radio and television transcripts—is the primary database that AidData has historically used for
the systematic review of media articles that report on Chinese ODA- and OOF-financed

68
AidData researchers update the OCPs on an annual basis as additional sources are identified or
become available. OCPs for the 2.0 dataset contained about 25 sources each, wheres OCPs for the 3.0
dataset include over 100 sources on average. The latest OCPs also contain archived records from
previous OCPs so that country-specific data collection information, advice, and challenges can be
passed on to future coding teams. noffi
69
Factiva and Dow Jones DNA sources often provide coverage of smaller grants, in-kind contributions,
technical assistance and medical team projects. The articles help identify or confirm implementation
details of potential projects (e.g., dates of signing or handover ceremonies, officials and ministries
present at these ceremonies, and other organizations or contractors involved in the project).

95
projects/activities. However, as the scope of our data collection efforts has expanded to a
larger number of recipient countries and a wider set of commitment years, we have turned to
Dow Jones DNA to more efficiently extract and process media articles when our Factiva
queries return more than 1,000 results for a single recipient country in a single commitment
year. Whereas Factiva was not designed to support machine learning applications, Dow Jones
DNA—a cloud-based content processing and storage platform—makes the entire, 30-year
Factiva archive and approximately 1 million income news articles per day accessible to users
who wish to use this information in machine learning applications.

We use a standardized set of search criteria to query Factiva and Dow Jones DNA.70 The
queries generate a long list of media articles, but only a subset of these “candidate sources”
contain information about Chinese ODA- and OOF-financed projects/activities. We therefore
use a machine learning algorithm71 to identify the subset of DNA articles that are most likely to
contain information about Chinese government-financed projects/activities.72 We refer to this
machine learning tool as the “TUFF Robot.” It combs through millions of search results at a rate
of approximately 115,000 results per hour—or 2.7 million results a day. It categorizes search
results as either “relevant” or “irrelevant” based on whether they seem to contain information
about Chinese ODA- and OOF-financed projects/activities. AidData coders then review each of
the Factiva articles returned by the query and Dow Jones DNA records that the machine
learning algorithm has classified as “relevant” and make case-by-case determinations about
whether those sources do indeed contain information about Chinese ODA- and OOF-financed
projects/activities.

For each commitment year, AidData coders review approximately 32,000 articles from Factiva
and 16,000 articles from Dow Jones DNA (across the full set of 165 countries in the dataset)73.

70
All of these queries rely on a standardized set of keywords (such as grant, loan, and donate), but we
run them independently for each recipient country.
71
To classify the documents, the machine learning software uses the LGBMClassifier from the lightgbm
package, which is a gradient boosting model developed by Microsoft, with balanced TRUE/FALSE files
as the training set. To balance these files a targeted artificial data augmentation library was used
(NLPAUG), which slightly altered a random selection of the existing TRUE files by replacing a number of
words in each selected document with synonyms to generate enough TRUE files to match the number of
FALSE files.
72
To train the machine learning tool, we use large amounts of training data (articles that we identified via
Factiva/Dow Jones DNA and then classified as containing or not containing information about projects
financed by the official donor/lender of interest) to “teach” the algorithm to accurately classify hundreds
of thousands of articles into “relevant” and “irrelevant” categories. Use of this tool significantly reduces
the amount of time that researchers would otherwise spend reviewing false positives—articles that
contain no information about projects financed by the official donor/lender of interest. To continuously
improve the accuracy of the TUFF Robot in classifying articles as either relevant or irrelevant, we update
the training set each data collection cycle with all new articles that our team of coders manually sorted
during the previous data collection cycle.
73
In cases where Factiva returns less than 1,000 search results for a single recipient country in a single
commitment year, we prefer to use this database for the systematic review of media articles. This
database is designed to have a user-friendly interface which is ideal for this type of manual review
process, and coders can easily manage the systematic review of less than 1,000 search results in this
interface while working in a time-efficient manner. Performing a systematic review of thousands of search
results per recipient country per commitment year would be too costly and less than ideal for this kind of
interface, which is why in cases where Factiva returns more than 1,000 search results for a single recipient
country in a single commitment year we employ the use of Dow Jones DNA and the TUFF Robot.

96
Across 22 commitment years (2000-2021), this amounts to over one million articles. In order to
conduct this manual review process, the coder assigned to the OCP for a specific recipient
country will also review the news article search results for that specific country. To conduct this
review in a time efficient manner, coders are advised to scan article titles, and if the title
indicates the possibility that the article may contain information on Chinese ODA- or
OOF-financed activities in the recipient country, the coder will then scan the text of the article.
If the coders identify information related to Chinese ODA- or OOF-financed activities in the
text of the article, they conduct a more thorough review and ensure the relevant information
from the news report is added to the relevant project/activity records in AidData’s data
management platform. In cases where multiple news reports provide the exact same text (e.g.,
republished stories from the Associated Press), coders are instructed to include only attach one
of these sources to the relevant record, with the objective of identifying the article from the
original news outlet that reported the information, or otherwise from the donor/creditor
country news outlet or recipient country news outlet that would be more closely tied to the
project/activity.

When these news databases are queried, AidData coders search for media reports that include
(a) some derivation of the name of the country/government from which the financial or in-kind
transfer originates; (b) some derivation of the name of the country/government to which the
financial or in-kind transfer is directed; and (c) at least one keyword related to financing or
development projects, such as “grant,” “loan,” or “medical team.” An illustrative set of search
terms for Chinese ODA- and OOF-financed activities in Chile is provided below:

(China or chinese or chin*) near5 (Chile or Chilean or Chile* or Santiago) AND (assist* or grant*
or loan* or lend* or lent or donat* or donor* or fund* or invest* or financ* or economic
package or development package or aid or scholarship* or capacity building or training* or
joint* near5 train* or train* near5 program* or technical cooperat* or exchange* or medical
team* or experts or provid* or provision* or support or debt* near5 forgive* or debt* near5
relie* or debt* near5 cancel* or export credit* or mixed credit* or buyer* credit* or disburse* or
feasibility stud* or relief effort* or disaster relief or humanitarian relief or emergency relief or
relief supplies or relief materials or sign* near5 agreement)

The TUFF Robot uses a similar query for Dow Jones DNA but only includes some derivation of
the name of the country/government from which the financial or in-kind transfer originates.

3.2 - Stage 2: Record Enhancement and Verification

In Stage 2, AidData coders populate as many “empty” dataset fields (i.e., financing and
implementation details) as possible for each project/activity record that was identified in Stage
1. They also seek to corroborate key project details by cross-checking them with new sources.
Additionally, in Stage 2, AidData coders identify and remove duplicate project/activity records.

3.2.1 - De-duplication and Detailed Searches

Stage 2 is undertaken in four steps:
1. Coders review the project information collected during Stage 1 and conduct duplicate
checks to ensure that newly-generated project/activity records capture new/unique

97
 projects/activities that are not already captured elsewhere in the data management
platform.
2. Coders review and seek to verify the coding and categorization determinations that
were made during Stage 1 with the same set of sources that were identified during
Stage 1. They also review the “Staff Comments” field to identify key information gaps
that need to be addressed.
3. Coders conduct targeted searches with English, Mandarin Chinese, and local language
terms. These searches include the following steps:
a. With English and Mandarin Chinese terms, search Chinese government
sources74 to verify the existence of the project/activity, dates related to key
variables, the flow type, the transaction amount, and the official project title in
Mandarin Chinese;
b. With English and local language terms, search recipient government sources75 to
verify transaction amounts, borrowing terms, loan disbursements, locational
details, and the official project title in local language of the recipient country;
c. With English, Mandarin Chinese, and local language terms, search implementing
agency sources76 (e.g. annual reports of Chinese state-owned enterprises) to
collect locational details and up-to-date information about the implementation
status/progress of the project;
d. With English, Mandarin Chinese, and local language terms, search additional
official sources (e.g. IMF Article IV report) to cross-check and verify the
transaction amount, the commitment date, loan pricing details, and loan
disbursement information; and
e. With English and local language terms, search local media to collect information
about project implementation progress
4. Coders fix broken sources (hyperlinks) that were identified during Stage 1. They also
update the Staff Comments field to flag unresolved discrepancies and key information
gaps that require the attention of Stage 3 (Quality Assurance) coders.

As coders conduct these searches and identify new information, they systematically populate
126 fields for each record. That process and the field-specific coding instructions that are given
to AidData coders are detailed in the next section.

74
One example of using Chinese search terms to identify information from Chinese government sources
is as follows: “site:mofcom.gov.cn 中加友谊体育场贷款." This combination of search terms, which
includes the url of a Chinese government site, the project name in Chinese (Sino-Gabon Friendship
Stadium), and the word “loan” in Chinese, will yield information about loans provided to fund the
Sino-Gabon Friendship Stadium.
75
One example of using local language search terms to identify information from recipient government
sources is as follows: “site:gub.uy acuerdo de cooperación económica y técnica China 2009." This
combination of search terms, which includes the url of a recipient government site, the type of
agreement in Spanish, the word “China” in Spanish, and the year the agreement was signed, will yield
information about the Economic and Technical Cooperation agreement signed between the Chinese
government and Uruguayan government in 2009.
76
One example of using English language search terms to identify information from implementing
organizations’ sources is as follows: “site:gwcl.com.gh China Kpong Water Supply Expansion Project."
This combination of search terms, which includes the url of a local contractor’s web page, the name of
the financier, and the project name in English, will yield information about the Kpong Water Supply
Expansion Project financed by China Eximbank in Ghana.

98
3.2.2 - Dataset Variables and Structure
As part of Stage 2, AidData coders seek to verify existing information and collect new
information in order to accurately populate all 126 fields (variables) in the 3.0 dataset for each
record. See Section 1.4 for a complete set of field names and definitions. AidData coders are
instructed to adhere to the following guidelines in order to populate these fields.

● Umbrella marker (Yes/No): There are two primary reasons a project and/or activity
record’s “Umbrella” field can be set to Yes: (1) to avoid double counting commitment
amounts across the dataset; and (2) to capture an overarching agreement/pledge of
funding that ultimately leads to multiple, subsidiary projects (such as a master loan
facility, an Economic and Technical Cooperation Agreement (ETCA), or a framework
agreement). The most common types of umbrella projects involve (a) debt forgiveness
of loans that were contracted (or may have been contracted) during the 2000-2021
period (and thus may be captured elsewhere in the dataset); and (b) grants/interest-free
loans issued through ETCAs for unspecified purposes/projects (which since multiple,
unknown projects may have been financed through these agreements, which creates a
risk of double-counting of other projects and commitment amounts). There are also a
few exceptions where AidData coders confirmed the complete disbursement of a large
line of credit, but could not identify the full range of sub-projects.77 Umbrella records
should be linked to related project/activity records through their title field and
description fields by a reference to their unique identification numbers (‘AidData Record
ID’). They should also be linked by an AidData Parent ID (or IDs).
● Commitment Year: The commitment year field captures the year in which an official
financial commitment (or official commitment to provide in-kind support) was codified
through the signing of a formal agreement by an official donor/lender in China and one
or more entities in a recipient country or a set of recipient countries. Whenever
possible, this field is based on the precise calendar day when the official commitment
was issued, which is captured in the 'Commitment Date' field. However, in some cases,
the official commitment date is unknown. In such cases, the commitment year is based
on whichever proxy is used for the commitment date. The process for identifying a
proxy date is explained in the ‘Commitment Date/Commitment Date Estimated’
section. For projects/activities that are assigned to the Pipeline: Pledge status category,
the commitment year will reflect the year in which the informal pledge was announced.
● Commitment Date/Commitment Date Estimated: The commitment date field captures
the date on which an official donor/lender in China and one or more entities in a
recipient country or set of recipient countries makes an official financial commitment (or
official commitment to provide in-kind support) by signing a formal agreement. When
the month, day, and year of the formal signed commitment are all known, it is recorded
in the commitment date field. If the precise month, day, and/or year is unknown,
AidData use the following proxies to estimate the commitment date: (1) The first day of
a known month in a known year if we do not know the day (i.e. 04/01/2017); (2) January
1st of a known year if we do not know the month or day (i.e. 01/01/2017); (3) the first
day, first month, and first year of a central bank’s financial fiscal year (i.e. the State Bank
of Pakistan’s financial fiscal year is from July 1 to June 30 so AidData would code July 1

77
In these rare and specific cases (Project ID#66806 and #40135), AidData coders have marked the main
commitment/disbursement project as a non-umbrella project, and removed the transaction amount from
any individual subsidiary projects to avoid double-counting.

