The University of Chicago Press

Big Constraints to Small Firms Growth?
Business Environment and Employment Growth

across Firms
Author(s): Reyes Aterido /Mary Hallward-Driemeier /Carmen Pags
Source: Economic Development and Cultural Change, Vol. 59, No. 3 (April 2011), pp. 609-647
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/10.1086/658349 .
Accessed: 31/03/2014 09:24
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
.
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to
Economic Development and Cultural Change.
http://www.jstor.org
This content downloaded from 210.56.13.7 on Mon, 31 Mar 2014 09:24:45 AM
All use subject to JSTOR Terms and Conditions
2011 by The University of Chicago. All rights reserved. 0013-0079/2011/5903-0005$10.00
Big Constraints to Small Firms Growth? Business
Environment and Employment Growth across Firms
reyes aterido
World Bank
mary hallward-driemeier
World Bank
carmen pages
Inter-American Development Bank
I. Introduction
An issue of particular relevance for economic growth is how the development
of markets and institutions inuences the growth of rms. This issue is even
more important in developing countries, where markets and institutional in-
frastructure are less developed. A related concern is whether an adverse business
environment affects the efciency and the dynamism of the business sector.
This study uses rm-level data on more than 56,000 enterprises in 85
developing countries and ve high-income economies to assess the effects of
the broader business environment on employment growth by rms, focusing
on differences across rm size. This study is related to two strands of literature.
First it relates to recent studies (Hsieh and Klenow 2009; Restuccia and
Rogerson 2008) that show that there is considerable resource misallocation
across rms, which is driven by rm-specic distortions. While these studies
are silent about the nature of the distortions, they show that such distortions
lead to substantial aggregate productivity losses and that differences in the
degree of distortions would explain a signicant portion of why India and
China exhibit lower manufacturing productivity than the United States. This
study examines differences in business environment across rm size, and there-
The views expressed here do not necessarily represent the views of the World Bank, the Inter-
American Development Bank, or their Boards of Executive Directors. We would like to thank
Francis Teal and participants at the World Bank and Inter-American Development Bank seminars
and at the Latin American and Caribbean Economics Association (LACEA) conference in Bogota
and anonymous referees for valuable comments and suggestions. Contact the corresponding author,
Mary Hallward-Driemeier, at [email protected].
610 economi c development and cultural change
fore it helps to provide a better understanding of which dimensions of the
business environment matter and of the heterogeneity of distortions across
rms.
This study is also related to a number of studies that assess the effect of
different dimensions of the business environment on rms performance. Several
studies in the literature examine the effect of employment regulations on rms
adjustment and on labor market outcomes more generally (a few recent ex-
amples are Besley and Burgess 2004; Botero et al. 2004; Haltiwanger, Scar-
petta, and Schweiger 2006; Micco and Pages 2006; Petrin and Sivadasan 2006;
Autor, Kerr, and Kugler 2007; and Almeida and Carneiro 2009). Others look
at the effect of regulations of entry of rms on rm creation and growth (e.g.,
Djankov et al. 2002; Klapper, Laeven, and Rajan 2004). A number of them
investigate the importance of access to nance for rmdevelopment and growth
(e.g., Demirgu c-Kunt and Maksimovic 1998; Rajan and Zingales 1998; Beck,
Demirgu c-Kunt, and Maksimovic 2005; Galindo and Micco 2007). Overall,
these studies indicate that regulations on labor, entry barriers for rms, and
nancing conditions can have a direct impact on rms employment decisions.
However, other aspects of the business environment can affect these decisions
as well, raising costs and risks associated with doing business or by creating
barriers to competition. By affecting the overall growth prospects, issues from
the quality of infrastructure, the strength of property rights, the nature and
enforcement of business regulations, and the overall openness in the manage-
ment of public resources can all affect rms growth (World Bank 2004). This
study looks at the joint and relative importance of a number of dimensions
of the business environment on rms growth.
Unlike most existing studies that rely on variation in business environment
conditions across countries, sectors, or subnational territories (states), this study
exploits rm-level data on the business environment conditions faced by rms.
Yet there are a number of methodological issues that need to be addressed,
including potential measurement error, omitted variable bias, and endogeneity.
To minimize the risk of the rst two problems, the analysis is conducted with
a full set of sector-survey interaction dummies. In addition to controlling for
some potentially important omitted variables and measurement errors, they
also control for survey differences and differences in demand conditions. In
addition, incorporating multiple dimensions of the broader business environ-
ment simultaneously deals seriously with concerns of omitted variable bias of
papers that include only a single dimension, such as labor regulations or
nance.
To address potential endogeneity, this study uses location-sector-size averages
(minus individual rms own responses) of the business environment measures
Aterido, Hallward-Driemeier, and Page s 611
rather than individual rms own responses. The study shows that there is a
large potential degree of endogeneity in individual responses and that the
results vary considerably if such endogeneity is not accounted for. The paper
also makes the point that how these averages enter into the regressions matters
greatly. Using the location-sector-size average does not necessarily address the
endogeneity problemif the averages used are based on current size. As
objective conditions can be more challenging for smaller rms (e.g., smaller
rms are less likely to access nance and more likely to pay bribes), assigning
rms that grow substantially their current more favorable investment climate
(IC) conditions, rather than those faced by rms of their initial size, can
introduce a mechanical relationship between employment growth and many
of the IC measures. Matching the averaged indicators to initial size is needed
to avoid this problem. This methodological contribution is most visible in
estimating the effects of access to nance and corruption, which are the two
areas where endogeneity is most likely.
Using rm-level data allows us to focus on the question of whether the
impact of the business environment is neutral across rm size. Market failures
or policy-created distortions can create xed costs in the operation of businesses,
creating cost disadvantages for smaller rms (see Tybout [2000] for a review
of issues). The costs of dealing with credit market information imperfections
or with a complex and nontransparent regulatory environment are some ex-
amples of such nonlinearities. Large rms may also have more political inuence
and shape regulations and policies in their favor. On the other hand, smaller
rms may benet from lax enforcement of regulations or lower harassment
from corrupt public ofcials. Across the world, a sizable amount of public
resources is devoted to the development of micro, small, and medium rms,
yet unlocking potential constraints to their performance is another, perhaps
equally effective, development strategy. The main objective of this study is to
assess whether there are heterogeneous effects of unlocking business environ-
ment constraints.
Differential effects across rms of different size can stem from two sources.
First, there can be differences in the underlying objective conditions faced by
rms. The study focuses on the effect of four areas of the business environment:
access to nance, the regulatory environment, corruption, and access to infra-
structure. Second, the same conditions can potentially have differential or
nonlinear effects on employment growth by size. Thus, it could be that the
same extent of external nance is more benecial to smaller rms or that the
same extent of power outages is less damaging to larger rms because they
have access to generators.
We nd evidence of both effects. The study does nd that micro and small
rms have less access to formal nance, face signicantly greater interruptions
in infrastructure services, and pay more in bribesas a percentage of sales
than do larger rms. In turn, larger rms spend signicantly more time dealing
with ofcials and red tape. The results indicate signicant nonlinearities across
rm size in the impact of these conditions, too. A lack of nance and poor
infrastructure reduce the growth of medium and large rms. Business regu-
lations affect mostly the growth of small rms, which seemingly prefer to
remain small to keep below the radar of regulators.
The results also showthe importance of distinguishing between micro (fewer
than 10 employees) and small rms (1049 employees). By rening the def-
inition of small rms, the message is rened to show that while small rms
can face more challenging conditions, the very small or micro rms can actually
benetin relative, and in some cases even in absolute, termsfrom an
adverse business environment. This can potentially explain why very small
rms account for a larger share of employment in countries with an adverse
business climate.
The paper is structured as follows. Section II reviews the related literature.
Section III describes the data. Section IV examines how objective business
conditions vary by types of rms, emphasizing the different patterns across
different sizes of rms. Including lagged performance variables in these re-
gressions reinforces the need to address endogeneityand indicate the likely
directions that employment growth might have on different dimensions of
the business environment. Section V describes the impact of objective con-
ditions on employment growth using location-sector-size averages instead of
individual rm responses. Section VI conducts a number of robustness checks.
Section VII concludes.
II. Literature Review
There is a growing literature that assesses the effects of the set of factors,
policies, and institutions, commonly known as business environment or in-
vestment climate, on the performance of rms and economic growth. The
methodologies used by these studies are very diverse. A number of studies
have focused on cross-country variation to identify the effect of labor regulations
(Botero et al. 2004; Heckman and Pages 2004), regulation of entry (Djankov
et al. 2002), or a wide set of regulations (Loayza, Oviedo, and Serven 2006).
These studies relate objective (de jure) measures of regulation at the country
level to aggregate country outcomes. Although the results are suggestive of
the importance of appropriate regulations for business development, they suffer
from important methodological constraints, ranging from omitted variable
bias to endogeneity concerns.
Another group of studies employ a difference-in-differences method rst
developed by Rajan and Zingales (1998). These studies analyze the effect of
different aspects of the business environment at the country-industry level.
They show that nancially dependent industries grow faster in nancially
developed markets (Rajan and Zingales 1998) and that rms in industries
dominated by small rms grow faster in more nancially developed markets
(Beck et al. 2008). Studies have also found that industries in which entry is
more important exhibit less growth in countries with restrictions to rm entry
(Klapper et al. 2004). Finally, Micco and Pages (2006) show that industries
that are inherently more volatile create fewer jobs and are less developed in
countries with very restrictive hiring and ring regulations.
While improving on cross-country studies, the former studies suffer from
three potential shortcomings that can be addressed in our study. First, in most
cases variation in business environment conditions is captured with country-
level variables that reect de jure regulations or conditions, that is, the pro-
cedures and costs that would be incurred if rms fully complied with what
is on the books. However, there can be large gaps between what is on the
books and what is experienced on the ground. This is particularly true in
lower income countries and those with higher levels of corruption (Hallward-
Driemeier and Aterido 2007; Kaufmann, Kraay, and Mastruzzi 2007). In that
regard, it is desirable to have measures of de facto regulation.