99
 of the first calendar year as the estimated commitment date) if the fiscal year is known
but the month, day, and year of the signed financial agreement is unknown; (4) the first
year of project implementation if the month, day, and year of the signed financial
agreement are unknown; (5) the year in which the underlying commercial contract
(supported by the official financial commitment) was signed if the month, day, and year
of the signed financial agreement are unknown; (6) the first year in which an informal
pledge was made if the month, day, and year of the signed financial agreement are
unknown.
○ If any of these proxies are used to estimate the commitment date, the
“Commitment date estimated” field is set to “Yes.” For projects that are
assigned to the Pipeline: Pledge status category, AidData coders are instructed
to record the date on which the informal pledge was announced.
● Title: In this field, AidData coders are instructed to include the following information
when they create or edit a record title: the funding agency name, the flow type (e.g.,
loan, grant, technical assistance), the commitment amount (if applicable), and the formal
English-language project title. If a formal English-language project title is unavailable,
the purpose of the transfer is identified. Example: “China Eximbank provides $267.2
million preferential buyer's credit for 500 kV North-South Power Transmission Line
Project."
● Description: AidData coders are instructed to include the following information when
they are creating or editing the description field:
○ Basic project/activity information. Coders should identify the funding agencies,
direct and indirect receiving agencies, the amount of funding they officially
committed for the project/activity, the currency of denomination, the type of
financial agreement that was signed, the official commitment date, and a
description of the purpose of the funding. Sometimes this is straightforward
(e.g., a grant to build a hospital). However, at times, the financial commitment
indirectly supports a project/activity (e.g., a loan is issued to a company which
then invests in a project/activity). In such cases, coders are instructed to clearly
document how the financing is used, and for what purposes.
○ Financial details. Coders should identify the specific terms and conditions that
govern the financing agreement, the timing and monetary value of commercial
contracts and financial disbursements, and the existence of any special
arrangements (such as on-lending agreements or offshore escrow accounts into
which project revenues must be deposited). They should also document any key
calculations or assumptions, such as the way in which an “all in” interest rate was
measured. For example: “The CDB loan had a variable interest rate of 6-month
LIBOR + a 2.90% margin. The interest rate of the loan that supported this
project is coded based on the 6-month LIBOR at time of signing of the loan
agreement. The average 6-month LIBOR rate in June 2012 was 0.736%. The
interest rate was therefore coded as 0.736% + 2.90%, or 3.636%.”
○ Physical implementation details. Coders should provide a description of major
project/activity events and the precise calendar dates when they took place
(e.g., commercial contract signing date, implementation start date, project
completion date, commercial operations date, and major milestones such as
when a project passes a midterm or final inspection by the funding agency).
Additionally, coders should identify all implementing agencies, co-financing
agencies, receiving agencies, or accountable agencies.

100
 ○ Financial implementation details. Whenever possible for loan records, coders
should record loan disbursement rate, debt repayment schedule, outstanding
debt amount, any events of default, maturity extension, and point to any debt
cancellation or debt restructuring records related to this project/activity. (e.g.,
The $2.9 billion USD syndicated facility was fully drawn (disbursed) at its 2009
inception. In 2012, Yancoal Australia repaid $100 million USD of the $2.9 billion
USD facility to CDB. In 2013, it repaid $100 million USD; in 2014, it repaid $99
million USD. As of December 31, 2015, the outstanding balance was down to
$2,600,000,000 USD. In 2017, Yancoal repaid $150 million USD under the
syndicated facility, reducing the balance to $2.450 billion USD. On September
17, 2018, it repaid $75 million USD; on October 17, 2018, it repaid $50 million
USD, lowering the outstanding balance to $1.525 billion USD.)
○ Geographical information. Coders should record geographical details that
accurately and precisely document project’s physical footprint, including location
names and types; the position or distance of the project’s location vis-à-vis other
geographical features (e.g., the building is located across the street from the
country’s parliamentary complex in the capital city); the name, length, and start
points and end points of physical infrastructure supported by the project (e.g.,
the 115.85 kilometer A1 highway runs from Colombo to Kandy); the total land
area occupied by the project site (e.g., the industrial park occupies a 10 square
kilometer area); and latitude and longitude coordinates of specific project
features (e.g., the coordinates of Olkaria IV Geothermal Power Station Kenya at
Hells Gate National Park are -0.918056, 36.334444).
○ Name the sources. When key project/activity details are provided in the
description (e.g., transaction amounts, commitment years), coders should
identify the specific source or sources which provide such information. Example:
“According to the 2017 Annual Debt Report published by Sierra Leone’s Ministry
of Finance, the Government of Sierra Leone signed a government concessional
loan (GCL) agreement with China Eximbank on March 13, 2016 for a National
Broadband Project that carries the following borrowing terms: 20 year maturity,
5 year grace period. 2% interest rate.”
○ Risks, Achievements, Failures, and Setbacks. Whenever possible, coders should
include a detailed overview of the various challenges that arose during project
design and implementation (such as controversies, strikes, riots, public protests,
wars, corruption scandals, natural disasters, public health restrictions, political
transitions, bankruptcies, debt defaults, contractual disputes, lawsuits, and
ruptures in diplomatic relations) and how funding, receiving, implementing, and
accountable institutions responded to these challenges. Additionally, the
description should include information about project achievements and failures,
contractor performance vis-à-vis deadlines and deliverables, and findings from
project audits and evaluations.
● Staff Comments: This field is used to ​identify the assumptions, logic, and evidence that
coders used to address coding and categorization determinations. It also provides
contextual information and source materials that users (and future AidData coders) may
find helpful if and when they seek to collect supplementary information about the
project/activity (or revisit previous coding and categorization determinations). More
specifically, this field seeks to document:
○ Discrepancies between sources: Coders should explicitly identify discrepancies
across different sources when they relate to key fields (such as the commitment

101
 year, the transaction amount, the funding agency, the receiving agency, and
loan pricing details). They should also specify how they adjudicated between
competing sources and resolved discrepancies. Example: “In the database of
Chinese loan commitments that SAIS-CARI released in July 2020, it identifies a
$264 million China Eximbank loan in 2009 for this project. However, AidData
relies on the face value of the CMEC supplier’s credit ($551,507,000) and the
commitment year (2005) of the supplier’s credit agreement that is specified in
the agreement itself. It is possible that CMEC obtained an export seller’s credit
from China Eximbank to finance its supplier’s credit with the Republic of Congo;
however, AidData has not yet independently confirmed that this occurred. [...]
The CMEC supplier's credit agreement can be accessed in its entirety via
https://www.dropbox.com/s/g4bkuq71ezrpjoy/L%20n%C2%B019-2019%20du%
2024%20mai%202019.pdf?dl=0 or
https://www.documentcloud.org/documents/20488096-cog_2005_490.”
○ Project titles in English, Chinese, and local languages. Example: This project is
also known as the Third Bridge Construction Project. The Chinese project title is
马里共和国巴马科第三大桥 or 援马里巴马科第三大桥项目. The French project
title is La construction du 3ème pont de Bamako or Pont de l'amitié
sino-malienne.
○ Any assumptions, logic, and evidence used to calculate the transaction amount,
interest rate, or other financial detail variables. For Example: “The individual
contributions of China Eximbank, and Credit Suisse AG to the syndicated loan
are unknown. For the time being, AidData has estimated the contribution of the
China EXIM bank by assuming that the two lenders contributed equal amounts
($42,000,000) to the loan syndicate.” or “The individual contributions of the
syndicate members are unknown. For the time being, AidData has estimated the
contribution of the four Chinese state-owned banks by assuming that the
thirteen lenders contributed equal amounts ($76,923,076.92) to the loan
syndicate.”
○ Specific justifications for challenging coding and categorization determinations.
Example: “AidData has coded this transaction as a collateralized loan because
ICBC was selected as the security agent (i.e. collateral agent) for the loan. When
lenders take collateral as security for their loans, a collateral/security agent is
often appointed to enforce rights against the collateral in the event of the
borrower’s default under the loan.”
○ Issues that were not fully resolved and/or that require further investigation.
Example: “​​According to the financial reports published by JP EPS, it contracted
a $35,938,868.58 loan with the Chinese Government on June 25, 2010 to
finance the imports of goods and services—including transformers, transmission
lines, conveyors and bulldozers—from China. This loan also has an 11 year
repayment period (between July 21, 2010 and January 21, 2010) and it also
carries an interest rate of 6-month LIBOR plus a 1.3% margin. It is unclear if this
loan is distinct from the China Eximbank loan that was rescheduled on February
20, 2009. For the time being, in order to err on the side of caution, AidData
does not record a separate loan to capture the loan that JP EPS reportedly
contracted on June 25, 2010. However, this issue merits further investigation.”
○ Hyperlinks to uniquely important sources, such as unredacted EPC contracts,
unredacted grant and loan agreements, official correspondence between
lenders and borrowers, and direct correspondence between AidData and

102
 government officials in recipient countries. Example: “The 2011 prepayment
facility agreement between PetroChina and PetroEcuador can be accessed in its
entirety via
https://www.dropbox.com/s/gfl6z608d9abalf/CONTRATO_PRINCIPAL_2011048.
pdf?dl=0.”
● Status: AidData coders are instructed to follow a two-step process to make status field
determinations. First, coders are instructed to determine if an official commitment has
taken place. Second, if an official commitment has taken place, coders are instructed to
determine if the project/activity has reached implementation or completion—or if it was
subsequently suspended or canceled.
○ Guidance on when to assign records to the Pipeline: Pledge, Pipeline:
Commitment, Implementation, or Completed status category:
■ If a commitment from an official sector institution in China took place,
and the project/activity was reportedly implemented or completed,
assume that the financial or in-kind transfer took place (at least in part)
and assign the record to the Implementation or Completion status
category.
■ If a pledge was issued by an official sector institution in China, and the
project/activity was reportedly implemented or completed, do not
assume that the financial or in-kind transfer took place (in part or in
whole). Assign the record to the Pipeline: Pledge status category.
Additionally, if negotiations with an official sector institution in China are
ongoing, keep the record in the Pipeline: Pledge status category. If
negotiations with an official sector institution in China do not result in an
official commitment, keep the record in the Pipeline: Pledge status
category.
■ If a commitment from an official sector institution in China took place but
the financial or in-kind transfer never materialized, and the
project/activity was reportedly completed (with an alternative source of
financing), the record should either be assigned a status code of Pipeline:
Commitment (with an explanation that there is no evidence of
disbursements taking place) or a status code of Suspended or Canceled
(if there is clear evidence that the official sector institution in China
withdrew its support).
■ If a project/activity’s status is unknown, but sources indicate that an
official sector institution from China fully disbursed the funds that it
previously committed to the project/activity (with no reports of
suspension/cancellation), assign the record to the Completion status
category.
○ Guidance on when to assign records to the Suspended or Canceled status
category:
■ Only projects/activities that previously secured a commitment from an
official sector institution in China can be assigned to the Suspended or
Canceled status category. A record that was assigned to the Pipeline:
Pledge category and never secured a commitment from an official sector
institution in China should not be assigned to the Suspended or
Canceled status category (even if it is known that the project/activity was
never carried out). Such records should remain in the Pipeline: Pledge
category.

103
 ■ If a loan agreement was signed but subsequently rejected by the
parliament or judicial body in the recipient countries, then it should be
assigned to the Canceled status category.
■ If a loan agreement was suspended and then a new loan request was
made by the borrowing institution, two separate records should be
created: one for the suspended loan agreement and another for the new
loan request (which should be assigned to the Pipeline: Pledge status
category).
■ If financial disbursements took place prior to the cancellation/suspension
of a project/activity that previously secured a commitment from an official
sector institution in China, coders should record the original financial
commitment amount in one record and assign it to the Suspended or
Canceled status category, and create a separate record to capture
disbursed amount prior to cancellation/suspension and assign it to the
Completed status category.
■ If a loan agreement was issued by an official sector institution in China
but official sources indicate that no disbursements were ever made, the
record should be assigned to the Suspended status category.
■ If no official sources explicitly report a suspension or cancellation of the
financing agreement and information about the project/activity’s
progress is either sparse or absent, coders should seek to identify (1)
evidence of the receiving agency (or another entity in the recipient
country) identifying an alternative source of funding to finance the
commercial contract; (2) evidence of the receiving agency (or another
entity in the recipient country) signing a new commercial contract with a
different contractor; and (3) whether the financing agreement is recorded
from the country’s aid/debt information management system (registry). In
any of these 3 scenarios, it is possible that the financing agreement was
never finalized (i.e., an official commitment never took place), and coders
should assign the record to the Pipeline: Pledge status category.
● Intent: Identifying the intent of a project/activity should be based on three
considerations:
○ Does the project/activity seek to improve the economic development or welfare
in the host country? If so, the project/activity should be coded as having
development intent, which is consistent with OECD-DAC guidelines.
Development intent is determined independently from the concessionality
calculation for loans. A project/activity can have development intent and also be
non-concessional. Any infrastructure projects that can improve economic
development or welfare in host countries that are financed with Chinese debt
and without Chinese equity should be coded as development intent, regardless
of whether the debt is offered on concessional or non-concessionary terms.
Most projects/activities are assumed to have development intent unless there is
specific evidence of commercial, representational, or military intent.
○ Does the project/activity seek to enhance the commercial interests of the
financier country (China)? If so, the project/activity should be coded as having

104
 commercial intent.78 Loans to help shipping companies acquire vessels that will
allow them to move ocean containers from country to country or shipping
equipment should be coded as commercial intent as they are designed to
support the commercial operations of the companies and not to advance an
economic development objective in the host country. A loan to help a company
finance its general operations, or a loan to help a company service its existing
debts should also be coded as having commercial intent. Working capital loans
or "working credit facilities" provide funds for a borrower's day-to-day
operations should be coded as commercial intent. Cross-currency interest rate
swaps are a form of corporate financing and such they represent transactions
with commercial intent.
○ Does the project/activity seek to disseminate or promote Chinese culture,
language, or values? If so, the project/activity should be coded as having
representational intent. Donations of equipment that apparently will be used to
spread Chinese culture, language or values should be coded as representational
intent. This includes donations of Chinese books on Chinese traditions,
donations of lion dance props, and even donations of luxury items from China
(e.g., the Hongqi L5 vehicle). Projects to establish or upgrade Confucius
Institutes and Chinese cultural centers are considered projects with
representational intent. Projects that involve the dispatch of Chinese language
instructors to recipient countries are considered projects with representation
intent.
● Sector Code/Sector Name: Based on the OECD’s 3-digit sector codes and names,
AidData coders should assign each project/activity to the sector that it is meant to
support. If the specific activities of a transaction are unknown, the record should be
assigned to the sector of the receiving agency (i.e., financing with unspecified purpose
to a major oil and gas company should be coded to the energy sector). Records with
unknown activities/purposes and unknown receiving agencies are assigned to the
unspecified sector (998).
● Infrastructure: This flag provides a marker of whether a project/activity is an
infrastructure project. In order to populate this field in the 3.0 dataset, staff isolated
likely infrastructure projects by identifying all projects with title fields or description
fields that included one or more of the following keywords: construc*, build, rehabilit*,
upgrad*, renovat*, exten*, restor*, built, groundbreaking, fiber, power plant, expansion,
electrification, hydro*, instal*, foundation. All projects that are assigned to the following
flow type categories: debt forgiveness, debt rescheduling, scholarships, training, or
free-standing technical assistance activities, as well as umbrella records, were also
excluded. Coders then performed a manual review of the project/activity records that
contained keyword matches to determine whether the infrastructure field should be set
to ‘Yes’ or ‘No’. This was followed by a review of all project/activity records without a
keyword match (though still excluding debt forgiveness, debt rescheduling,
scholarships, training, or free-standing technical assistance activities, as well as umbrella
records).

78
Projects assigned to the “Commercial intent” status category are those that primarily seek to promote
the commercial interests of the country from which the financial transfer originated (e.g., encouraging
the export of Chinese goods and services).

105
 AidData coders were instructed to read project/activity record title and description
fields to make their determinations. More specifically, coders were instructed to set the
Infrastructure field to ‘Yes’ if the project involved one of the following:
○ building a new physical structure
○ rehabilitating or adding onto an existing physical structure, and/or
○ maintaining an existing physical structure.
Coders were instructed to not set the infrastructure field to ‘yes’ when a project/activity
involved the provision of cash, technical assistance, scholarships, equipment, or
supplies.
● COVID: This field provides a marker of whether it is known that the project/activity is
part of China's global COVID-19 response efforts. In the 3.0 version of the dataset, this
field was populated by first using artificial intelligence (AI) to identify all project/activity
records that matched the OECD’s 5-digit sector code for COVID-related projects
(12264). The COVID field for project/activity records that received this sector
designation were systematically set to ‘Yes’. In future iterations of the dataset, AidData
coders will be instructed to evaluate the sources that describe a project/activity for
evidence that the purpose of the project/activity is related to COVID-19 control,
including providing information, education and communication as well as activities or
materials enabling testing, prevention, immunization, treatment, or care.
● Funding agency: Only the official sector institution in China providing the financial or
in-kind support should be identified as a funding agency. Co-financing agencies should
be identified in the co-financing field (regardless of whether they are official sector
institutions from China). The transaction (commitment) amount should correspond to
the financial or in-kind transfer from only the official sector institution in China identified
as the funding agency. If the project/activity was financed by multiple official sector
institutions from China, and the respective financial commitments of each institution are
known, a separate project/activity record should be created for each commitment
amount and corresponding funding agency; all other contributors should be added as
co-financing agencies. If, however, the respective financial commitments of each
institution are not known (but the total commitment amount from all official sector
institutions in China is known), then the Equal Contribution Assumption should be
applied (see ‘Equal Contribution Assumption’ below).
● Co-financing agency/marker: If a project has a co-financier, the co-financing field
(marker) should be set to “Yes,” including in cases where the specific co-financing
agencies are unknown. The co-financing agency name(s), organization type(s), and
origin(s) should also be identified. Counterpart funding from the recipient
agency/company is not considered co-financing.
● Direct and Indirect Receiving Agency: The 'Direct Receiving Agencies' field identifies
the agency designated to receive and manage the financial or in-kind transfer. The
Indirect Receiving Agencies' field provides the name of the agency or agencies that
receive and manage a financial transfer (loan) from the entity captured in the 'Direct
Receiving Agencies' field. If a receiving agency (borrower) on-lends the proceeds of a
loan to an additional entity or entities, then the borrower is captured in the 'Direct
Receiving Agencies' field and the additional entity or entities which receive loans from
the borrower is captured in the 'Indirect Receiving Agencies' field. If more than one
entity is responsible for receiving and managing incoming grant funds or an in-kind
transfer, all of these entities are identified in the 'Direct Receiving Agencies' field. For
seller's credits, the Chinese state-owned enterprise receiving an export seller’s credit
from the Chinese state-owned bank should be coded as direct receiving agencies; and

106
 the foreign customer (borrower) in the recipient country which is in turn receiving the
supplier’s credit from the Chinese state-owned enterprise should be coded as indirect
receiving agencies. When a Chinese state-owned enterprise uses its own funds (without
any known export seller’s credit from a Chinese state-owned bank) to provide a
supplier’s own financing to a foreign customer (borrower) in the recipient country, only
the foreign customer should be coded as the direct receiving agency. If an official
sector institution in China provides a loan to a bank or financial institution in the
recipient country, and that bank or financial institution in-turn uses the proceeds of the
loan to on-lend to state-owned companies or agencies in the recipient country, then
the entity who is on-lending the proceeds of the loan should be coded as the direct
receiving agency; and the entity (or entities) receiving the loan should all be coded as
indirect receiving agencies.
● Implementing Agency: The organizations/entities involved in carrying out the project
should be identified in this field (and in the description field).
● Insurance Provider/Insurance Provided: An insurance provider is a third-party
organization (i.e., not the funding agency or the receiving agency) that issues a credit
insurance policy to ensure repayment in the event that the borrower (i.e. direct receiving
agency) cannot service its debt. A common scenario is one in which Sinosure issues a
credit insurance policy in support of an export buyer’s credit from a Chinese
state-owned bank. However, the insurer need not be an official sector institution in
China. If an insurance provider is identified, the ‘insurance provided’ field should be set
to ‘Yes.’ If it is known that a credit insurance policy was taken out for a loan but no
insurance provider could be identified, the ‘insurance provided’ field should also be set
to ‘Yes’. All credit insurance policies correspond to loans; only project/activity records
with flow types set to ‘loan’ should identify insurance provider(s). Insurance provider
information should also be recorded in the project/activity record’s description field.
● Guarantor/Guarantee Provided: A guarantor is an agency that provides a repayment
guarantee in the event the borrower (i.e. direct receiving agency) cannot meet its debt
repayment obligations. By providing a guarantee, the guarantor is promising to repay
the loan if the receiving agency (primary borrower) defaults on its repayment
obligations. Government agencies from the recipient country may issue a sovereign
guarantee,79 though guarantees can also come from private companies, state-owned
companies, or other types of agencies (in China, the recipient country, or another
country). Coders are provided the following guidance:
○ If a third-party repayment guarantee is issued, the ‘guarantee provided’ field
should be set to ‘Yes’ and the organization(s) providing the guarantee should be
added as agencies, their role set to ‘guarantor’. In the case of a sovereign
guarantee, the guarantor should be the recipient government.
○ If the recipient government issues a sovereign guarantee in support of a loan
issued to another entity, this information should be recorded in the description
field.
○ Only project/activity records with flow types set to ‘loan’ should identify
guarantors and/or set the ‘guarantee provided’ field to ‘Yes’.

79
Sovereign guarantees are legally binding commitments by a sovereign government to assume
responsibility for servicing a debt on behalf of another entity under specific conditions (e.g. default). A
sovereign guarantee represents a contingent liability on the recipient government's balance sheet.
Sovereign guarantees can be provided to public or private entities. Sovereign guarantees are also
sometimes referred to as government guarantees or public guarantees.

107
 ○ The ‘guarantee provided’ field may be set to ‘Yes’ even if no guarantor is
identifiable, provided source(s) indicate this is the case. However, coders should
verify a guarantee was actually provided, rather than collateral.
○ Coders should use caution to prevent mistaking the provision of collateral for a
guarantee and vice versa. They are instructed that:
■ A guarantee is typically not related to a specific asset or revenue stream.
Rather, the issuance of a sovereign guarantee allows the creditor to
secure repayment by pursuing any assets or revenue streams controlled
by the sovereign government in the event of default (assuming the
assets/revenue streams in question are not protected by sovereign
immunity). If a guarantee is mentioned in relation to a specific revenue
stream, it may be a reference to collateral (see section on “Collateralized
Lending” in Section 2.5.3.3).
■ If the word ‘guarantee’ is mentioned in relation to the repayment of the
loan by a non-English language source, this may indicate collateralization
rather than a guarantee as AidData defines it. In languages other than
English, collateralized debt arrangements are sometimes referred to as
‘guarantee’ or ‘guaranteed.’ However, in English, a (third-party)
guarantee is a concept that is distinct from collateralization. Similarly, if a
loan is “guaranteed by” a specific asset, this may indicate
collateralization rather than a guarantee.
● Collateral Provider/Collateralized/Securitized: If an entity pledges one or more sources
of collateral for a loan that can be seized in the event the borrower defaults on its
repayment obligations, it should be identified as an accountable agency. The
‘Collateralized/Securitized’ field should also be set to ‘Yes’. When it is known a loan is
collateralized, AidData coders should identify the exact source(s) of collateral and
populate the ‘collateral’ field, describing the nature of the collateral that was pledged.
However, coders may set the ‘Collateralized/Securitized’ field to yes without knowing
the collateral provider and/or the exact collateral pledged if they cannot be identified.
Only project/activity records with flow types set to ‘loan’ should identify collateral
providers, set the ‘collateralized/securitized’ field to ‘Yes’, and/or list sources of
collateral.
○ In order to identify projects that are collateralized but may not explicitly use the
term “collateral” or deviations thereof, AidData coders should reference
‘Collateralized Lending’ in Section 2.5.3.3.
● Security Agent/Collateral Agent: The security agent or collateral agent is the
organization that was appointed to enforce rights against the collateral in the event that
the borrower defaults on its debt repayment obligations. Only project/activity records
with flow types set to ‘loan’ should include agencies with the role ‘Security
Agent/Collateral Agent’.
● Agency Type: For each kind of agency involved in a project (i.e., funding, cofinancing,
receiving, implementing, or accountable agency), coders are instructed to identify the
agency type. Specific considerations for certain agency types include the following:
○ Government agencies. These agency type designations are given to any entities
that are a part or an extension of the governmental structure in the country
(whether in China or in the recipient country).
○ State-owned policy bank/state-owned commercial bank. These agency type
designations are only used for Chinese state-owned policy banks and
commercial banks that provide overseas funding. The policy banks include