Second, it is important to explore variation in the business environment
not only across countries but also within country boundariesacross subna-
tional areas and especially, across rm size and ownership.
1
Having disaggre-
gated variables allows for hypotheses to be tested without having to assume
that one country is the benchmark or best case against which to compare
other countries.
Third, while most studies look at the effects of individual dimensions of
the investment climate, our focus is on regressions that combine multiple
dimensions in a single regression. This not only addresses potential bias in
the estimates, it allows us to test directly which dimensions have the biggest
impact on rm performance.
A number of recent studies make use of the World Business Environment
1
A few studies assess differences in business environment conditions within countries. Besley and
Burgess (2004) and Ahsan and Pages (2009) exploit differences in labor market regulations across
Indian states to assess how different labor market regulations affect economic performance across
Indian states. Almeida and Carneiro (2009) assess the effect of variation of enforcement in labor
regulations within Brazil.
Survey (WBES)
2
a rm-level data set with 4,000 rms across 54 countries
to study the effect of the business environment on rm growth. Using sub-
jective rm-level data measures of the business environment, these studies
show the importance of nance, corruption, and property rights (Batra, Kauf-
man, and Stone 2003; Ayyagari, Demirgu c-Kunt, and Maksimovic 2008).
Some other studies examine the relationship between business environment
and rm growth using the newer Enterprise Survey, but for individual or
small groups of countries (Dollar, Hallward-Driemeier, and Mengistae [2005]
for India, Pakistan, Bangladesh, and China; Fisman and Svensson [2007] and
Rienikka and Svensson [2002] for Uganda; and Bigsten and Soderbom [2006],
which reviews the literature for Africa).
A central area of focus for this study is whether there are signicant dif-
ferences in the impact across rm types, particularly by rm size. The literature
has largely examined this issue with regard to access to nance, generally
nding that smaller rms are more constrained (Clark, Dollar, and Micco
2004; Love and Mylenko 2005; Galindo and Micco 2007). Beck et al. (2005)
is of particular interest here, as these authors include additional measures of
corruption and property rights using the WBES rm-level data set. Using
rms perceptions on potential constraints, they also nd patterns across coun-
tries, with small rms beneting most from greater nancial and institutional
development.
Our work expands on this literature in several dimensions. This study nearly
doubles the number of countries covered, with country samples approximately
ve to 10 times larger. Whereas Beck et al. (2005) use subjective rmresponses
as measures of the business environment at the rm level, we use objective
measures (i.e., time spent with ofcials dealing with permits and licenses
rather than a ranking of how burdensome red tape is on a scale of 15). Beck
et al. (2005) include a nal table to address potential endogeneity of business
environment conditions on rm growth using location-size averages of re-
sponses as instruments. However, as the averages are based on current size,
we show that this strategy does not actually address endogeneity. Once we
control for endogeneity, matching average conditions based on initial size, we
no longer nd that smaller rms benet most from improved access to nance.
We also use narrower bands on rm-size categories, which lead to some in-
2
Despite their similar names, the World Business Environment Survey (WBES), a one-off survey
in 1999, is not part of the larger and ongoing World Bank rm survey effort, the Enterprise
Surveys (ES). The earlier round primarily collected perception data regarding constraints and some
information on rm performance. In some regions rms only provided information on their em-
ployment range, that is, small, medium, large, making it impossible to use those data to study
differences between small rms below and above 10 employees.
teresting extensions on their results, and look at how results vary across country
groupings, focusing on measures of institutional development.
III. The Data
Our study is primarily based on the World Bank Enterprise Surveys (ES), a
newly available collection of rm-level data sets for developing countries as
well as a limited number of high-income countries.
3
Questionnaires are ad-
ministered within a framework of common guidelines in the design and
implementation.
4
The data set is assembled using the core survey, which is a
module of identical questions included in all questionnaires, thereby enabling
cross-country analyses.
While, in principle, the data set contains information for approximately
80,000 rms in more than 110 countries, complete data for the variables used
in our estimations are available for a subsample of approximately 56,000 rms
in 90 countries in seven different regions surveyed during the period 2000
2006. For some countries more than one survey is available (see table A1 in
the appendix for a complete list of countries and surveys available, as well as
the number of observations in each country).
5
The median sample size is 350
rms, with several large countries having substantially larger samples (Brazil,
China, India, Turkey, and Vietnam have samples over 1,500). The sample of
rms in each country is stratied by size, sector, and location.
6
Because of this
stratication, large enterprises are in general oversampled in the ESs compared
to their share in the number of rms, but not in terms of their contribution
to GDP. The unit of analysis is the Establishment in the manufacturing
and service sectors.
7
Most rms are registered with local authorities, although
they may be only in partial compliance with labor and tax authorities. Unlike
most rm-level data sets, these data span the manufacturing and service sectors.
Approximately 61% of the observations reect establishments in manufac-
turing, while the rest are in the service sector.
3
The term Enterprise Surveys includes surveys that have been implemented regionally under
other names, such as BEEPS, Investment Climate Surveys, and RPED surveys (World Bank 2004
and World Bank 2008).
4
From https://www.enterprisesurveys.org/Methodology/.
5
While efforts are shifting to building a panel data set, most repeat surveys have been additional
cross sections. There are approximately 2,000 rms that enter twice in the data set.
6
From http://www.enterprisesurveys.org/Methodology/default.aspx#weights: ES have been con-
ducted following simple random sampling or random stratied sampling. In a simple random
sample, all members of the population have the same probability of being selected and no weighting
of the observations is necessary. In a stratied random sample, all population units are grouped
within homogeneous groups and simple random samples are selected within each group.
7
In Europe and Central Asia, the unit is the rm. In all other regions it is the plant or
establishment. As over 90% are single-plant rms, the distinction is not likely to affect the results.
The World Bank enterprise survey data were developed to provide infor-
mation on rms performance as well as on aspects of the investment climate
faced by rms. Regarding rms performance, the data set contains information
on a number of variables such as current employment, investment, or sales.
It also contains retrospective information for these three variables, in most
cases for up to three periods before the year of reference of the survey, but in
some few cases only for two or one period before. We focus on the employment
level or growth of permanent workers as our outcome variables of interest.
We do not include other forms of employment, such as contract labor or
temporary workers, because there are differences in the consistency of reporting
of total employment (i.e., permanent plus temporary) across some countries.
Other important additional reasons to focus on the level and growth of per-
manent employment are that these variables are more likely to reect the
long-run performance of the rm, and their evolution is of higher concern for
policy makers, as nonpermanent forms of employment tend to be considered
substandard or less preferred by workers.
Following Davis and Haltiwanger (1992, 1999), our measure of employment
growth refers to the change of permanent employment during the period t
and 3 years before, divided by the rms simple average of permanent em-
ployment during the same period.
8
The measure is symmetric around zero
and bounded by values 2 and 2 and, while monotonically related to the
conventional growth rate and a second-order approximation of the logarithmic
rst difference, it allows computing meaningful growth rates for rms suffering
sharp expansions or contractions, avoiding any arbitrary treatment of outliers.
This is quite useful, because even when by construction our sample does not
register rm exit, some rms experience sharp changes in employment due
to their early/late stage of development.
9
Young, small rms may experience
very large growth in their labor force; bankrupt rms or rms on the verge
of closure may suffer very large contractions prior to their nal exit from the
market. Due to differences in data availability, employment growth refers to
different periods for different rms. We account for this difference in the
period of reference in our empirical specications.
The data also contain subjective and objective information on the business
8
For rms that have been in business less than 4 years, we compute employment growth based
on a 2-year difference or, if in business for only 2 years, 1-year difference. We then dene controls
EG(s) as one if we have computed employment growth s years apart and zero otherwise and include
these variables in our growth estimates to account for such differences in our estimations.
9
In principle, our sample allows identifying new entrants, that is, rms that have been in business
1 year or less. However, as the sampling frame can be a year or two old, there is an underrepre-
sentation of new entrants.
climate faced by individual rms, reecting conditions at the time of the
survey (period t). Regarding objective measures, the ones we focus on in this
study, the data set contains a set of measurements relative to various aspects
of the investment climate as they are being experienced by the rm. We focus
on three measures of access to nance and two measures of each of remaining
three areas of the business climate (see table 2 for a description and summary
statistics of these variables). We measure access to nance by means of the
share of investment nanced with funds external to the rm (sh-invest-n) or,
alternatively, with the share of working capital nanced externally (sh-work-
cap) or the share of sales on credit (sh-sales-cr). The sample means of those two
variables are 21%, 20%, and 42%, respectively.
We measure business regulatory conditions with the percentage of time
managers (mng-time) devote to dealing with authorities or, alternatively, with
the days rms devote to inspections. On average, rms spent 8.9% of their
time dealing with government authorities and almost 16 days dealing with
inspections during the year previous to the survey. We measure the extent of
corruption by means of either a dummy variable (bribe y-n) that captures
whether a rm had to pay a bribe or the percentage of sales paid on bribes
during the year previous to the sample (bribe %). To minimize the amount of
subreporting, the survey inquires about the amount paid in bribes by asking,
How much do rms similar to yours pay in bribes in order to get things
done? The survey means show a considerable incidence of corruption: 42%
of rms declare having paid bribes, with an average amount of 1.5% of sales
paid. Finally, we capture the extent of infrastructure bottlenecks by measuring
the number of days rms remained without power service (days-no power) as
well as the share of the average cargos value lost while in transit (loss-transit
%) during the last year. Again, the survey means show nonnegligible bottle-
necks, with an average of 38 days during the year previous to the survey
without power service, and 1.5% of sales lost during transit.
Finally, the data set also includes information on a number of rm char-
acteristics such as size, age, location, export activity, ownership, and sector.