108
 China Eximbank and CDB. The state-owned commercial banks include those
that are at least 50% owned by the Chinese government. The state-owned
commercial banks also include so-called shareholding commercial banks that are
subsidiaries of state-owned enterprises (e.g., China CITIC Bank) and city
commercial banks (i.e., Bank of Shanghai).
○ State-owned bank. This agency type designation is only used in recipient
countries when the bank is at least 50% owned by the recipient government.
○ State-owned company. We consider all companies with the state as their largest
shareholder to be state-owned companies. These include companies that are
wholly-owned or majority-owned or partially-owned by the state. Wholly-owned
subsidiaries of state-owned companies are also coded as state-owned. This
definition holds for Chinese and recipient state-owned companies.
○ Private sector. This agency type designation encompasses all companies with no
state ownership or where the state was not the largest shareholder.
○ State-owned fund: This agency type designation is only used for funds set up
and financed by Chinese governmental and state-owned banking entities.
Certain funds are region-specific, such as the Africa Growing Together Fund, and
the China Co-financing Fund for LAC. Some funds may serve a specific purpose,
such as the Silk Road Fund, that is dedicated to providing financing in countries
participating in the Belt and Road Initiative. Others are set up to support
intergovernmental organizations, such as the People’s Republic of China Poverty
Reduction and Regional Cooperation Fund, and China Trust Fund.
○ Special purpose vehicle/joint venture. This agency type designation captures
project companies (independent legal entities) that are established to manage
the financing and implementation of a particular project. Owner organizations,
when known, should be attached to the SPV/JV’s organization record with their
ownership stake documented. This information should also be noted in the
description field of the record(s) associated with the SPV/JV and in the
organization’s description field.
○ Intergovernmental Organization. This agency type designation captures
organizations made up of governments from multiple countries. Coders should
ensure funding agencies do not fall in the intergovernmental organization
category.
○ NGO/CSO/Foundation. This agency type designation captures
non-governmental organizations, civil service organizations, and foundations.
AidData coders are instructed to ensure funding agencies do not fall into the
NGO/CSO/Foundation category, as these fall outside the scope of official sector
financing.
○ Miscellaneous Agency Type. This agency type designation captures all other
agencies that do not fit in one of the above categories. AidData coders are
instructed to use this designation sparingly.
● Agency Origin: For each kind of agency involved in a project/activity (i.e., funding,
cofinancing, receiving, implementing, or accountable agency), coders are instructed to
identify the agency origin. The agency origin is no longer captured in a separate field.
However, it is reflected in the ‘Agency Type’ fields (i.e. ‘Guarantor Agency Type’,
‘Implementing Agencies Type’, etc).
○ For example, for a record where a state-owned enterprise involved in project
implementation has its origin set to China, the ‘Implementing Agencies Type’
variable will be set to ‘Chinese state-owned enterprise’.

109
 ○ Specific considerations for certain agency types include the following:
■ The agency origin field should be set to China if the agency is
wholly-owned by the Chinese government or a Chinese company.
■ The agency origin field should be set to “recipient country” if the agency
is wholly-owned by the government or companies in the recipient
country.
■ The agency origin field should be set to “other” if (1) the organization is
wholly-owned by an entity not from the recipient country or China; (2)
the organization is an intergovernmental organization; or (3) the
organization is partially owned by an entity from China and an entity from
the recipient country.
● Transaction Amount: In most circumstances, only the face values of loans and grants (or
the monetary values of in-kind transfers) from official sector institutions in China should
be recorded as transaction amounts. When an official commitment amount is identified,
AidData coders are instructed to record it as the official transaction amount even if the
disbursed amount is different from the official committed amount. However, if an official
commitment amount is not provided and a disbursement amount is provided, AidData
coders are instructed to record to use the disbursement amount as a proxy for the
official commitment amount.80 However, this coding rule only applies to completed
projects.
○ There are cases where a separate project/activity record should be created for
the sole purpose of capturing disbursements: (1) disbursement (drawn down
amount) for a line of credit that has been coded as umbrella project - see section
on Master Loan Facilities/Agreements and Credit Lines (see Section 2.1.3.2), and
(2) disbursements made prior to a project cancellation (one project/activity
record should already be created for the cancelled project).81
○ If a coder is not certain that an official sector institution in China fully financed
the entire project/activity, the total project/activity cost should not be recorded
as the transaction amount. The transaction amount should only capture the
financial amount provided by the official sector institution in China.
○ If there are two transaction amounts in a single project/activity (e.g. two
disbursements in different currencies), coders are instructed to create two
separate project/activity records. The title field and description field should
identify the project/activity records as linked through their unique identification
numbers (‘AidData Record ID’) and should further share an AidData Parent ID.
○ For transaction amount coding practices associated with specific lending
instruments, see Section 2.5.3.3.
● Amount Estimated: For the vast majority of project/activity records in the 3.0 dataset,
the reported transaction amount is based on information in the primary (hyperlinked)
source materials. However, there are a few unique scenarios in which transaction

80
Sometimes there are cost overruns on projects and the Chinese government ends up providing top-up
funding beyond the original commitment amount to address these cost overruns. So, in effect, a higher
disbursement figure compared to the original commitment amount serves as a proxy for the final official
commitment amount after the top-up funding is included. There is often an agreement
revision/addendum to officially increase the face value of the loan (but we usually don’t have access to
these agreement revisions/addenda).
81
We do not count the face value of a line of credit (or master loan facility) as an official commitment
amount if it is not fully utilized by the borrower.

110
amounts are estimated by AidData coders using information that is provided in the
project/activity record itself. In these cases, the ‘Amount Estimated’ field should be set
to ‘Yes’ and a Staff Comment should be added explaining how the estimation was
generated.
○ Equal Contribution Assumption for Syndicated Loans
● If the members of a loan syndicate are known but their individual
contributions are not known, AidData coders are instructed to divide the
total value of the syndicated loan by the total number of known
syndicate members. This value is then used as the (estimated) transaction
amount for the individual project/activity record capturing the
contribution of an individual syndicate member. This is done to avoid
artificially deflating commitment aggregates, as would occur if the
Chinese official financing was coded as zero for these projects/activities.
When applying the equal contribution assumption, coders should leave a
staff comment explaining the calculation used.
○ Example, AidData Record ID #89485: "The exact size of Bank of
China and ICBC’s respective financial contributions to the $8
billion syndicated bridge loan are unknown. For the time being,
AidData assumes that all 19 members of the lending syndicate
contributed equal amounts ($421,052,631)." Coders should add
one Chinese bank for that project entry as the funding agency (in
this case, Bank of China) and all others as the co-financing
agencies (see the linked Record #89485 for an example).
● If coders know some but not all syndicate members, and know the total
value of the loan, coders are instructed to estimate individual bank
contributions to the syndicated loan based on the total number of known
members of the syndicate – even if there are only two or three.
○ For example, in AidData Record ID #98335, it is known that at
least twelve banks participated in this syndicate, but the exact
number of participants is unknown. AidData applies the equal
contribution assumption, and a staff comment notes: “AidData
assumes that the twelve known syndicate members each
contributed an equal amount to the syndicate (USD 20,833,333)”.
● We will not apply the equal contributions assumption if we do not know
the total transaction/commitment value of the syndicated loan.
● For each tranche of a loan, the equal contribution assumption is applied
independently. This means that if a Chinese bank is not involved in a
particular tranche, we do not count it for the purposes of equal
contribution assumption calculations.
● If multiple branches of the same bank are participating in a syndicated
loan, each branch counts as a distinct lender. For example, a syndicate
loan with DBS Bank and three different branches of ICBC has 4 members
of a syndicate, not 2. The same rule applies for subsidiaries, e.g. ICBC
London PLC and ICBC are two distinct lenders.
○ This impacts how the transaction amount is coded for Chinese
banks. If there are three branches of the same Chinese bank, your
assumed transaction amount should be the sum of all the
branches' loans.

111
 ■ For example, in AidData Record ID #52623, there are two
separate ICBC members [ICBC Shanghai Pilot Free Trade
Zone Branch and Industrial and Commercial Bank of China
Limited (ICBC)] participating in the syndicate. We code the
transaction amount as the combined contribution from
both the Shanghai Pilot Free Trade Zone Branch and the
main ICBC, like so: “The individual financial contributions
of ICBC Shanghai Pilot Free Trade Zone Branch, Industrial
and Commercial Bank of China Limited (ICBC), the
Export-Import Bank of China to the syndicated loan are
unknown. For the time being, AidData assumes equal
financial contributions ($161,666,666) by the 6 known
members of the loan syndicate. The transaction amount
for ICBC is recorded as $323,333,333, as ICBC Shanghai
Pilot Free Trade Zone Branch and Industrial and
Commercial Bank of China Limited (ICBC) are two
separate members of the loan syndicate. This issue
warrants further investigation.”
● In cases where the Equal Contribution Assumption is applied, AidData
coders are instructed to set the ‘Amount Estimated’ field to ‘Yes’.
○ Imputed Transaction Amounts for In-Kind Donations of COVID-19 Supplies
■ In the wake of the Coronavirus pandemic, Chinese official financiers
began providing in-kind donations of supplies for preventing or
mitigating the spread of the COVID-19 virus to recipient countries. This
included donations of Personal Protective Equipment (PPE), medical
devices, and diagnostic tools. Per the OECD, “Aid in kind…should where
possible be valued at prevailing international or national market prices
for the goods in question at the time of the transfer.”82 Therefore, in
order to both mitigate undercounting 2020 and 2021 financial
commitments and bring our dataset further in line with OECD guidelines
vis a vis in-kind donations, AidData has taken steps to impute transaction
amounts for in-kind donations of COVID-19 related supplies where
possible.
■ To do this, AidData identified per-unit prices – specific to the month-year
level – for fourteen commonly donated types of supplies using the World
Health Organization’s (WHO’s) Emergency Global Supplies Catalogues.
These catalogues were produced for the WHO’s COVID-19 Supply Chain
System (CSCS) Supply Portal and contain lists of purchasable supplies
and their estimated per-unit costs.83 These estimates were regularly
updated, allowing AidData to account for changes in per-unit price over
time. In cases where a catalogue was not available for a particular month,
the prices from the preceding catalogue most recently available were

82
See https://one.oecd.org/document/DCD/DAC/STAT(2023)9/FINAL/en/pdf
83
The purpose of the CSCS Supply Portal was to allow “national authorities and all implementing
partners supporting COVID-19 National Action Plans to request critical supplies” for combatting
COVID-19. See
https://www.who.int/docs/default-source/coronaviruse/covid-19-supply-chain-system-requesting-and-rec
eiving-supplies.pdf

112
 used. In cases where a catalogue was available for a given month but the
specific item donated was not included, an average of that items’ cost in
other CSCS catalogues was used.
■ Next, AidData took steps to pinpoint all in-kind donations of
COVID-related health supplies for which a transaction amount needed to
be, and could be, imputed. First, a broad set of potentially eligible
donations were identified using a keyword search in internal databases
for “COVID”. Additionally, all projects for which the COVID variable field
was set to “Yes” were reviewed. Next, AidData coders reviewed these
projects to identify which met the following criteria:
● The number of units donated were known for at least one type of
supply
● No available source provided a monetary value for the donation
● At least one type of supply donated was included in the WHO’s
Emergency Global Supplies Catalogue for COVID-19
■ AidData coders then multiplied the per-unit price of the relevant item by
the number of units donated in order to calculate the estimated
transaction amount. Coders were instructed to code certain specific
cases in the following ways:
● In cases where multiple types of supplies were provided in the
same donation, the “units donated times per unit price”
calculation was repeated for each supply type, and their value
summed for the total estimated transaction amount.
● In cases where the number of units donated are known for some
types of medical supplies but not for others, the transaction
amount is based only on the supplies with a known number of
donated units.
● In cases where both items that do and do not appear in a price
catalogue were donated together, the transaction amount is
based only on the known items.
● In cases where a group of donors provided the in-kind donation
jointly, with their respective contributions unknown, the total
value of the donation was calculated and split according to the
Equal Contribution Assumption (see above).
■ The ‘Amount Estimated’ field was systematically set to ‘Yes’ for in-kind
donations with imputed transaction amounts. Additionally, a staff
comment reading the following was systematically populated in the Staff
Comments field: “AidData has estimated the transaction amount for this
donation based on price catalogues from the World Health Organization
(WHO). Please see the TUFF Methodology for additional details.” In
future iterations of the dataset, coders will be instructed to apply this
coding manually to each project/activity record they create for an in-kind
COVID-related donation.
○ Other Scenarios: In all of these cases, AidData coders are instructed to set the
‘Amount Estimated’ field to ‘Yes’:
■ Scenario 1 (estimating transaction amounts for preferential or
non-preferential export buyer’s credits). If the underlying source materials
confirm that the financing for a project was issued in the form of an
export buyer’s credit (buyer’s credit loan) from an official sector institution

113
 in China, and the face value of the export buyer’s credit is unknown,
coders assume that it is equivalent to 85% of the commercial contract
cost. AidData recognizes that Chinese state-owned banks may
sometimes deviate from this practice and provide an export buyer’s
credit that covers as little as 60% of a commercial contract or as much as
95% of a commercial contract, but for estimation purposes, we adhere to
the “85% rule.” If coders record a transaction amount that is estimated
based on the 85% rule, they should include a staff comment that reads
“The face value of the buyer’s credit loan is not reported by any of the
underlying sources. AidData estimates that the face value by taking 85%
of the value of the underlying commercial (EPC) contract supported by
the buyer’s credit loan.” AidData coders are instructed to not make any
inferences or assumptions based upon the amount of export credit
financing that is insured by Sinosure (since Sinosure typically insures the
loan’s principal and interest, but the transaction amount field in the 2.0
dataset is only intended to capture the loan’s principal).84
■ Scenario 2 (estimating transaction amounts for government concessional
loans from China Eximbank). If AidData coders are confident that the
financing for a project is in the form of a government concessional loan
(GCL) from China Eximbank, they can assume that the proceeds of the
GCL were used to finance 100% of the commercial (EPC) contract costs
and code the transaction amount field accordingly. The absence of a
counterpart financing requirement is a core design feature of the GCL
lending instrument/program. If AidData coders record a GCL
transaction amount that is estimated, they should include a staff
commend that reads “The face value of the government concessional
loan is not reported by any of the underlying sources. AidData estimates
that the face value by taking 100% of the value of the underlying
commercial (EPC) contract supported by the government concessional
loan.”
■ Scenario 3 (estimating transaction amounts for MOFCOM’s interest-free
loans). If AidData coders are confident that the financing for a project is
in the form of a interest-free loan from MOFCOM, they can assume that
the proceeds of the loan were used to finance 100% of the commercial
(EPC) contract cost. The absence of a counterpart financing requirement
is a core design feature of MOFCOM’s interest-free loan lending
instrument/program. Therefore, the transaction amount field can be set
to 100% of the commercial contract cost. If AidData coders record a
transaction amount that is estimated, they should include a staff
comment that reads “The face value of the interest-free loan is not
reported by any of the underlying sources. AidData estimates that the
face value by taking 100% of the value of the underlying commercial
(EPC) contract supported by the interest-free loan.”

84
If a guarantee was provided for the loan and the guarantee amount is known while the loan’s face
value is not known, AidData coders are instructed to infer the face value of the loan (commitment
amount) from the monetary value of the guarantee. This is because the guarantee amount is equivalent
to the face value of the loan (i.e. the loan’s principal).

114
 ■ Scenario 4 (estimating transaction amounts of MOFCOM/Chinese
government grants for infrastructure projects). If AidData coders are
confident that an infrastructure project is being financed with a grant
from MOFCOM/the Chinese government, they can assume that the grant
was used to finance 100% of the commercial (EPC) contract cost.
Therefore, the transaction amount field can be set to be 100% of the
commercial contract cost. If AidData coders record a transaction amount
that is estimated, they should include a staff comment that reads “The
face value of the grant is not reported by any of the underlying sources.
AidData estimates that the face value by taking 100% of the value of the
underlying commercial (EPC) contract supported by the grant.” These
fully funded projects are often referred to in official Chinese source
materials as “China-aided projects.”
● Implementation Dates/Implementation Dates Estimated: The “Planned Implementation
Start Date” field captures the day on which a project/activity supported by an official
financial (or in-kind) commitment from China was originally scheduled to begin
implementation; the “Actual Implementation Start Date” field records the day on which
a project/activity supported by an official financial (or in-kind) commitment from China
began implementation. The “Planned Completion Date” field captures the day on
which a project/activity was originally scheduled to reach completion; and the “Actual
Completion Date” field captures the day on which a project/activity was completed.
○ All of these fields seek to capture precise calendar dates. However, in cases
when AidData coders are only able to identify the month and year in which a
project implementation start date or completion date took place (or was
scheduled to take place), the first day of the month is used as a proxy measure.85
If any of these proxies are used to estimate the implementation dates, the
“Actual Implementation Start Date Estimated” field or the “Actual Completion
Date Estimated’ should be set to “Yes.”
● Maturity: This field captures the total number of years it will take the borrower to repay
a loan or export credit, as specified in the original loan or export credit agreement.
AidData includes loans with maturities less than 1 year in duration to ensure
comprehensive coverage of official financial flows to China. However, users of the 3.0
dataset who wish to exclude these loans from their analysis to ensure strict
comparability with OECD-DAC statistics can use the maturity field to filter out loans
with values less than 1.
● Interest Rate: This field captures the rate of interest (in percentage terms) that applies
to a loan, as specified in the original loan agreement. Loans can have fixed interest
rates or variable interest rates. Variable interest rates are also referred to as floating
interest rates. These rates are based on market rates that float over time added to a
fixed margin. The actual interest paid back is determined by the trends of the market
rate over the term of the loan. We are not able to measure these trends in the market
rate over the term of the loan; therefore, to calculate the grant element, we instruct
AidData coders to convert the variable interest rate to a “fixed” interest rate at a single
point in time. Specific rates (floating interest rates at a single point in time) and the
number of basis points86 (the fixed margin) are sometimes detailed on official

85
See Section 1.4 for further details.
86
500 basis points = 5%, 100 basis points = 1%, 50 basis points = 0.5%; 3 Month LIBOR + 100 bps = 3
Month LIBOR + 1%.

115
 documents published by relevant agencies. If a specific rate is not provided in official
sources, then AidData coders use the rate at the time that the project agreement was
finalized (i.e., the time of the official commitment).
○ To calculate the “all in” interest rate of a loan with a floating rather than fixed
interest rate, AidData coders anchor the floating market interest rate to the
value of the rate at the time the loan was issued. They first identify the average
rate during the month of the official commitment (e.g. average 6-month LIBOR
rate in January 2017 of 1.34%), and then add it to the margin specified in the
loan agreement. If a loan or swap was committed in a specific calendar year, and
the specific month is unspecified (e.g. 2017 as opposed to March 2017), the
average monthly rate during that calendar year is used. If a floating interest rate
is calculated in this manner, the details should be included in the project
description.
■ Website for LIBOR and EURIBOR:
https://www.global-rates.com/en/interest-rates/libor/american-dollar/hist
orical/
■ Website for JIBAR:
https://www.jse.co.za/downloadable-files?RequestNode=/Safex/mtmdat
a
● Grant Element: This field captures the grant element of a loan or export credit at the
time that the original loan or export credit agreement was signed. For each loan where
AidData coders identify loan pricing details (in particular the maturity and interest rate),
AidData uses the OECD’s grant element formula to calculate the grant element. If a
grace period is available, the grant element formula will include that information.
However, if no grace period is available, AidData assumes a grace period of 0 years.87
● JV/SPV Government Ownership: For each project/activity received by a JV or SPV,
AidData coders are instructed to record the ownership breakdown of the recipient
JV/SPV, including ownership organization name, organization type, percentage of
ownership, and country of origin. These variables subsequently determine the value of
the JV/SPV Host Government Ownership field, JV/SPV Chinese Government Ownership
field, and Level of Public Liability field.88
● Source information: For each source identified in any stage of data collection or
verification, AidData coders are instructed to attach the source to the record. Included
in the source material is the public URL where the source can be accessed, title,
author(s), published date, publisher name and location, language of source and type of
source (including whether it is an official donor or recipient source, a media article, an
academic source, etc). The information on a project’s sources is published alongside the
project information to allow for transparency in how the record was compiled.
● Geographic location: Coders should record geographical details that accurately and
precisely document project/activity’s physical footprint, including location names and
types; the position or distance of the project/activity’s location vis-à-vis other
geographical features (e.g., the building is located across the street from the country’s
parliamentary complex in the capital city); the name, length, and start points and end

87
See Section 2.1 for further details regarding the calculation of the grant element.
88
The value of the The “Level of Public Liability” field in the 3.0 version of AidData’s GCDF dataset
captures the extent to which the host government may eventually be liable for debt repayment. See
Section 2.3 for more details on the “Level of Public Liability” field.

116
 points of physical infrastructure supported by the project (e.g., the 115.85 kilometer A1
highway runs from Colombo to Kandy); the total land area occupied by the project site
(e.g., the industrial park occupies a 10 square kilometer area); and latitude and
longitude coordinates of specific project features (e.g., the coordinates of Olkaria IV
Geothermal Power Station Kenya at Hells Gate National Park are -0.918056,
36.334444). Whenever possible, coders should record OpenStreetMap URLs that
capture the geographical locations and the following features of projects (see section
4.1 for a full description): (i) the precise physical boundaries and exact locations of
buildings and facilities (e.g., schools, hospitals, stadiums, government buildings, power
plants, and factories) with polygons or points; (ii) the precise geographical scope of
special economic zones, industrial parks, mining concessions, protected areas, and plots
of land under cultivation via polygons or points; and (iii) the exact routes of linear
infrastructure (e.g., roads, bridges, tunnels, railways, power lines, canals, and pipelines)
via line vectors.

3.3 - Stage 3a: Project-Level Quality Assurance

Once Stage 2 is completed for a given record, it advances to Stage 3a ( Quality Assurance, or
QA).89 AidData coders assigned to Stage 3a should assess (a) whether a record’s sources,
variables, title, and description tell a coherent narrative; (b) whether the record is complete
(with respect to the 127 fields in the 3.0 dataset); and (c) whether the underlying sources
support the coding and categorizations determinations that were made in prior stages. Every
record newly created or amended during the 3.0 data collection process is subjected to Stage
3a.

Stage 3a consists of a series of rigorous and systematic QA procedures that are designed to
identify and eliminate common mistakes, coding errors, biases, false assumptions, and
information gaps. Stage 3a coders also ensure that there is sufficient evidence from official
sources to confirm key project details. AidData staff conduct Stage 3a for (a) countries receiving
especially high volumes of Chinese ODA and OOF and (b) and countries with many complex
transactions. AidData’s strongest and most experienced coders quality assure the remaining
project/activities records.
● Logical consistency: Some fields depend on the coding of other fields. For example, a
record’s flow class is a function of intent, concessionality, flow type, and funding agency.
Therefore, export credits by definition cannot have an ODA-like flow class, regardless of
concessionality, because they can only have commercial or mixed intent. Stage 3a
coders are responsible for resolving these logical inconsistencies.
○ Auto-fill logic: In 2021, AidData transitioned to a new internal data management
platform. One of the most consequential upgrades involved the introduction of
“auto-fill logic,” which has reduced the frequency of Stage 1 and Stage 2
coding errors and made it easier for Stage 3a coders to perform logical
consistency checks. For example, if Sinosure is coded as an Insurance Provider,
then the Insurance Provided field automatically populates to “Yes.” In other
cases, the coding of one field limits the coding options for a different field. For

89
Records with flow type equal to Foreign Direct Investment (FDI), Joint Venture (JV), or Official
Investment undergo Stage 2 but not QA because they are excluded from AidData’s final Global Chinese
Development Finance Dataset, 3.0.

117
 instance, if a private entity is coded as the Funding Agency, then only unofficial
flow classes can be selected by coders.
○ However, the auto-fill logic does not address record titles or descriptions, so
Stage 3a coders must carefully scrutinize those fields. For example, if a
description field mentions a handover ceremony for a finished project, then the
Actual Completion Date field should not be blank and the Status field should be
coded as Completed. If the description field mentions linked records, then those
project identification numbers should be added to the same Parent ID.
● Reduce and eliminate double counting: In light of the fact that the TUFF methodology
draws information from a range of sources and tracks Chinese ODA- and OOF-financed
projects/activities over time, there is a risk of capturing the same transactions multiple
times. AidData staff and coders eliminate instances of double-counting by deactivating
duplicates and assigning some records to the Umbrella category.
○ Duplicate checks: During Stage 1 and Stage 2, the data management platform is
searched for duplicates before further amending or creating records. Stage 3a
coders practice strategic filtering and keyword searches to identify and
sometimes deactivate duplicate records.
○ Umbrella: When a record’s Umbrella field is set to “Yes,” it usually means that it
is capturing a signed financial agreement but the funds are not allocated for a
specific project/purpose until a subsequent date. Umbrella records serve as a
placeholder until separate, subsidiary records are created recording the entire
financial breakdown. Stage 3a coders verify whether the umbrella marker is
necessary or not.90
● High value checks: Stage 3a coders are instructed to pay special attention to records
with transaction amounts over $100 million. This means verifying the financial details in
official donor/creditor and/or recipient government sources but also ensuring that the
record as a whole is as close to correct and complete as possible.
● Verify calculations: Some record entries demand a little arithmetic. Stage 3a coders
performing QA should always check a loan’s floating interest rate calculations to make
sure the previous coder(s) correctly identified the “all in” interest rate.
● Clarify assumptions: Only AidData staff and coders who have demonstrated the
strongest grasp of the TUFF methodology and most reliable judgment are asked to
conduct QA activities because this stage requires making judgements that other coders
cannot not be expected to make. For example, if a loan meets the minimum
requirements for assuming that it is an export buyer’s credit, the Stage 3a coder is
responsible for recognizing that possibility, re-coding fields, and justifying their
assumptions in the description or staff comments fields.91

Once Stage 3a is complete, records are passed on to senior AidData faculty and staff for review
and feedback. Those records are then passed back to AidData staff and coders for feedback
incorporation and another round of QA. Stage 3a is the last comprehensive record-by-record
review before regional and global checks are undertaken.