This is an important feature of the data, which allows expanding previous
aggregate cross-country analyses attempting to assess the economic impact of
different aspects of the business climate to an analysis of whether business
environment conditions differ across, and affect differently, different types of
rms. The rm characteristics used in this study include four size classes (micro,
110 permanent employees; small, 1150; medium, 51200; and large, more
than 200); three age categories (young, 15 years; mature, 615; and older, more
than 15); two location types (smallcity, or cities with fewer than 1 million
inhabitants; and largecity, i.e., capital cities or with 1 million or more inhab-
TABLE 1
EMPLOYMENT GROWTH
Initial Size, t 3 Mean 25th Percentile Median 75th Percentile
Size 110 .220 .000 .000 .444
Size 1150 .081 .000 .000 .261
Size 51200 .033 .087 .000 .200
Size 200 .022 .115 .000 .122
Total .117 .000 .0000 .293
itants); ownership (whether foreign or government represents 10% or more of
ownership); whether the rm is an exporter (10% or more of sales); and a two-
digit industry classication.
The sample contains a large proportion of micro and small establishments
(39% micro and 34% small rms), with size measured with retrospective data
generally referring to 3 years before the current period.
10
Only 11% are large
rms. The sample also consists largely of young rms, with age also referred
to the initial period for which information for a rm is available. Table A2
shows that 31% of rms in our sample were 5 or fewer years old, and 38%
were between 6 and 15 years. According to the 10% thresholds noted above,
11% of the rms are foreign owned, and 23% of rms are exporters.
The sample is well diversied across regions; 27.2% of the sample corre-
sponds to East Europe and Central Asian countries; 20.9% of rms in the
sample are from Latin America and the Caribbean; 16% are from sub-Saharan
Africa; 14.5% from East Asia and 10.2% from South Asia (see table A2 in
the appendix). There is a smaller, but still signicant representation of North
Africa and the Middle East region (6.7%) and high-income European countries
(4.6%).
Table 1 reports the mean and standard deviation of the employment growth
measure across rm size, with size measured as the initial size (3, 2, or 1 year
ago). There is positive employment growth across all rm sizes except for
larger rms, which on average experience negative growth. The average rm-
level employment growth, according to our measure, is 12%. Micro rms
grew over the average, while small and medium growth was below average.
IV. Variations in Objective Investment Climate Conditions by Firms
We start our analysis by examining differences in the reported business en-
vironment conditions across rms for the different investment climate variables
dened in Section III and summarized in table 2. To do so, we estimate the
following specication:
10
In some cases the reference period is less than 3 years due to rm age or data availability.
TABLE 2
VARIABLE DESCRIPTION
Variable Description Mean SD
Emp-gr Employment growth .117 .411
Labor,t Number of employees in period t (log) 3.112 1.618
Sh-invest-n Share of investments nanced externally .209 .347
Sh-work-cap-n Share of working capital nanced externally .201 .137
Sh-sales-cr Percentage of sales sold on credit .424 .396
Mng-time % of managements time dealing with regulations 8.872 15.104
Days-inspections Total days spent on inspections during last year 15.990 31.721
Bribe y-n Bribes given to get things done (yes-no) .425 .494
Bribe (%) Amount of bribe paid to get things done 1.507 4.242
Days-no power Number of power outages experienced during the last year 38.143 83.993
Loss-transit (%) Percentage of the average cargos value lost while in transit 1.547 5.736
ICvariable pd d Small d Medium d Large
ict 1 2 ict 3 ict 4 ict
d Foreign d Exporter d Mature d Older
5 ict 6 ict 7 ijct 8 ijct
(1)
d Government d Smallcity d Expand
9 ijc 10 ijc 11 ict1
d Contract l #l .
12 ict1 c j ict
To capture differences in the objective conditions facing rms of different
sizes we include size dummies (micro-enterprises is the omitted category),
with the size measured by number of employees at the time of the survey
(period t). We also control for whether the rm is foreign or government
owned, whether it is an exporter, whether it is located in a large city (or
capital) or not (large city is the omitted), as well as age categories (young is
the excluded category), all measured at period t.
We also include variables Expand
ict-1
and Contract
ict-1
, the rate of employment
expansion or contraction in rm i, country c, between period t 1 and period
t 3. The omitted category is that employment remained unchanged in the
last period. The measures of contraction and expansion are thus lagged one
period, to have them predetermined, but the claims here are not causal. We
are simply interested in describing the patterns across rms, examining
whether there are differences in the reported objective characteristics across
expanding and contracting rms. These coefcients will also provide infor-
mation on whether endogeneity is likely to be an issue in the employment
regressions.
Our focus is on within-country variations rather than across countries; there-
fore, we include a full set of sector-survey interaction dummies ( # ). l l
j S
This allows us to control not only for potential omitted variables at the country
level but also for some possible measurement errors or differences in survey
implementation across countries and sectors.
Table 3 reports the results of estimating this specication for the nine IC
variables. The results show that there are signicant differences across rms
in most of the variables in each of the four areas of the investment climate
(access to capital, business regulations, corruption, and infrastructure). Again,
the focus is on differences across rms of different size. Regarding access to
nance, our estimates indicate that controlling for rm characteristics and
sector-survey interaction dummies, access to nance, measured as share of
investment or working capital nanced externally, grows monotonically with
rm size, while the share of sales on credit peaks for medium-sized rms.
Overall, large rms have more access to nance than medium, small, and
micro enterprises.
Regarding business regulations, micro rms report a lower share of man-
agement time devoted to dealing with government regulations than small,
medium, and large rms. The pattern again is increasing in rm size, although
in this case medium rms spend more time dealing with regulations than
large rms. Days of inspections also increases monotonically with size, with
large rms spending more than triple the time dealing with inspections than
small rms.
Regarding corruption, the incidence of bribes is the highest among small
rms. In addition, medium and large rms pay less in bribesas a percentage
of salesthan small or micro rms. This can reect small rms having rel-
atively larger xed payments or less recourse to avoid making payments. The
payments may also be correlated with the degree of compliance with regu-
lations; micro and small rms may not meet all of the requirements and have
to pay ofcials to maintain this position.
For infrastructure, the frequency of power outages hits small and medium
rms hardest and large rms the least. Larger rms are more likely to have
alternative sources of electricity (given the substantial xed costs of owning
and operating a generator). It is also possible that larger rms have a greater
choice among possible locations and thus choose to operate in areas with more
reliable electricity, a hypothesis that is tested for below. Similarly, losses in
transit are largest among small and medium-sized rms.
The specications reported in table 3 provide some other important results.
Perhaps the most important one is that there are indications that reported
business climate conditions are related to rm performance. For example, rms
that expanded or contracted in the period previous to the survey report a
higher dependency on external funds and more management time dealing with
ofcials relative to rms that were stable in size. On their part, contracting
rms spend more time in inspections than expanding or stable rms. Similarly,
the incidence of corruption (whether a rm pays bribes) appears to be more
severe for rms that are expanding or contracting relative to stable rms. It
appears corruption and dealing with authorities are the price to pay to either
get ahead or preserve the rm in bad times. Finally, rms that are growing
or contracting report more blackouts and more losses in transit than rms
that are stable in size. While it may be that objective conditions vary across
stable rms and the rest, it is also likely that power outages and transit losses
are more distortive for growing or contracting rms, and therefore, they recall
more such events and report higher values in those questions than rms that
are stable.
The former regressions suggest that objective business conditions vary sys-
tematically across rms of different size and economic performance. Yet none
of these claims should be interpreted as causal, as one could also expect causality
to run in the opposite direction, that is, the constraints rms face determine
their size and their performance. However, they do caution of a possible
endogeneity of reported investment climate conditions relative to rm per-
formance. We address this topic in the next section of the paper.
V. Methodology to Estimate the Impact of Investment Climate
Conditions on Firm Employment Growth
We estimate the effect of investment climate (IC) conditions on employment
growth by means of the following specication:
Empg pb Small b Medium b Large
ict,ts 0 ict3 1 ict3 2 ict3
b IC b Small #IC b Medium #IC
3k ikct 4k ict3 ikct 5k ict3 ikct
k k k
b Large #IC b Foreign b Government (2)
6k ict3 ikct 7 ict 8 ict
k
b Exporter b Mature b Older b Smallcity
9 ict 10 ict3 11 ict3 12 ict
b EG(s) b Empg-cell l #l ,
13 ic 14 ic j s ict
where Empg
ict,ts
refers to the growth of permanent employment of rm i in
country c between period t and t s (generally 3 years before, but in a few
cases, where information for period t 3 is not available, at periods t 1
or t 2). Differences in the period of denition of employment growth are
identied and controlled for by the dummies EG(s)
i
, which take the value of
1 if employment growth is computed between periods t and t s and zero
otherwise. Investment climate conditions faced by rm i in country c are
measured along four dimensions (k): nance, regulations, corruption, and in-
frastructure, and are denoted by . While most previous studies focus on IC
ikc
the effect of one aspect of the investment climate, our concern is that omitted
622
T
A
B
L
E
3
I
N
V
E
S
T
M
E
N
T
C
L
I
M
A
T
E
B
Y
F
I
R
M
S
C
H
A
R
A
C
T
E
R
I
S
T
I
C
S
D
e
p
e
n
d
e
n
t
V
a
r
i
a
b
l
e
:
I
n
v
e
s
t
m
e
n
t
C
l
i
m
a
t
e
V
a
r
i
a
b
l
e
S
h
-
i
n
v
e
s
t
-
n
(
%
)
(
1
)
S
h
-
w
o
r
k
-
c
a
p
i
t
a
l
-
n
(
%
)
(
2
)
S
h
-
s
a
l
e
s
-
c
r
(
%
)
(
3
)
M
n
g
-
t
i
m
e
(
%
)
(
4
)
D
a
y
s
-
i
n
s
p
e
c
t
i
o
n
s
(
l
o
g
)
(
5
)
B
r
i
b
e
(
Y
e
s
/
N
o
)
(
6
)
B
r
i
b
e
(
%
S
a
l
e
s
)
(
7
)
D
a
y
s
-
n
o
p
o
w
e
r
(
l
o
g
)
(
8
)
L
o
s
s
-
t
r
a
n
s
i
t
(
%
)
(
9
)
S
m
a
l
l
.