90
See Section 1.4 for more details on the Umbrella field.
91
See Section 2.5.3.3 for more details on these minimum requirements.

118
3.4 - Stage 3b: Dataset-Level Quality Assurance

Following Stage 3a, AidData staff perform a rigorous set of protocols (Stage 3b) to remove any
errors and biases in order to produce the most consistent, complete and replicable dataset
possible. These procedures are detailed below:

● Targeted Review: After a record-by-record review during Stage 3a, the dataset
undergoes another layer of review that focuses on high-value projects/activities (as
indicated by especially large commitment amounts). A staff member reviews all records
greater than $1 billion for accuracy and missing information. This review is meant to add
an additional layer of scrutiny to ensure no additional data can be identified, field
codings are correct, financial values are accurate, and no duplication of records has
occurred. A staff member will also review any records still marked as “suspicious” after
the QA stage and update records as needed. In addition, a staff member reviews the
dataset for any incorrect inclusions or exclusions of projects/activities that could
substantially influence analysis that involves aggregate financial commitment amounts. .
● Data Logical Consistency Checks: After Stage 3a is completed and the Targeted Review
is carried out, AidData staff perform a series of data checks to make sure all fields are
correctly coded and to any outstanding information gaps (of special importance to
analysts) are addressed. This process include (1) reviewing variable fields such as
Commitment Date, Flow Type, Flow Class, and Sector Code that were not coded; (2)
reviewing records with Flow Type designations of Vague TBD; (3) reviewing the flow
class of loan projects/records; (4) reviewing any Grants/Technical
Assistance/Scholarships with non-ODA/non-OOF/non-Vague OF Flow Class
designations for funding agencies that should be set as official funding agencies (thus
updating the Flow Class coding); (5) reviewing records that have a status designation of
Pipeline: Pledge and Pipeline: Vague and specific implementation dates or completion
dates; (6) reviewing records with Suspended or Canceled status designations that
should have been assigned to the Pipeline: Pledge category (because no financial
commitment was ever issued)e; (7) reviewing records with Pipeline: Commitment stage
designations where only a framework agreement was signed; (8) reviewing any ETCAs
that were assigned to the Pipeline: Pledge or Pipeline: Vague status categories; (9)
reviewing any lines of credits that should not be coded as umbrella projects; (10)
ensuring consistent coding between official flow type and flow class; (10) ensuring
consistent coding of participating organizations (e.g. funding agencies, co-financing
agencies), including their organization type and origin designations; (11) ensuring
consistent application of coding guidelines to key variable fields (e.g., guidelines to
estimate transaction amounts); and (12) reviewing health of record scores for each
record and targeting an extra layer of review for records with lower scores on any of the
4 measures (see Appendix E for more details on the Health of Record scores).
● Extended Review: Once the dataset has gone through all of the previously described
steps, it is reviewed by a new set of AidData staff and a different cohort of external
coders. These reviewers vet the dataset using various methods, including but not
limited to (1) generating descriptive statistics with the dataset to identify anomalies or
suspicious results; (2) comparing the dataset and the resulting financial amounts to
other published estimates of Chinese development finance (or subsets thereof) to
identify significant deviations from other estimates, including White Papers published

119
 by the Chinese Government and estimates published by third parties; (3) comparing
individual records to official sources to ensure comprehensive and accurate coverage;
(4) reviewing individual records for errors or missing data; and (5) identifying biases in
the data and identify potential ways to address them.
● Deflation & Financial Review: To ensure the financial commitment values are
comparable across years, all of these values are calculated in constant 2021 U.S. dollars
using the deflation methodology that is described in Appendix D. As part of this
process, potential local currency changes and revaluations are identified and the
currency exchange rates are adjusted accordingly.

Section 4 - Geospatial Data Collection Process

Upon completion of Stage 3 (Quality Assurance, or QA) for a designated region,
projects/activities advance to the geospatial data collection stage. This data collection stage is
geared towards identifying financial and in-kind transfers that frequently underpin physical
assets or activities at specific locations characterized by geographical features, serving as the
ultimate destination for the financial transfer (flow). Projects/activities with no geolocation
information or geofeatures available or Projects/activities without specific financial destinations
are not included in the geospatial data collection process.

To compile projects/activity locations for the 3.0 version of the GCDF dataset, we leverage
existing features from OpenStreetMap (OSM, the world’s largest catalogue of open source,
community-driven geospatial information), and contribute updates or new features reflecting
projects/activities. OSM records countless geographic features, from jurisdictional borders to
the exact routes of roads and the precise locations of individual buildings. The geospatial data
that are provided by OSM (referred to as features in OSM) fall into three primary categories:
nodes, areas, and ways. Below is a description of the four types of OSM geographical features
representing physical footprints of Chinese ODA- and OOF-financed projects/activities:

OSM Geographical Feature Example assets/activities

Node (point) Water wells, oil derricks, wind turbines,

statues/monuments, telecommunication towers,
and (some) buildings. Also used to represent
administrative areas without exact area
definitions.

Opened Way (line) Roads, bridges, tunnels, railways, electricity

transmission lines, canals, and pipelines

Closed Way (polygon) Schools, hospitals, airports, seaports, dams,

power plants, substations, factories, stadiums,
and office buildings, zoos, public parks, protected
areas, special economic zones, farms, mining
concessions, and industrial parks, as well as
administrative areas.

120
 Relationship (Mixed) A combination of one or more
nodes/ways/relationships.

AidData’s process for identifying, collecting, and conducting quality assurance on the
geofeatures of Chinese ODA- and OOF-financed projects/activities involves three stages of
workflow.

4.1 - Stage 1: Geospatial Data Collection and Precision Level

Labeling

To identify OSM features associated with projects/activities, AidData utilizes documentation

from established quality assured records and sources to conduct targeted searches. For
example, a project description of a hydropower station being built in the east of a city along a
river would be cross-referenced with satellite imagery of the area and site photos provided in
the project sources to determine its exact location. Coders then search OSM for existing
features associated with the hydropower station, edit or add new features if needed, and
record the corresponding OSM feature IDs and URLs. Due to limits on the availability of
information on open source platforms, it is not always possible to identify or track a precise
geofeature. A larger ADM area (or administrative boundaries) may be available in some cases
to fill the informational gap. The objective of AidData coders is to identify geofeatures at the
most precise level possible. Whenever geofeatures are unavailable at the most precise level,
coders follow a hierarchical order (described below) to fill in less precise geolocational
information. Below is a description of the various levels of precision that AidData identifies in
the dataset, based on the availability of information.

Level of Precision Feature Example

Precise The precise boundary of the The boundaries of an airport.

feature itself

Approximate (5km buffer) Areas, landmarks near the target A power station is not
feature within a 5km radius. identifiable with existing
information but a substation is
identifiable in a 5km radius. In
this case, the substation would
be retrieved and Approximate
level will be labeled.

ADM 8 Village/City/Town (Refer to the Neither the precise boundary

OSM ADM level92) nor a nearby landmark could be
identified. A smallest ADM level
boundary will be the next
priority.

92
OSM ADM level for all countries: https://wiki.openstreetmap.org/wiki/Tag:boundary%3Dadministrative

121
 ADM 6 City/County (Refer to the OSM Neither the precise boundary
ADM level) nor a nearby landmark could be
identified. A smallest ADM level
boundary will be the next
priority.

ADM 4 Province/Territory (Refer to the Neither the precise boundary

OSM ADM level) nor a nearby landmark could be
identified. A smallest ADM level
boundary will be the next
priority.

ADM 3 (*rare) Region (Refer to the OSM ADM Neither the precise boundary
level) nor a nearby landmark could be
identified. A smallest ADM level
boundary will be the next
priority.

After coders record the OSM feature IDs and URLs, they then refer to the level of precision
system to label each project accordingly.

4.2 - Stage 2: Geospatial Data Enhancement and Quality

Assurance

Once the Stage 1 geospatial data collection is completed for a designated region, AidData
coders conduct a review of all the retrieved geospatial data to ensure the geofeatures
accurately reflect the final destination of the ODA/OOF flows at the most precise level
available. During this Stage 2 review process, AidData coders cross-check OSM features and
the level of precision system with the established records.

4.3 - Stage 3: Geospatial Data Cleaning and Dataset Generation

Once all regions are reviewed and quality assured, AidData staff reformat the data collection
worksheet with R scripts to automatically check for human error. Projects/activities with missing
or mismatched levels of precision with OSM features are flagged for a second round of review
and troubleshooting.

After cleaning the OSM feature data, AidData staff generate GeoJSONs using the OSM URLs
through a Python workflow leveraging a combination of web scraping, the Overpass API, and
the osm2geojson package. Individual GeoJSONs are then combined into multi-polygons,
allowing AidData to represent various features within a project/activity as a single
multi-polygon, as well as ensure a consistent feature type across all extracted geospatial
features (lines, points, and polygons) in the final dataset. During the process, AidData staff flag

122
a subset of invalid OSM links/features for a third round of review and troubleshooting. Since
OSM is an open source community, a previous feature may have been revised/removed by
other collaborators, which would result in the previous OSM link being invalid. The third round
of review aims to find alternative valid OSM features and code corresponding levels of
precision and rerun to get the updated GeoJSONs (Goodman et al. 2023). The code to
replicate the GeoJSON generation can be found at:
https://github.com/aiddata/gcdf-geospatial-data.

After the GeoJSONs are processed, AidData processes ADM files with centroid points for data
users. See detailed documentation in the read.me file in the download.

123
Appendices

Appendix A: Classification of Official Finance

AidData seeks to assign projects/activities to an official finance classification (for the ODA and
OOF projects/activities we capture) based on the OECD-DAC guidelines. Doing so allows users
to make comparisons between Chinese development finance and development finance from
other donors. Projects/activities are assigned to the ODA-Like category if they meet three
criteria. First, the primary purpose of the project/activity must be the promotion of economic
development and welfare in the recipient country (i.e., have development intent). Second, the
project/activity must take place in a country that qualifies for ODA based on its income level.
Third, the official commitment supporting the project/activity must be concessional in nature
(i.e., grant, technical assistance, scholarship, debt forgiveness, or loan with a grant element
meeting a specified threshold). For official commitments issued (flows reported) between 2000
and 2017, we follow the OECD's practice to use the cash-flow methodology to define ODA,
which included a threshold level of 25% grant element with a discount rate of 10% for all loans.
For official commitments issued (flows reported) in 2018 and subsequent years, we use the
OECD's grant-equivalent methodology, which relies upon a tiered concessionality threshold
system for loans. Under the grant-equivalent methodology, the concessionality threshold for
loans to the official sector in the recipient country is 45% for LDCs and other LICs (using a
discount rate of 9%), 15% for LMICs (using a discount rate of 7%) and 10% for UMICs (using a
discount rate of 6%). Loans to the private sector, however, continue to use the 25% threshold
used in the cash-flow methodology (in alignment with OECD-DAC practices). Users can refer to
the "OECD ODA Concessionality Threshold" field to identify the threshold used for a particular
loan record in the dataset. Projects/activities that are supported by an official financial or
in-kind transfer but do not meet all three of these criteria are assigned to the OOF-Like
category. Projects/activities that are backed by an official commitment but cannot be reliably
categorized as ODA-like or OOF-like because of insufficiently detailed information are

124
assigned to the “Vague (Official Finance)” category. Projects/activities in this residual category
primarily consist of (a) those with an unspecified “Flow Type” (i.e., values of “Vague TBD”); and
(b) those financed with development-intent loans for which AidData lacks the borrowing terms
(interest rates, grace periods, or maturity dates) needed for concessionality determinations. We
do not capture Official Investment flows at this time.