0
4
9
*
*
*
.
0
4
5
*
*
*
.
0
7
2
*
*
*
1
.
2
3
9
*
*
*
.
2
3
8
*
*
*
.
0
2
9
*
*
*
.
0
5
4
.
0
6
1
*
*
*
.
1
4
3
*
(
.
0
0
4
)
(
.
0
0
3
)
(
.
0
0
4
)
(
.
1
5
2
)
(
.
0
1
4
)
(
.
0
0
6
)
(
.
0
5
1
)
(
.
0
1
5
)
(
.
0
7
4
)
M
e
d
i
u
m
.
0
9
3
*
*
*
.
0
8
7
*
*
*
.
0
9
6
*
*
*
1
.
7
6
7
*
*
*
.
5
0
8
*
*
*
.
0
1
4
*
.
2
5
0
*
*
*
.
0
5
0
*
*
*
.
2
0
2
*
*
(
.
0
0
5
)
(
.
0
0
4
)
(
.
0
0
5
)
(
.
1
9
9
)
(
.
0
1
7
)
(
.
0
0
8
)
(
.
0
6
0
)
(
.
0
1
8
)
(
.
0
8
6
)
L
a
r
g
e
.
1
1
9
*
*
*
.
1
1
7
*
*
*
.
0
9
0
*
*
*
1
.
4
6
8
*
*
*
.
7
1
5
*
*
*
.
0
0
1
.
3
6
1
*
*
*
.
0
8
8
*
*
*
.
0
3
9
(
.
0
0
7
)
(
.
0
0
5
)
(
.
0
0
6
)
(
.
2
3
4
)
(
.
0
2
2
)
(
.
0
1
0
)
(
.
0
6
7
)
(
.
0
2
2
)
(
.
0
9
9
)
M
a
t
u
r
e
.
0
1
3
*
*
*
.
0
0
6
*
.
0
0
9
*
*
.
2
7
8
*
.
0
3
8
*
*
.
0
0
4
.
1
1
4
*
.
0
6
3
*
*
*
.
1
9
2
*
*
(
.
0
0
4
)
(
.
0
0
3
)
(
.
0
0
4
)
(
.
1
5
0
)
(
.
0
1
5
)
(
.
0
0
7
)
(
.
0
5
8
)
(
.
0
1
6
)
(
.
0
8
3
)
O
l
d
e
r
.
0
1
0
*
*
.
0
1
1
*
*
*
.
0
0
6
.
4
9
4
*
*
*
.
0
5
9
*
*
*
.
0
2
0
*
*
.
2
3
6
*
*
*
.
0
7
2
*
*
*
.
2
0
4
*
*
(
.
0
0
5
)
(
.
0
0
4
)
(
.
0
0
4
)
(
.
1
8
0
)
(
.
0
1
7
)
(
.
0
0
8
)
(
.
0
6
5
)
(
.
0
1
8
)
(
.
0
8
9
)
E
x
p
o
r
t
e
r
.
0
3
0
*
*
*
.
0
3
7
*
*
*
.
0
4
8
*
*
*
.
5
0
4
*
*
*
.
0
8
5
*
*
*
.
0
4
0
*
*
*
.
0
5
7
.
0
5
8
*
*
*
.
2
1
7
*
*
*
(
.
0
0
5
)
(
.
0
0
3
)
(
.
0
0
4
)
(
.
1
5
9
)
(
.
0
1
4
)
(
.
0
0
7
)
(
.
0
5
0
)
(
.
0
1
5
)
(
.
0
6
7
)
F
o
r
e
i
g
n
.
0
4
1
*
*
*
.
0
2
5
*
*
*
.
0
4
3
*
*
*
.
0
9
4
.
0
6
1
*
*
*
.
0
1
6
*
*
.
0
8
0
.
0
3
8
*
*
.
2
1
3
*
*
(
.
0
0
6
)
(
.
0
0
4
)
(
.
0
0
5
)
(
.
1
8
3
)
(
.
0
1
8
)
(
.
0
0
8
)
(
.
0
6
4
)
(
.
0
1
8
)
(
.
0
8
8
)
S
m
a
l
l
c
i
t
y
.
0
1
0
*
*
*
.
0
0
9
*
*
*
.
0
1
4
*
*
*
.
1
9
0
.
0
2
5
*
.
0
5
3
*
*
*
.
2
2
8
*
*
*
.
1
1
8
*
*
*
.
0
7
2
(
.
0
0
4
)
(
.
0
0
3
)
(
.
0
0
3
)
(
.
1
4
4
)
(
.
0
1
3
)
(
.
0
0
6
)
(
.
0
4
3
)
(
.
0
1
3
)
(
.
0
6
3
)
623
G
o
v
e
r
n
m
e
n
t
.
0
3
5
*
*
*
.
0
1
3
*
*
.
0
4
1
*
*
*
.
5
9
9
*
*
.
0
0
0
.
1
3
2
*
*
*
.
4
9
6
*
*
*
.
1
1
2
*
*
*
.
0
7
8
(
.
0
0
7
)
(
.
0
0
5
)
(
.
0
0
6
)
(
.
2
4
2
)
(
.
0
2
6
)
(
.
0
1
0
)
(
.
0
6
4
)
(
.
0
2
5
)
(
.
1
2
5
)
E
x
p
a
n
d
.
0
4
1
*
*
*
.
0
2
9
*
*
*
.
0
2
9
*
*
*
.
4
9
1
*
*
.
0
1
8
.
0
5
3
*
*
*
.
1
4
2
*
*
.
0
9
8
*
*
*
.
2
4
3
*
*
(
.
0
0
6
)
(
.
0
0
4
)
(
.
0
0
5
)
(
.
2
0
9
)
(
.
0
1
9
)
(
.
0
0
8
)
(
.
0
6
9
)
(
.
0
2
0
)
(
.
0
9
8
)
C
o
n
t
r
a
c
t
.
0
2
7
*
*
*
.
0
4
0
*
*
*
.
0
4
4
*
*
*
1
.
7
7
1
*
*
*
.
1
4
0
*
*
*
.
0
8
8
*
*
*
.
5
2
3
*
*
*
.
0
5
9
*
*
.
4
3
5
*
*
(
.
0
0
9
)
(
.
0
0
6
)
(
.
0
0
7
)
(
.
3
0
8
)
(
.
0
2
7
)
(
.
0
1
2
)
(
.
1
0
5
)
(
.
0
2
8
)
(
.
1
6
9
)
S
u
r
v
e
y
#
S
e
c
t
o
r
x
e
d
e
f
f
e
c
t
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
C
o
n
s
t
a
n
t
.
1
3
3
*
*
*
.
1
2
4
*
*
*
.
3
5
9
*
*
*
7
.
4
7
7
*
*
*
1
.
5
1
6
*
*
*
1
.
8
1
2
*
*
*
1
.
5
7
0
*
*
*
1
.
4
6
5
*
*
*
(
.
0
0
5
)
(
.
0
0
3
)
(
.
0
0
5
)
(
.
1
8
4
)
(
.
0
1
8
)
(
.
0
6
6
)
(
.
0
1
8
)
(
.
0
9
7
)
O
b
s
e
r
v
a
t
i
o
n
s
4
2
,
5
1
9
6
3
,
0
2
0
6
5
,
5
3
1
6
3
,
2
9
5
5
6
,
7
9
4
5
,
1
7
0
7
5
0
,
6
1
0
6
1
,
3
0
9
4
5
,
2
7
2
R
2
.
1
9
.
2
1
.
3
1
.
1
8
.
3
1
.
1
7
.
1
0
.
5
3
.
0
7
N
o
t
e
.
F
i
r
m
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
a
r
e
a
t
t
i
m
e
t
.
C
o
l
u
m
n
6
i
s
a
d
p
r
o
b
i
t
,
s
o
c
o
e
f
c
i
e
n
t
s
a
r
e
m
a
r
g
i
n
a
l
e
f
f
e
c
t
s
a
n
d
R
2
i
s
P
s
e
u
d
o
R
-
s
q
u
a
r
e
d
.
R
o
b
u
s
t
s
t
a
n
d
a
r
d
e
r
r
o
r
s
i
n
p
a
r
e
n
t
h
e
s
e
s
.
*
S
i
g
n
i
c
a
n
t
a
t
1
0
%
.
*
*
S
i
g
n
i
c
a
n
t
a
t
5
%
.
*
*
*
S
i
g
n
i
c
a
n
t
a
t
1
%
.
variables may bias results. After all, environments characterized by ineffective
business regulations or high incidence of corruption are also often characterized
by poor infrastructure or low access to nance. To overcome some of these
concerns, we control for different indicators of business climate conditions and
their interactions with rm size simultaneously. Size dummies are dened in
four groups: micro, small, medium, and large, measured at the initial period
of observation. The omitted category is micro enterprises, employing between
1 and 10 employees. Age of the rm is specied in three categories: young,
from 1 to 5 years old; mature, between 5 and 15 years old; and older, above
15 years old; and it also refers to the initial period of observation. The other
rm characteristics are dated at period t, the only period for which such
information is available. We also control for ownership (foreign, government),
exporting status (exporter), and location, with variables dened as reported in
Section III.
To account for differences in demand conditions and productive structure,
the regressions also include a full set of sector-country interaction dummies.