Appendix B: Geographic Coverage of the 3.0 Dataset

List of Countries Covered in AidData's Global Chinese Development Finance Dataset, Version
3.0

Country Record Count

Afghanistan 169

Albania 64

Algeria 57

American Samoa No projects/activities found

Angola 403

Antigua and Barbuda 76

Argentina 189

Armenia 52

Aruba No projects/activities found

Azerbaijan 53

Bahamas 47

Bangladesh 175

Barbados 109

Belarus 159

Belize No projects/activities found

Benin 174

125
Bhutan No projects/activities found

Bolivia 147

Bosnia and Herzegovina 55

Botswana 131

Brazil 200

British Virgin Islands No projects/activities found

Brunei Darussalam 53

Bulgaria 60

Burkina Faso 62

Burundi 156

Cabo Verde 108

Cambodia 417

Cameroon 203

Cayman Islands No projects/activities found

Central African Republic 159

Chad 144

Chile 78

Colombia 101

Comoros 96

Congo 232

Cook Islands 41

Costa Rica 81

126
Cote d'Ivoire 189

Cuba 145

Curacao 5

Democratic People's Republic of Korea 149

Democratic Republic of the Congo 298

Djibouti 123

Dominica 110

Dominican Republic 32

Ecuador 232

Egypt 119

El Salvador 58

Equatorial Guinea 185

Eritrea 100

Eswatini No projects/activities found

Ethiopia 311

Fiji 211

French Polynesia 2

Gabon 109

Gambia 61

Georgia 52

Ghana 254

Grenada 141

127
Guam 1

Guatemala No projects/activities found

Guinea 143

Guinea-Bissau 123

Guyana 125

Haiti 35

Honduras 5

India 113

Indonesia 437

Iran 113

Iraq 52

Israel 30

Jamaica 114

Jordan 100

Kazakhstan 184

Kenya 249

Kiribati 52

Kosovo No projects/activities found

Kyrgyz Republic 143

Lao People's Democratic Republic 347

Lebanon 99

Lesotho 155

128
Liberia 213

Libya 14

Madagascar 165

Malawi 215

Malaysia 176

Maldives 101

Mali 178

Marshall Islands 66

Mauritania 158

Mauritius 162

Mexico 94

Micronesia 183

Moldova 61

Mongolia 232

Montenegro 34

Morocco 78

Mozambique 160

Myanmar 495

Namibia 232

Nauru 14

Nepal 222

New Caledonia No projects/activities found

129
Nicaragua 4

Niger 156

Nigeria 197

Niue 27

North Macedonia 62

Northern Mariana Islands No projects/activities found

Oman 27

Pakistan 496

Palau 1

Panama 70

Papua New Guinea 200

Paraguay 4

Peru 165

Philippines 267

Puerto Rico No projects/activities found

Romania 42

Russia 295

Rwanda 152

Saint Lucia 18

Samoa 167

Sao Tome and Principe 77

Senegal 132

130
Serbia 99

Seychelles 170

Sierra Leone 229

Sint Maarten (Dutch part) 1

Solomon Islands 40

Somalia 65

South Africa 314

South Sudan 189

Sri Lanka 301

St. Kitts and Nevis No projects/activities found

St. Martin (French part) No projects/activities found

St. Vincent and the Grenadines No projects/activities found

Sudan 337

Suriname 106

Syrian Arab Republic 84

Tajikistan 172

Tanzania 290

Thailand 116

Timor-Leste 131

Togo 165

Tonga 159

Trinidad and Tobago 58

131
Tunisia 94

Turkey 152

Turkmenistan 53

Turks and Caicos Islands No projects/activities found

Tuvalu No projects/activities found

Uganda 213

Ukraine 123

Uruguay 89

Uzbekistan 196

Vanuatu 157

Venezuela 179

Viet Nam 191

Virgin Islands (U.S.) No projects/activities found

West Bank and Gaza Strip 52

Yemen 89

Zambia 246

Zimbabwe 273

132
Appendix C: TUFF Source Prioritization Protocol
It is common for sources to have conflicting information on a certain project. In this case, it is
necessary to have a hierarchical ranking of how much we weigh in each source.

Ranking of Resource Types based on Reliability of Project Data

1. Official government source, from a donor or recipient government agency

2. Implementing or intermediary agency report/website
3. Other official source (e.g. IMF, World Bank, CIA, etc.)
4. Peer-reviewed scholarly article
5. Other scholarly output, including working papers and dissertations
6. NGO, civil society, or advocacy group report/website
7. Media reports, including Wikileaks
8. Social media, including blogs from any unofficial source

If any conflicting information exists on a project detail, then the AidData coder or staff member
will arbitrate by explicitly stating within the project description a) the source of this conflict and
b) the reasoning for the proposed solution in the Staff Comments field.

Appendix D: AidData’s Deflation Methodology

1. Currency Conversion and Deflation Purpose

Financial values collected as part of AidData’s data collection activities, including TUFF, must
be converted and deflated so that they are comparable across currencies and years. AidData’s
methodology follows after the OECD’s methodology.93 The full methodology involves two
steps: 1) Calculating nominal exchange rates and 2) calculating deflation rates detailed below.
We calculate the deflators based on the OECD’s methodology using World Bank sources for
exchange rates and inflation.

2. Exchange Rates

2.1 Exchange Rate Methodology

Before deflation, all values must first be expressed in nominal (current) U.S. dollars (USD). This
is done with an LCU per USD exchange rate, applied by:
(original value) / (LCU per USD) = (new value)

For example:
100 EU / .7 = 142.57 USD

93
Available at http://www.oecd.org/dac/aidstatistics/informationnoteonthedacdeflators.htm

133
2.2 Exchange Rate Sources
Global Economic Monitor -- Official exchange rate, LCU per USD, period average (annual),
https://databank.worldbank.org/source/global-economic-monitor-(gem)

2.3 Currency Revaluations/Changes in Currencies

The standard data from the World Bank does not take into consideration currency revaluations
and currency changes. So to reflect this nuance, we identified the complete list of countries
that had undergone currency changes or revaluations that would affect the exchange rates
used in TUFF datasets. In cases where the financial amount was quoted in old currencies, we
used historical exchange rates (annual period average) from OANDA to calculate the exchange
rate to USD.

3. Deflators
Deflation is necessary to take the USD nominal amount and deflate (or inflate) that amount into
a constant year across the whole dataset so all the financial values are comparable despite year
values. Deflators control for two changes over time: inflation in the donor country and change
in buying power in the donor country relative to the United States. Both of these changes are
calculated separately, and then multiplied together to get the final deflator used by AidData.
The formula is the following:

Inflation * Change in Buying Power = Deflator

3.1 Inflation

The first part of the deflator formula is to calculate the inflation value from the base year to the
constant year.

Inflation is measured as relative to a given base year. The below example calculations use a
base year of 2014.94 Percentages are then generated using the following formula:

Percentage Year = Percentage Previous Year + (Percentage Previous Year * Inflation Year )

For example, in 2014, Colombia’s GDP inflation was 4.2%. Taking 2014 as the base year, the
percentage for 2014 is 100%. So, to calculate the percentage for 2010, using 2014 as the start
year:

100 = Pprevious + ( Pprevious * .04 )

Pyear Iyear

This yields 96% as Colombia’s percentage for 2012. (Decimals have been rounded for this
example, but were not rounded for AidData’s deflator table.) In 2012, Colombia’s GDP inflation
was 8%. Then, to calculate 2012, 2013 is the start year:

96 = Pprevious + ( Pprevious * .08 )

Pyear Iyear

This yields 89% as Colombia’s percentage for 2012.

94
AidData’s GCDF 3.0 dataset uses a base year of 2021 for constant USD amounts.

134
The following sources are used to compile the inflation values: World Bank GDP Inflation --
http://data.worldbank.org/indicator/NY.GDP.DEFL.KD.ZG

3.2 Change in Buying Power

The second part of the deflators formula is to calculate the change in Buying Power for the
donor country.

The change in buying power is taken from the LCU per USD rate and expressed as:

Exchange Rate Base Year / Exchange Rate Transaction Year = Change in Buying Power

For example, the Korean Won to USD rate was 1273.9 in 2014 and 804.4 in 1996. The
subsequent change in buying power is:

1273.9 / 804.4 = 1.58

Note that this methodology yields a ratio of 1 for all currencies pegged to the USD.
The data used for the buying power formula are generated from the historical exchange rates
described above.

3.3 Finalized Deflators

The GDP inflation and change in buying power numbers are combined to create annual
deflators for donor countries:

Inflation * Change in Buying Power = Deflator

4. Examples on Using GDP Deflators

Amounts in LCU should be converted to nominal USD, using the LCU per USD exchange rates
found in sheet A1. Then, the values should be divided by the percentages in sheet “E1." For
example, in 1975, Kuwait funded an electrification project in Bangladesh worth 6,400,000
KD1975 (AidData ID 2427051). To convert this amount to USD 2014, first, convert it to USD
1975:

6,400,000 KD 1975 / (.29003 KD/USD 1975) = 22,066,505.30 USD 1975

Next, divide it by the AidData deflator:

22,066,505.30 USD 1975 / 20.83% = 105,936,175.20 USD 2014

Note that amounts that are already reported in USD do not need to be converted. They only
need to be deflated (divided by the appropriate deflator).

135
Appendix E: Health of Record Scores
AidData’s ‘health of record’ scores are meant to signal the quality of each record in four
dimensions: 1) the quality of the sources used to underpin the record, 2) the completeness of
the record in terms of foundational project information, 3) the level of detail available on
project implementation, and 4) the financial details available for the project. The details of each
score are listed below.

Source Quality Score: This metric varies on a scale of 1 to 5, with 1 indicating that the record is
exclusively underpinned by unofficial sources and 5 indicating reliance upon multiple, official
sources. This score is meant to communicate which projects meet our preferred threshold for
reliability/quality of the sources. We would consider records with a score of 3 or lower as
records that have a lack of authoritative sources underpinning the record, flagging to users
records that may have reliability issues. The average score across the entire 3.0 dataset is 4.3,
indicating that most records successfully meet our threshold for quality of sources (scoring a 4
or 5). The full scoring criteria is detailed below:

Criteria for Source Quality Score:

Source Categorizations:
● Official sources include Donor/Recipient Official Source, Implementing/Intermediary
Organization Source, Other Official Source
● Tier 1 non-official sources include Academic Journal Article, Other Academic
● Tier 2 non-official sources include Media Report, NGO/Civil Society/Advocacy, Social
Media, Other

Assign a value based on the following criteria:

● 1 = Only media sources
○ Source type = Media Report (any number).
● 2 = Only Tier 2 non-official sources or non-official sources + any media sources (not
required)
○ Source types = NGO/Civil Society/Advocacy OR Social Media OR Other (at least
one). Can have Media Report source type as well (any number).
● 3 = At least 1 Tier 1 non-official source (but no official sources) + any Tier 2 non-official
sources (not required)
○ Source types = Academic Journal Article OR Other Academic (any number). Can
have Media Report, NGO/Civil Society/Advocacy, Social Media, or Other (any
number).
● 4 = Only 1 official source (no additional official sources) + any non-official sources,
either Tier 1 or 2 (not required)
○ Has only 1 source type that matches Donor/Recipient Official Source OR
Implementing/Intermediary Organization Source OR Other Official Source. Can
have Media Report, NGO/Civil Society/Advocacy, Other, Social Media,
Academic Journal Article, or Other Academic (any number).
● 5 = At least 2 official sources
○ Has at least 2 sources with source type Donor/Recipient Official Source,
Implementing/Intermediary Organization Source, or Other Official Source. Can
have Media Report, NGO/Civil Society/Advocacy, Social Media, Other,
Academic Journal Article, or Other Academic (any number).

136
Data Completeness Score: This metric varies on a scale of 0 to 5, with 5 indicating that the
basic fields of the record are complete. The “threshold” for a score of 5 is similar to the key
fields in the OECD-DAC’s Creditor Reporting System: an actual rather than estimated
commitment year, a non-missing transaction amount, a flow type/flow class that is not defined
as “Vague," and identifiable funding, implementing, and receiving agencies. The average Data
Completeness Score for the 3.0 dataset is 3.3.

Criteria for Data Completeness Score:

Start at 5, then
● Projects with year uncertain = -1
● Projects with no transaction amount (include umbrella projects) = -1
● Projects with vague flow class or flow type = -1
● Projects with a missing or unspecified Funding Agency = -1
● Projects with EITHER a missing Implementing Agency OR Receiving Agency= -1
*Min: 0, Max = 5

Project Implementation Score: This metric varies on a scale of 0 to 5, with higher scores
indicating that more implementation details have been captured in the record. The following
implementation details are considered: whether the project’s implementing agency (or
agencies), implementation start and completion dates (actual or planned), and geographical
locations are specified; and whether the project has a specified sector allocation. Project
Implementation Scores are only calculated for records with a “Recommended for Aggregates”
value of “Yes” and a “Status” value of “Implementation” or “Completion." The average Project
Implementation Score in the 3.0 dataset is 3.7.

Criteria for Project Implementation Score:

Add an additional point for each of the criteria met below where Status =
Implementation/Completion AND Recommended for Aggregates = Yes
1. The presence of implementing agency when Status = Implementation/Completion
2. The presence of start/end dates based on Status = Implementation
a. Actual start date = 1 point
b. Planned end date = 1 point
3. The presence of start/end dates based on status = Completion
a. Actual start date = 1 point
b. Actual end date= 1 point
4. The presence of location details when Status = Implementation/Completion
5. Sector != Unallocated/Unspecified

Loan Detail Score: This metric varies on a scale of 0-5, with higher values indicating that more
financial transaction details are captured in the record. Loan Detail Scores are only calculated
for records with a “Recommended for Aggregates” value of “Yes” and a “Flow Type”
designation of “Loan." A score of 5 indicates that a loan’s interest rate, maturity, transaction
value, loan type, funding agencies, and receiving agencies are all specified (i.e., not missing).
The average Loan Detail Score in the 3.0 dataset is 3.6.

Criteria for Loan Detail Score

Add an additional point when each of the fields below is not blank. Calculated only for projects
where Flow Type = Loan AND Recommended for Aggregates = Yes
○ Interest rate != blank

137
 ○ Maturity != blank
○ Transaction value != blank
○ If no interest rate or maturity, and Loan type != blank or “No Information”
○ Funding agency != blank OR Receiving agency != blank
○ Grace Period != blank

Appendix F: AidData’s Global Chinese Development Finance

Dataset, Version 3.0 at a glance
Global Chinese Development Finance Global Chinese Development Finance
AidData Dataset Dataset, Version 2.0 Dataset, Version 3.0
(Published September 2021) (Published November 2023)
Sectors All All
165 countries globally 165 countries globally
(including 145 countries with projects (including 146 countries with projects
Country Coverage identified) identified)

334 Chinese official sector donors and 791 Chinese official sector donors and
Financiers lenders lenders
Loans (with categorization of 23
Loans, grants, scholarships, technical distinct loan instruments), grants,
Scope & assistance, debt rescheduling, debt scholarships, technical assistance, debt
Coverage Financial Instrument forgiveness rescheduling, debt forgiveness
Number of Records 13,427 20,985
Number of Fields 70 133
Sources Publicly 91,125 147,703
Available (including 62,750 unique sources) (including 99,393 unique sources)
$1.34 trillion (2021 prices)
(excluding short-term “rollover”
Total Financial Value $851 billion (2017 prices) facilities, or $1.5 trillion when included)
Dataset 2000-2017 (with implementation 2000-2021 (with implementation
Summary Timeframe details through 2021) details through 2023)
Transaction amount, collateral,
interest rate, default interest rate,
grace period, maturity, commitment
Transaction amount, collateral, interest fee, management fee, insurance fee,
rate, grace period, maturity, first and last loan repayment dates,
Financial Details commitment fee, management fee level of public liability

Project
Details

138
 Funding agencies, co-financing
agencies, direct receiving agencies,
Funding agencies, co-financing indirect receiving agencies,
agencies, receiving agencies, implementing agencies, guarantor,
implementing agencies, accountable insurance provider, collateral provider,
Participating Agencies agencies security agent/collateral agent
Commitment date, status, planned and
actual start and end dates, deviation
Commitment year, status, planned and from planned start and completion
Implementation Details actual start and completion dates dates, infrastructure project flag
Description Average of 142 words per project Average of 166 words per project
Sector, flow class, recipient country
income classification, grant-equivalent
OECD Classifications Sector, flow class measure
Sub-national Details 3,285 physical locations 9,497 physical locations

139
References
Anaxagorou, Christiana, Georgios Efthyvoulou, and Vassilis Sarantides. 2020. Electoral Motives
and the Subnational Allocation of Foreign Aid in Sub-Saharan Africa. European Economic
Review 127: 103430.

Asmus, Gerda, Vera Z. Eichenauer, Andreas Fuchs and Bradley C. Parks. Forthcoming. Does
India Use Development Finance to Compete with China? A Subnational Analysis. Journal of
Conflict Resolution.

Baehr, Christian, Ariel BenYishay, and Bradley C. Parks. 2022. Environmental Impacts of
Chinese Government-Funded Infrastructure Projects: Evidence from Road Building in
Cambodia. Working Paper #115. Williamsburg, VA: AidData at William & Mary.

BenYishay, Ariel, Bradley Parks, Daniel Runfola, and Rachel Trichler. 2016. Forest Cover Impacts
of Chinese Development Projects in Ecologically Sensitive Areas. AidData Working Paper 32.
Williamsburg, VA: AidData at William and Mary.

Blair, Robert A., Robert Marty, and Philip Roessler. 2022. Foreign Aid and Soft Power: Great
Power Competition in Africa in the Early Twenty-first Century. British Journal of Political Science
52 (3): 1355-1376.

Bluhm, Richard, Axel Dreher, Andreas Fuchs, Bradley C. Parks, Austin Strange, and Michael J.
Tierney. 2018. Connective Financing: Chinese Infrastructure Projects and the Diffusion of
Economic Activity in Developing Countries. AidData Working Paper #64. Williamsburg, VA:
AidData at William & Mary.

Bluhm, Robert, Axel Fuchs, Bradley C. Parks, Austin M. Strange, and Michael J. Tierney. 2020.
Connective Financing: Chinese Infrastructure Projects and the Diffusion of Economic Activity in
Developing Countries. CEPR Discussion Paper 14818. London, UK: CEPR.

Dreher, Axel, Andreas Fuchs, Bradley Parks, Austin M. Strange, and Michael J. Tierney. 2018.
Apples and Dragon Fruits: The Determinants of Aid and Other Forms of State Financing from
China to Africa. International Studies Quarterly 62(1): 182–194.

Dreher, Axel, Andreas Fuchs, Roland Hodler, Bradley Parks, Paul A. Raschky, and Michael J.
Tierney. 2019. African Leaders and the Geography of China’s Foreign Assistance. Journal of
Development Economics 140: 44–71.

Dreher, Axel, Andreas Fuchs, Bradley Parks, Austin Strange, and Michael J. Tierney. 2021a. Aid,
China, and Growth: Evidence from a New Global Development Finance Dataset. American
Economic Journal: Economic Policy 13 (2): 135-174.

Dreher, Axel, Andreas Fuchs, Bradley Parks, Austin Strange, and Michael J. Tierney. 2022.
Banking on Beijing: The Aims and Impacts of China’s Overseas Development Program.
Cambridge, UK: Cambridge University Press.

Export-Import Bank of China. n.d. Government Concessional Loan & Preferential Buyer’s Credit
Brochure. Beijing, China: Export-Import Bank of China. Retrieved from

140
https://www.dropbox.com/s/sptor0fju165k3j/392125621-GovernmentConcessionalLoan-Prefer
entialBuyer-sCredit-Brochure-pdf.pdf?dl=0

Gelpern, Anna, Sebastian Horn, Scott Morris, Brad Parks, Christoph Trebesch. 2022. How China
Lends: A Rare Look into 100 Debt Contracts with Foreign Governments. Economic Policy
eiac054.

Goodman, Seth, Sheng Zhang, Ammar Malik, Bradley C. Parks, and Jacob Hall. 2023.
AidData's Geospatial Global Chinese Development Finance Dataset. Mimeo.

Horn, Sebastian, Bradley C. Parks, Carmen M. Reinhart, and Christoph Trebesch. 2023. China
as an International Lender of Last Resort. NBER Working Paper #31105. Cambridge, MA:
NBER.

Iacoella, Francesco, Bruno Martorano, Laura Metzger, and Marco Sanfilippo. 2021. Chinese
Official Finance and Political Participation in Africa. European Economic Review 136: 103741.

International Monetary Fund and World Bank. 2020. Collateralized Transactions: Key
Considerations for Public Lenders and Borrowers. Washington DC: International Monetary Fund
and World Bank.

Isaksson, Ann-Sofie and Andreas Kotsadam. 2018a. Chinese Aid and Local Corruption. Journal
of Public Economics 159: 146–159.

Isaksson, Ann-Sofie, and Andreas Kotsadam. 2018b. Racing to the Bottom? Chinese
Development Projects and Trade Union Involvement in Africa. World Development 106 (C):
284–298.

Isaksson, Ann-Sofie. 2020. Chinese Aid and Local Ethnic Identification. International
Organization 74 (4): 833–852.

Kose, M. Ayhan, Sergio Kurlat, Franziska Ohnsorge, and Naotaka Sugawara. 2022. A
Cross-Country Database of Fiscal Space. Journal of International Money and Finance 128
(November): 102682.

Malik, Ammar A., Bradley Parks, Brooke Russell, Joyce Lin, Katherine Walsh, Kyra Solomon,
Sheng Zhang, Thai-Binh Elston, and Seth Goodman. 2021. Banking on the Belt and Road:
Insights from a New Global Dataset of 13,427 Chinese Development Projects. Williamsburg,
VA: AidData at William & Mary.

Martorano, Bruno, Laura Metzger, and Marco Sanfilippo. 2020. Chinese development
assistance and household welfare in Sub-Saharan Africa. World Development 129 (C): 104909.

Marty, Robert, Seth Goodman, Michael Le Few, Carrie B. Dolan, Ariel BenYishay, and Dan
Runfola. 2019. Assessing the Causal Impact of Chinese Aid on Vegetative Land Cover in
Burundi and Rwanda Under Conditions of Spatial Imprecision. Development Engineering 4:
100038.

141
Morris, Scott, Brad Parks, and Alysha Gardner. 2020. Chinese and World Bank Lending Terms:
A Systematic Comparison across 157 Countries and 15 Years. CGD Policy Paper 170.
Washington, DC: Center for Global Development.

Muchapondwa, Edwin, Daniel Nielson, Bradley Parks, Austin M. Strange, and Michael J.
Tierney. 2016. “Ground-Truthing” Chinese Development Finance in Africa: Field Evidence from
South Africa and Uganda. Journal of Development Studies 52 (6): 780–796.

Organisation for Economic Co-operation and Development (OECD). 2023. Converged

Statistical Reporting Directives for the Creditor Reporting System (CRS) and the Annual DAC
Questionnaire, Chapter 1-6. 27 April 2023. Paris: OECD. Accessed on 14 September 2023 at
https://one.oecd.org/document/DCD/DAC/STAT(2023)9/FINAL/en/pdf

Parks, B. C., Malik, A. A., Escobar, B., Zhang, S., Fedorochko, R., Solomon, K., Wang, F., Vlasto,
L., Walsh, K. & Goodman, S. 2023. Belt and Road Reboot: Beijing's Bid to De-Risk Its Global
Infrastructure Initiative. Williamsburg, VA: AidData at William & Mary.

Pedregosa, Fabian, Gaël Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion,
Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, Jake
Vanderplas, Alexandre Passos, David Cournapeau, Matthieu Brucher, Matthieu Perrot, and
Édouard Duchesnay. 2011. Scikit-learn: Machine Learning in Python. Journal of Machine
Learning Research 12: 2825−2830.

Ray, Ray and Kevin P. Gallagher. 2020. Scope and Findings: China’s Overseas Development
Finance Database. Global Development Policy Center at Boston University. Accessed at
https://www.bu.edu/gdp/2020/12/13/scope-and-findings-chinas-overseas-development-financ
e-database/

Ray, Rebecca, Kevin P. Gallagher, William Kring, Joshua Pitts, and B. Alexander Simmons. 2021.
Geolocated dataset of Chinese overseas development finance. Nature Scientific Data.
Accessed at https://osf.io/gfwhj/?view_only=49a5f67a97ff4996812cc68612cc5cff

Strange, Austin, Bradley Parks, Michael Tierney, Andreas Fuchs, Axel Dreher, and Vijaya
Ramachandran. 2013. China’s Development Finance to Africa: A Media-Based Approach to
Data Collection. CGD Working Paper 323. Washington, DC: Center for Global Development.

Strange, Austin, Axel Dreher, Andreas Fuchs, Bradley C. Parks, Michael J. Tierney. 2017.
Tracking Underreported Financial Flows: China’s Development Finance and the Aid-Conflict
Nexus Revisited. Journal of Conflict Resolution 61(5): 935-963.

Wellner, Lukas, Axel Dreher, Andreas Fuchs, Bradley C. Parks, and Austin Strange.
Forthcoming. Can Aid Buy Foreign Public Support? Evidence from Chinese Development
Finance. Economic Development and Cultural Change.

World Bank. 2020. International Debt Statistics 2021. Washington DC: World Bank. Accessed
on 20 February 2023 via https://openknowledge.worldbank.org/handle/10986/34588

World Bank. 2021. International Debt Statistics 2022. Washington DC: World Bank. Accessed
on 20 February 2023 via https://openknowledge.worldbank.org/handle/10986/36289

142
World Health Organization (WHO). 2020. COVID-19 Supply Chain System: Requesting and
Receiving Supplies. Geneva: WHO. Accessed on 31 October 2023 via
https://www.who.int/docs/default-source/coronaviruse/covid-19-supply-chain-system-requestin
g-and-receiving-supplies.pdf

143