11
Thus, our focus is on within-country rather than across-country variation. This
decision reects several factors. First, there is substantial variation of investment
climate indicators within countries, and we want to test whether its impact
is signicant. Second, it controls for potential omitted variables at the country
level. In addition, it could control for some possible measurement errors or
differences in implementation across countries. As Haltiwanger and Schweiger
(2005) note, in the ES there are some countries with particularly high net
employment growth rates. However, there is a strong country-specic com-
ponent, such that country effects yield relative magnitudes of job creation and
employment growth. They also noted that the relationships of employment
variations between size and age classes appear right. Thus, small and younger
rms have higher job creation and destruction than older and larger rms
within the country, when considering total employment. Therefore, statistical
methods that remove country level effects are appropriate to draw conclusions
within countries on the difference of the impact of investment climate on
different types of rms.
The construction of the investment climate variables requires some IC
ikc
explanation. A number of previous studies assess the investment climate ex-
perienced by rm i, using rm-level subjective measures. Such studies rely
on answers to questions of the form, How much (the dimension k of the
business climate) is an obstacle for the growth of your rm? with possible
answers ranking between 1 and 5 (Johnson, McMillan, and Woodruff 2002;
11
See table A2 for the list of sectors included.
Beck et al. 2005; Pierre and Scarpetta 2006; Ayyagari et al. 2008). Subjective
assessments, while insightful, have a number of possible drawbacks. There is
a concern that rms perceptions of the business environment reect idiosyn-
cratic differences in the degree of optimism or pessimism of the individuals
responding to the survey. Answers are also likely to be strongly inuenced by
the experience and performance of the rm (see Hallward-Driemeier and Ate-
rido 2007).
Our analysis uses objective measures to gauge the relative importance of
the investment climate to rms. While, in principle, objective measures may
be less likely to be inuenced by rms performance than subjective measures,
the results presented in table 3 suggest that individual rms reported objective
investment climate conditions are likely to be endogenous to rmperformance,
a feature that we need to take into account when estimating the effect of
investment climate variables on rm performance.
To account for endogeneity, we construct a measure of the IC conditions
faced by rm i on dimension k in country c, by averaging the responses of
rms in the same country-location-sector-size cell, excluding the observation
from rm i in computing that average.
12
To ensure adequate numbers of rms
in each location-sector-size cell average, we drop one dimension of the cell
until an adequate number is reached (i.e., rst we substitute country-location-
size bins, then country-sector-size).
13
This captures the broader environment in which the rm operates and allows
the individual rms own contribution to the average to be excluded. The
approach also has the benet of not losing those observations where a rm
did not answer all the individual investment climate questions.
14
12
In some specications we also assess whether results vary if we instead compute IC conditions
by averaging the responses of rms in the same country-location-sector bins (i.e., excluding size
as a dimension). Here we use a ner disaggregation of location. Above, locations are aggregated
by city size (e.g., capital cities and those with a population above 1 million, and cities below 1
million people). In the country-location-sector averages, we construct the cells based on the actual
cities or towns the rms are located in. If the representation is low, locations are aggregated based
on ve size groups (capital cities or more than 1 million; 500,0001 million; 250,000500,000;
50,000250,000; less than 50,000).
13
If we were to keep always the original four dimensions same country-location-sector-size cell
there are approximate 5,000 observations that are lost and results do not change substantially. We
also used different minimum cell-size requirements (410 rms) and found results were robust.
Results presented use the cutoff of at least four rms in a cell.
14
This approach is very similar to using location-sector-size dummies as instruments (except that
the rms own value is not excluded in this calculation, the number of observations averaged in
a cell may be very small, and the additional observations cannot be recovered if a single investment
climate variable is not available). The test of overidentifying restrictions could not be estimated
using the full specication due to the large number of dummies and instruments. However, in
In constructing these measures, we actually take two steps before entering
the averages into the regressions, with the second one being particularly im-
portant if endogeneity concerns are to be addressed. An issue is how the
investment climate indicators are calculated and matched to rms in the
regression. This requires differentiating between a rms current sizethat
can be important in the conditions they report facing nowand their initial
sizeand the average conditions faced by rms of the size they were initially.
Simply giving rms the average value of rms in their current size category
is giving them the ex post conditions, not the initial conditions. To justify
this, one has to assume that the conditions faced by rms have remained
constant over time, regardless of how they may have grown. The alternative
we present assumes instead that conditions faced by rms are likely to vary
as they grow. Rather, we make the less restrictive assumption that, within a
location and sector, conditions facing rms of a certain size now are the same
as the conditions facing the same-sized rm in the earlier period. For example,
we can use the information on conditions facing micro rms now to measure
what micro rms faced 3 years before.
Thus, in the rst step, we use the information that rms report about
investment climate conditions they face in the current period. While it would
be desirable to have the conditions they faced in the initial period 3 years
ago, this information was not collected. However, we also recognize that rms
that have recently changed size may report very different IC conditions relative
to what is typical for rms that have been in a given size category for a while.
To account for this we use a rms average size in constructing the location-
sector-size averages of investment climate conditions. We do so by constructing
averages of investment climate indicators within cells, which are dened ac-
cording to rms average size.
15
In the second step, we match these averaged
indicators to rmsbased on their initial, not current, size.
16
This procedure has a number of advantages. The rst is that as advanced
specications not using a full set of sector-survey interaction dummies, the restrictions could not
be rejected at the 0.3 level.
15
We repeated the exercise using current size in dening the cells for calculating the IC variables
and did not nd signicant differences. For rms that remain in the same size category there is
no difference between the two approaches.
16
An example may be useful to clarify our procedure. Let us consider, for example, a textile-
producing rm in the capital city of Chile, which was small in period t 3 and has grown to
be of medium size in period t, and the average size in the two periods is medium. We use its
responses on the investment climate indicators to construct the average of Santiago-textile-medium-
sized rms. However, this average is not what is then used for that rm in the regressions. Since
we want to capture the average IC conditions of that rm in period t 3, when it was small, we
match it to the average reported IC investment climate conditions of Santiago-textile-small rms.
in the former paragraph, using IC averages based on rms average size allows
us to obtain a more accurate measure of the IC conditions facing a rm in a
given size category. Thus, if, for example, rms that have recently expanded
report lower access to capital than rms that have been large for a while, then
accounting for the observations of growing rms in the construction of the
IC may underestimate the average access to capital of large rms. A similar
effect may happen for rms that have recently contracted and report access to
capital or other IC which are more representative of rms that are large than
of rms that are small. Considering rms based on their average size ameliorates
this problem.
Second, constructing a measure of IC conditions based on the initial instead
of the current size is a key feature to minimize endogeneity problems. Thus,
if we consider again the example of a rm that expands from being small in
period t 3 to medium in t, it matters whether the rm is assigned the
average IC conditions of small or medium rms. The main point is that
matching based on current size reintroduces endogeneity, as rms that grow
are assigned higher access to credit just because they have grown, which creates
a positive correlation between the error term and the IC conditions. Our
procedure assumes that the conditions that matter for employment growth
are those prevalent in the past and that objective conditions remain constant
over time for size classes of rms, not for individual rms themselves. These
hypotheses allow us to construct measures that are exogenous to the rm and
control for the fact that rms may choose to grow/or not grow to face a
determined set of IC. We show in our estimates the importance of how the
IC variables are constructed for the results.
While we believe our constructed IC variables go a long way toward ad-
dressing endogeneity, a lingering concern is that our results may be capturing
the effects of shocks that are correlated across rms, which drive our constructed
average IC as well as individual rm performance. We reduce any possible
biases brought by these correlations by introducing as an additional regressor
the average employment growth of the cell used to compute the IC conditions
of rm i (excluding as usual, the employment growth of rm i in that cal-
culation). This variable is denoted as Empg-cell
i
.
VI. Results
We now describe the results attained with our estimation procedure and
presented in table 4, column 1. The results indicate important nonlinearities
in the effects of business environment constraints across rm size. They also
suggest that a weak business environment tends to hurt the growth of small,
medium, and large rms and benet in relative termsand for some variables,
628
TABLE 4
IMPACT OF INVESTMENT CLIMATE ON EMPLOYMENT GROWTH
Dependent Variable: Employment Growth (Emp(t) Emp(t 3))/((Emp(t) Emp(t 3))/2)
Investment Climate Variable
(ICvar)
ICvar avg
Matched
to Initial
Size
(1)
Individual
Firm Re-
sponse
(2)
ICvar as in
Col. 1,
with Col. 2
Sample
(3)
ICvar avg
Matched
to Current
Size
(4)
ICvar Cell Av-
erage: Survey-
Sector-
Location
(5)
Sh-invest-n .055 .083*** .095 .583*** .008
(.059) (.018) (.074) (.075) (.033)
(Small) # Sh-invest-n .092 .022 .090 .039 .047
(.065) (.023) (.077) (.070) (.040)
(Medium) # Sh-invest-n .130* .028 .159* .024 .115**
(.075) (.024) (.086) (.083) (.045)
(Large) # Sh-invest-n .136* .019 .221*** .134 .110**
(.071) (.028) (.085) (.087) (.056)
Sh-mng-time .003** .001*** .005*** .011*** .002***
(.001) (.000) (.002) (.002) (.001)
(Small) # Sh-mng-time .005*** .002*** .008*** .004*** .004***
(.001) (.000) (.002) (.001) (.001)
(Medium) # Sh-mng-time .003 .001** .005** .003* .002**
(.002) (.001) (.002) (.002) (.001)
(Large) # Sh-mng-time .003* .002*** .007*** .005*** .002**
(.002) (.001) (.002) (.002) (.001)
Bribe y-n .051 .027*** .043 .015 .025
(.041) (.010) (.046) (.045) (.021)
(Small) # Bribe y_n .004 .035*** .023 .031 .012
(.045) (.014) (.047) (.048) (.024)
(Medium) # Bribe y-n .073 .058*** .032 .104* .041
(.054) (.016) (.053) (.058) (.030)
(Large) # Bribe y-n .056 .026 .025 .099* .029
(.051) (.017) (.055) (.058) (.033)
Days-no power .008 .009*** .016 .002 .006
(.008) (.003) (.010) (.009) (.004)
(Small) # Days-no power .006 .008* .011 .007 .008**
(.005) (.004) (.007) (.006) (.004)
(Medium) # Days-no power .014* .010** .021** .024*** .015***
(.008) (.005) (.009) (.008) (.005)
(Large) # Days-no power .015* .010** .020** .019** .014**
(.008) (.005) (.009) (.009) (.006)
(Small) .110*** .108*** .071** .136*** .103***
(.028) (.013) (.028) (.031) (.017)
(Medium) .149*** .162*** .151*** .143*** .162***
(.032) (.017) (.032) (.035) (.021)
(Large) .210*** .239*** .217*** .181*** .224***
(.031) (.019) (.034) (.037) (.023)
Firm controls Yes Yes Yes Yes Yes
Survey # Sector xed effects Yes Yes Yes Yes Yes
Observations 56,880 28,619 28,616 56,880 55,612
R
2
.15 .18 .18 .17 .15
Note. Firm controls: dummies for rm age (t 3), foreign or government owned, exporter status, if
in small city, I base years, average employment growth in cell, and a constant term. Size is t 3. See
denitions in Sec. III. Columns 1 and 3 cells based on survey-sector-location-avg-size (excluding own);
col. 4 uses current size. Standard errors are clustered on survey-location-sector-size.
* Signicant at 10%.
** Signicant at 5%.
*** Signicant at 1%.
even in absolute termsthe growth of micro rms. As addressing endogeneity
is one of our chief concerns, columns 25 demonstrate the signicance of
alternative means of addressing itor of failing to do so.
For all specications, coefcients on rm characteristics (other than size),
the set of sector-survey interaction dummies, and the constant are not reported
due to space constraints. As expected, the coefcients on rm sizes categories
indicate that employment growth declines monotonically with rm size.
17
We
focus next on the coefcients on the investment climate and their interactions
with rm size.
A. Access to Finance
The impact of better access or lower cost of nances on rms growth has
been studied by a number of authors.
18
Beck et al. (2005) argue that nancial
access should favor small rms. In their work, they nd that rms self-reported
(subjective) constraints on access and cost to nance are associated with lower
growth of small rms relative to large rms. Table 3 showed that there are
differences in the amounts of nance available, which could already explain
why smaller rms are more likely to complain about nance. Here we test
whether, even for the same amount of nancing (measured as percentage of
investment nanced externally), the impact would be different across different
types of rms. Note that nance is dened such that a larger number implies
better access to nance. The results show strong nonlinearities across rm size,
and unlike earlier results, we nd a positive effect of increased access to nance
on the employment growth of medium and large rms and no signicant
effect among micro and small enterprises.
19
We investigate the sensitivity of
these results to different ways of estimating the IC measures and nd the
results to be quite sensitive to whether and how we account for endogeneity.
It is useful to compare our results in column 1 with those obtained using
rms individual own responses (table 4, col. 2). While the number of ob-
servations declinessince not all rms report information for such variables
17
There is a debate in the literature whether it is appropriate to use initial size or average size
in rm growth regressions (see Teal 1999; Bigsten and Soderbom 2006). Proponents of Gibrats
Law favor the latter, although even using average size can still lead to the result that smaller rms
grow faster than larger rms. We repeated our regressions dening the size controls and the
investment climate interactions based on average size and found very robust results. (Note that,
to keep from reintroducing endogeneity, what one must avoid is using a measure of size from a
time period prior to the one used in matching the investment climate averages to rms.)
18
See, e.g., Demirgu c-Kunt and Maksimov (1998), Rajan and Zingales (1998), and Beck et al.
(2005).
19
The test that the overall effect of access to nance on medium and large rms is equal to zero
is rejected at the 10% signicance level for large rms and at the 12% level for medium rms.
we nd that using reported individual measures, access to nance has a positive
effect across all rms that does not differ substantially across rm size, sug-
gesting that not controlling for endogeneity may overestimate the effect of
nance on micro and small rms. Note that differences in coefcients across
columns 1 and 2 are not due to the smaller number of observations in column
2. Column 3 reports the results of reproducing the estimates presented in
column 1 for the same sample as in column 2, and the results do not vary
much: if anything, the nding that nance helps medium and larger rms if
endogeneity is properly controlled for becomes even stronger.
In table 4, column 4, we report the results when using IC measures that
are constructed as reported in Section III, but with the only difference that
the averages within country-city-sector-size cells are matched to current rather
than initial size. As mentioned, this is likely to reintroduce some level of
endogeneity, as growing rms are matched with higher levels of access to
nance. In this case we nd that all rms benet from access to nance, and
while the coefcients point to smaller benets for larger rms, differences
across rm size are not statistically signicant.
The fact that results are fairly similar when using individual rm responses
or cell averages matched to current size, and at the same time quite different
from the results when measuring IC based on cell averages matched to initial
size, suggests that (i) endogeneity is likely to bias the estimates when not
accounting for it and that (ii) using averages matched on current size does
not solve the endogeneity problem. Quite the opposite, it reintroduces en-
dogeneity in the IC measures due to the match.
Our ndings also suggest that the results found by previous studies that
access to nance benets mostly the smaller rms may be driven by endo-
geneity. In our sample, there are as many as 16% of the rms that start up
as micro and transit to larger rms. Depending on the procedure, such rms
either report higher access to capital for endogenous reasons (i.e., because they
have grown) or are matched with higher access to capital than the average of
initially small rms (i.e., are given the average access to capital of larger rms).
This creates a positive bias in the coefcient for micro and small rms. Con-
versely about 7% of large rms shrink and become smaller rms in our sample.
As they become smaller, they are likely to report lower access to nance or,
depending on the method, be matched to the access to capital of smaller rms.
This creates a downward bias in the coefcient of larger rms.
Table 4, column 5, presents the results of further assessing the robustness
of our main results when constructing the IC averages with country-sector-
location averages. Results conrm the ndings presented in our basic speci-
cation (col. 1): increased access to external nance to fund investments seems
to benet mostly medium-sized and larger rms.
Why would micro or small rms not be positively affected by increased
access to nance? Jeong and Townsend (2007) show that with a wide range
of talent among micro-entrepreneurs, access to nance may not be the binding
constraint they face in growing their rms. If access to nance is restricted,
individuals with high entrepreneurial talent but low wealth are size con-
strained, and the demand for wage employment is low, which drives many
individuals to become entrepreneurs even when they do not have much en-
trepreneurial talent. Increased access to nance sorts the good from the less
talented entrepreneurs. Those who are good can grow their rms; those who
have lower entrepreneurial talent suffer as wages increase due to an increased
demand for labor. Our estimates, particularly the negative, albeit not statis-
tically signicant coefcient on micro rms, may then reect this selection
effect.
B. Regulatory Environment
Measures of the regulatory environment show that business regulations mea-
sured as the share of time that management devotes to dealing with govern-
ment regulations have differential effects across rm size. Our results suggest
that regulations create growth bottlenecks for small rms, which may fear/
avoid expansion to limit their exposure to red tape and increased enforcement.
Testing the overall signicance by size conrms that the overall effect of
regulations on employment growth for small rms (the sum of the coefcient
on micro and small rms) is signicantly different from zero, while those for
medium and large rms are not. Thus, business regulations do not appear to
affect the growth of larger rms. On the other hand, the positive effect for
micro rms suggests that micro rms benet from a generally lower enforce-
ment stand for micro rms, which may help divert some economic activity
to micro rms. Results are quite similar across all specications regardless of
how the IC variables are computed, suggesting that endogeneity may be a
less important concern in this case.
C. Corruption
Moving on to results for corruption, we nd that, as in the case of access to
nance, results vary substantially whether and how we account for endogeneity.
Results based on individual responses (table 4, col. 2) or cell averages matched
to current size (col. 4) would suggest that corruption helps micro rms, while
having a much smaller, and in some cases negative, effect on the growth of
larger rms. Measuring the incidence of corruption based on country-industry-
location size averages matched to initial size or on country-industry locations,
in our opinion, two better ways of accounting for endogeneity, yields results
that go in the same direction but are not statistically signicant. This suggests
that while corruption may have differential effects across rm size, and possibly
adverse effects on medium-sized rms, the presence of endogeneity creates
biases that increase the probability of nding this result.
D. Infrastructure
With regard to infrastructure bottlenecks, measured as the incidence of power
outages, results are again quite sensitive to whether we control or not for
endogeneity. Using our baseline specication (col. 1), we nd evidence that
infrastructure bottlenecks tend to stunt the growth of medium and large rms.
Instead, the coefcient for micro rms is positive, indicating that again, micro
rms may benet from the troubles of larger, possibly more productive rms,
as some production may be diverted to micro rms. The coefcient on micro
rms, however, is not statistically signicant. But the difference between micro
and small/medium rms is larger, taking into account that small and medium
rms report more frequent outages than micro rms. This is consistent with
micro rms using less capital- and energy-intensive modes of production and
beneting that these processes can be sustained due to hardships hitting the
small and medium enterprises (SMEs).
Results appear stronger when individual rms responses are used to com-
pute IC, but the coefcients are of similar magnitudes as those reported in
table 4, column 1. This suggests that endogeneity does not appear as a large
concern and that results in column 3 are estimated with more precision as IC
varies at the rm level. Computing IC out of the average within country-
location-sector cells also yields similar but stronger results than in column 1.
All in all, the results suggest that poor infrastructure has a detrimental effect
on the growth of all rms but the micro ones.
VII. Robustness
We now examine whether our results are robust to changes in the denition
of variables, set of controls, and sample of countries.
A. Addressing Other Possible Sources of Endogeneity
Another source of potential endogeneity is whether better-performing rms
choose locations based on the quality of the business environment. To see if
this could be signicant, column 1 of table 5 reports the results if we restrict
the sample to domestic-owned, micro, and SMEs who are least likely to be
footloose and where the location tends to be associated with the places of birth
or residence of the business owner. Foreign rms and large domestic rms are
the most likely to be selective in their choice of location. The results are very
robust.
B. Alternative Denitions of Investment Climate Indicators
The second set of robustness checks involves using alternative measures of the
different investment climate areas. The surveys generally provide various mea-
sures to choose from. Of particular interest are measures on access to nance.
There are two sets of results, that the nature of the nance being captured
matters and that the results are somewhat sensitive to whether variation across
sectors within a country is included or not.
To examine the different types of nance, two additional measures are
includedthe share of sales made on credit and access to working capital
from external institutions. The measures used so far capture a type of nance
that is relatively more difcult to access, namely, nancing for longer term
investment projects from formal nancial institutions. The other two alter-
natives presented in table 5 measure access to a more informal formof nancing
and to shorter term nancing from institutions. In the case of the share of
sales on credit, the results show that micro rms do not share in the benets
of access to this nancing, consistent with such credit being offered less by
micro rms. Small and medium-sized rms do benet. The impact on large
rms is not signicant, which again is not so surprising, as this type of nance
is relatively less important for them. For access to working capital (col. 3) the
benets are more widespread. If anything, this type of nancing is most
benecial for the micro rms, although the differences across sizes are not
signicant.
The second point is that these results are sensitive to whether one allows
for sectoral variation within countries to be included. Our results so far have
included the full set of interacted sector-survey dummies. However, here (table
5, cols. 24) the results are more signicant when sector and country dummies
enter separately. There are two interpretations that can be given to this nding.
First, the original rationale to include the broader set of dummies was in part
to control for demand shocks that could vary by country and sector. Failing
to control for sector-country shocks, endogeneity could be reentering the re-
gressions. Such shocks could affect rm performance and increase their demand
for nancing. This would imply caution in interpreting the results. However,
there is a second interpretation. Work by Rajan and Zingales (1998) points
to potentially important differences across sectors in the impact of improved
access to nance for growth. Beck et al. (2008) show that sectors with a larger
proportion of small rms stand to benet disproportionately from nancial
634
TABLE 5
ROBUSTNESS
Control for Possi-
ble Location Se-
lection Bias Alternative Measures of IC
Investment Climate Variable
(ICvar)
Sample Restricted
SMEs Domestic
(1)
Finance p Sh-
sales-credit
(2)
Finance p sh-
wkcap
(3)
Four Alterna-
tive IC
Variables
a
(4)
Finance .046 .039 .194** .272***
(.061) (.033) (.083) (.089)
(Small) # Finance .093 .092*** .108 .245***
(.065) (.033) (.072) (.075)
(Medium) # Finance .168** .066* .126 .237***
(.079) (.035) (.085) (.091)
(Large) # Finance .034 .111 .178*
(.037) (.084) (.093)
Regulations .003** .004*** .004*** .012
(.001) (.001) (.002) (.012)
(Small) # Regulations .005*** .006*** .005*** .029***
(.001) (.001) (.001) (.009)
(Medium) # Regulations .001 .004*** .003* .023*
(.002) (.001) (.002) (.013)
(Large) # Regulations .005*** .003* .017
(.001) (.002) (.014)
Corruption .047 .007 .033 .004
(.042) (.024) (.034) (.006)
(Small) # Corruption .013 .043* .007 .002
(.046) (.025) (.033) (.005)
(Medium) # Corruption .055 .001 .070 .003
(.057) (.029) (.044) (.007)
(Large) # Corruption .021 .043 .009
(.033) (.043) (.008)
Infrastructure .006 .008 .008 .000
(.009) (.005) (.008) (.004)
(Small) # Infrastructure .003 .000 .000 .000
(.005) (.004) (.005) (.004)
(Medium) # Infrastructure .012 .008 .006 .002
(.008) (.005) (.007) (.004)
(Large) # Infrastructure .007 .009 .003
(.006) (.007) (.005)
Firm controls Yes Yes Yes Yes
Survey # Sector xed effects Yes No No No
Survey and Sector xed effects No Yes Yes Yes
Observations 46,122 54,442 54,013 48,411
R
2
.15 .115 .122 .122
Note. Firm controls: dummies for rm size and age (t 3), foreign or government owned, exporter
status, if in small city, empg base years, average employment growth in cell, and a constant term. All
ICvar use cell average, matched to initial size; cells based on survey-sector-location-avg-size (excluding
own). Col. 1, nance pSh-invest-n; regulations pSh-mng-time; corruption pBribe y-n; infrastructure
p Days-no power. Results in cols. 2 and 3 use the same variables as col. 1 with exception of Finance
variable. Standard errors are clustered on survey-location-sector-size.
a
Column 4 IC alternative: nance psh-wkcap; regulations pdays-inspections; corruption pbribe (%);
infrastructure p loss-transit.
* Signicant at 10%.
** Signicant at 5%.
development. Including interactions sector-survey dummies then shuts down
this channel, explaining why these two variables, share of sales sold on credit
and the share of working capital nanced externally, lose their signicance
when all the interactions are included in the regression.
Column 4 of table 5 also includes alternative measures for the other three
investment climate measures. They are days of inspections, the size of bribes
paid, and losses incurred during transportation delays. The measure of in-
spections has the same effect as that of management time spent dealing with
ofcials, deterring small and medium-sized rms from growing. However, the
new measures of corruption do not offer new insights, and the use of the
transportation-related variable provides weaker results. One conclusion is that
power, not transportation (at least when measured as percentage of sales lost
in transportation), is the more important dimension of infrastructure affecting
rm growth.
C. Variations in Denition of Size Categories
Table 6 investigates whether results are robust to variations in size classi-
cations. For comparison purposes, column 1 reproduces the results of table 4,
column 1. Column 2 of table 6 reports the results of dening micro rms as
rms with 510 employees and excluding from the estimation rms with
fewer than 5 employees. The other size categories remain unchanged. The
results on access to nance weaken for medium-sized rms, suggesting that
the differential effects of nance between micro and medium-sized rms are
in part explained by the more negative effect of nance on the smallest rms.
This lends additional support to the selection story proposed by Jeong and
Townsend (2007), as the very small rms are the ones that are likely to be
less productive and, therefore, are more likely to be hurt by increasing costs
of inputs driven by a higher supply of credit.
In column 3 of table 6 we follow the size category cutoffs used by Beck
et al. (2005), which implies combining most of our micro with all our small
into one category and combining our medium with many of our large into
the medium category and leaving only the very largest in the large category.
With these cutoffs, and controlling for endogeneity, we do not replicate their
nding that the smallest rms benet the most (although their result is
reproduced if we match conditions based on current rather than initial size).
The results also show that it is not the very largest rms that benet from
nance relative to the small but, rather, those in the mid-sized (50500
employees) range.
The results on regulations show that the positive relationship is driven by
the smallest rms; excluding those with fewer than ve employees leaves the
636
TABLE 6
OTHER SIZE DEFINITIONS
Baseline (Table 4,
Col. 1),
Micro (!10),
Small (1050),
Medium (51200),
Large (200)
Exclude ! 5,
Micro (510),
Rest of Sizes
as in Col. 1
(1)
Small (550),
Medium (51
500),
Large (500)
(2)
Excluding
Micro (!10)
(3)
Sh-invest-n .055 .078 .047 .032
(.059) (.069) (.048) (.046)
(Small, t 3) # Sh-invest-n .092 .084
(.065) (.070)
(Medium) # Sh-invest-n .130* .132 .114** .057
(.075) (.085) (.053) (.052)
(Large) # Sh-invest-n .136* .149* .062 .070
(.071) (.081) (.073) (.059)
Sh-mng-time .003** .002 .001 .002
(.001) (.002) (.001) (.001)
(Small) # Sh-mng-time .005*** .004***
(.001) (.002)
(Medium, t 3) # Sh-mng-time .003 .001 .002 .003**
(.002) (.002) (.001) (.001)
(Large) # Sh-mng-time .003* .002 .000 .002
(.002) (.002) (.002) (.002)
Bribe y-n .051 .014 .007** .008**
(.041) (.044) (.004) (.004)
(Small) # Bribe y-n .004 .037
(.045) (.043)
(Medium) # Bribe y-n .073 .026 .004 .005
(.054) (.055) (.005) (.005)
(Large) # Bribe y-n .056 .005 .019* .012
(.051) (.053) (.011) (.008)
Days-no power .008 .002 .004 .001
(.008) (.009) (.009) (.008)
(Small) # Days-no power .006 .002
(.005) (.006)
(Medium) # Days-no power .014* .010 .007 .007
(.008) (.008) (.006) (.006)
(Large) # Days-no power .015* .010 .009 .002
(.008) (.009) (.010) (.008)
Firm controls Yes Yes Yes Yes
Survey # Sector xed effects Yes Yes Yes Yes
Observations 56,880 47,491 47,222 34,474
R
2
.15 .13 .12 .12
Note. Firm controls: dummies for rm size and age (t 3), foreign or government owned, exporter
status, if in small city, empg base years, avg. employment growth in cell, and a constant term. All ICvar
use cell avg, matched to initial size; cells based on survey-sector-location-avg-size (excluding own).
Standard errors are clustered on survey-location-sector-size.
* Signicant at 10%.
** Signicant at 5%.
effect insignicant for the remaining micro rms. Still, the effect remains
negative for the small rms. In column 3, when all the micro are excluded,
the omitted category is now small and the effect is negative but not signicant.
Medium and large rms still have a more positive response to regulations
than the small. However, combining the rms into the larger size categories
misses the entire story, with no results appearing signicant.
For corruption, however, the story does become stronger once the micro
rms are dropped (table 6, col. 3) or absorbed with small rms (col. 2). Now
there is a positive and signicant effect for the smaller rms that is offset for
the larger rms, signicantly so for the very large rms. Combined with the
regulation results, small rms above the 10-employee thresholdthat is, those
that are large enough to nd evasion from regulations more difcultnd
measures of regulation more detrimental to their growth, but that corruption
offers a way of mitigating these effects. For medium rms, regulations are not
as detrimental as for small rms (in fact, they have a small net positive effect)
and corruption is correspondingly less benecial for medium-sized rms.
D. Age versus Size
It is well known that size is strongly correlated with age, as rms tend to
start operations small and grow over time. We assess whether differences in
the impact of business environment variables could be related to age rather
than size by adding a full set of interactions of age dummies with IC variables
to our baseline results (table 4, col. 1). Results for size # IC interactions
remain very similar; however, some interesting results emerge from the size
interactions (see table 7). We nd that regulations tend to affect the growth
of older, more established rms, relative to younger ones. Combined with the
size interactions, these results provide some interesting qualications to our
baseline results. They suggest that the adverse effect of regulations on the
growth of small rms is more visible among rms that are small but have
been in business for a while, rather than for rms that are small and young.
Similarly, these results allow us to qualify the statement that micro rms
benet from a poor regulatory environment. In fact, the rms that have been
around for a while are the ones that benet from regulations, while micro-
young ones do not. It may well be that a poor regulatory environment increases
the survival of low productivity rms (i.e., rms that do not have the entre-
preneurial ability to grow) because it penalizes more severely higher produc-
tivity rms.
Another interesting result of the size interactions is that corruption seems
to disproportionately benet older, more established rms. Perhaps age is
correlated with the access to networks and inuence channels. It may also be
638
TABLE 7
IS IT SIZE OR AGE THAT MATTERS?
Investment Climate Variable (ICvar) Continuation
Sh-invest-n .053
(.064)
(Small, t 3) # Sh-invest-n .093
(.065)
(Medium) # Sh-invest-n .127* (Mature, t 3) # Sh-invest-n .004
(.076) (.037)
(Large) # Sh-invest-n .132* (Older) # Sh-invest-n .003
(.072) (.041)
Sh-mng-time .005***
(.001)
(Small) # Sh-mng-time .005***
(.001)
(Medium. t 3) # Sh-mng-time .002 (Mature) # Sh-mng-time .002***
(.002) (.001)
(Large) # Sh-mng-time .002 (Older) # Sh-mng-time .004***
(.002) (.001)
Bribe y-n .029
(.042)
(Small) # Bribe y-n .011
(.046)
(Medium) # Bribe y-n .087 (Mature) # Bribe y-n .037
(.056) (.024)
(Large) # Bribe y-n .071 (Older) # Bribe y-n .058**
(.053) (.029)
Days-no power .006
(.009)
(Small) # Days-no power .006 (Mature) # Days-no power .003
(.005) (.003)
(Medium) # Days-no power .015* (Older) # Days-no power .003
(.008) (.004)
(Large) # Days-no power .015* Mature .086***
(.008) (.013)
(Small) .110*** Older .137***
(.029) (.016)
(Medium) .148*** Firm controls Yes
(.032) Survey # Sector xed effects Yes
(Large) .209*** Observations 56,880
(.032) R
2
.15
Note. Firm controls: dummies for whether rm is foreign or government owned, exporter status, if
in small city, empg base years, average employment growth in cell, and a constant term. Size and
age are t 3. All ICvar are cell averages, matched to initial size; cells based on survey-sector-loc.-
avg-size (excluding own). This is a single regression, reported in two columns due to the many
variables included. Standard errors are clustered on survey-location-sector-size.
* Signicant at 10%.
** Signicant at 5%.
true that a corrupt environment makes the survival of certain types of rms
more likely, which then are observed as older at the time of the survey.
Addressing this form of sample selection is beyond the scope of this study.
E. Different Samples
The results so far have pooled countries together (with sector-survey xed
effects), but it could be that the effects vary by country groupings, along
dimensions that can reinforce or weaken the broader business environment.
To test whether the institutional development of countries matters, we divide
the sample according to the depth of credit markets (table 8, cols. 12), the
degree by which countries control corruption (cols. 34), and the strength of
their rule of law (cols. 56).
20
Quite interestingly, results for increased access
to nance turn out to be much stronger in countries with a more developed
nancial system, suggesting again that a marginal development in a countrys
nancial system tends to favor small, medium, and, to a lower extent, large
size rms, rather than micro rms. This also suggests that nance operates
better in economies with more developed nancial markets. This result is
further conrmed by the results in columns 34, which show that the effect
of increased access to nance is stronger in economies with higher control of
corruption and better rule of law, both hallmarks of better functioning econ-
omies.
Business regulations, measured as the percentage of management time spent
dealing with regulations, have a negative effect on the growth of small rms
throughout most samples of countries, suggesting that the conclusion that
stringent regulations may lead many small rms to grow less in order to hide
from the authorities is quite robust to the inclusion or exclusion of countries.
What is striking is that results are stronger in countries with either lower
control of corruption or a weaker rule of law. It may be that the quality of
the regulations in lower institutional environments is more burdensome, re-
sulting in the greater constraints on small rms. It is also possible that more
onerous regulations in a weak institutional setting open the way to corruption
or harassment on the part of ofcials, encouraging rms to remain small and
below the radar screen where possible.
The direct effects of corruption on the growth of rms tend to be more
signicant in countries with higher control of corruption. Corruption, in a
setting where corruption is not controlled, does not have much of an effect.
20
We also investigated whether there were signicant patterns by income or GDP per capital. As
the differences were less pronounced than those that capture the institutional environment, we do
not report them. We do note, however, that this implies that it is indeed the broader institutional
environment that matters rather than income itself.
640
T
A
B
L
E
8
I
N
S
T
I
T
U
T
I
O
N
A
L
E
F
F
E
C
T
S
(
I
C
O
N
E
M
P
L
O
Y
M
E
N
T
G
R
O
W
T
H
)
D
e
p
e
n
d
e
n
t
V
a
r
i
a
b
l
e
:
E
m
p
l
o
y
m
e
n
t
G
r
o
w
t
h
(
E
m
p
(
t
)
E
m
p
(
t
3
)
)
/
(
(
E
m
p
(
t
)
E
m
p
(
t
3
)
)
/
2
)
D
o
m
e
s
t
i
c
C
r
e
d
i
t
t
o
P
r
i
v
a
t
e
S
e
c
t
o
r
C
o
n
t
r
o
l
C
o
r
r
u
p
t
i
o
n
R
u
l
e
o
f
L
a
w
M
e
d
i
a
n
(
1
)
1
M
e
d
i
a
n
(
2
)
0
(
3
)
1
0
(
4
)
0
(
5
)
1
0
(
6
)
S
h
-
i
n
v
e
s
t
-
n
.
0
7
0
.
1
1
2
.
0
2
0
.
1
1
0
.
0
3
5
.
1
5
8
*
(
.
0
9
4
)
(
.
0
7
6
)
(
.
0
7
9
)
(
.
0
8
1
)
(
.
0
8
0
)
(
.
0
8
5
)
(
S
m
a
l
l
)
#
S
h
-
i
n
v
e
s
t
-
.
0
9
7
.
2
0
2
*
*
.
0
1
5
.
1
8
5
*
.
0
1
7
.
2
2
2
*
*
(
.
0
9
3
)
(
.
0
8
4
)
(
.
0
8
1
)
(
.
0
9
9
)
(
.
0
7
9
)
(
.
0
9
6
)
(
M
e
d
i
u
m
)
#
S
h
-
i
n
v
e
s
t
-
.
0
1
9
.
2
0
3
*
.
0
5
1
.
2
0
4
*
.
0
1
6
.
3
1
1
*
*
(
.
1
1
3
)
(
.
1
1
6
)
(
.
0
9
6
)
(
.
1
1
3
)
(
.
0
9
6
)
(
.
1
2
7
)
(
L
a
r
g
e
)
#
S
h
-
i
n
v
e
s
t
-
.
0
2
6
.
1
4
5
.
0
2
6
.
0
9
6
.
0
0
5
.
1
9
7
*
(
.
1
1
5
)
(
.
0
9
7
)
(
.
0
9
3
)
(
.
1
2
0
)
(
.
0
9
7
)
(
.
1
1
2
)
S
h
-
m
n
g
-
t
i
m
e
.
0
0
1
.
0
0
6
*
*
.
0
0
2
.
0
0
2
.
0
0
2
.
0
0
2
(
.
0
0
2
)
(
.
0
0
3
)
(
.
0
0
2
)
(
.
0
0
3
)
(
.
0
0
2
)
(
.
0
0
3
)
(
S
m
a
l
l
)
#
S
h
-
m
n
g
-
t
i
m
e
.
0
0
2
.
0
0
9
*
*
*
.
0
0
3
*
*
.
0
0
5
.
0
0
4
*
*
.
0
0
4
(
.
0
0
2
)
(
.
0
0
3
)
(
.
0
0
2
)
(
.
0
0
3
)
(
.
0
0
2
)
(
.
0
0
2
)
(
M
e
d
i
u
m
)
#
S
h
-
m
n
g
-
t
i
m
e
.
0
0
0
.
0
0
5
.
0
0
0
.
0
0
2
.
0
0
0
.
0
0
1
(
.
0
0
2
)
(
.
0
0
4
)
(
.
0
0
2
)
(
.
0
0
3
)
(
.
0
0
2
)
(
.
0
0
3
)
(
L
a
r
g
e
)
#
S
h
-
m
n
g
-
t
i
m
e
.
0
0
1
.
0
1
0
*
*
*
.
0
0
1
.
0
0
6
*
.
0
0
1
.
0
0
4
(
.
0
0
2
)
(
.
0
0
3
)
(
.
0
0
2
)
(
.
0
0
3
)
(
.
0
0
2
)
(
.
0
0
3
)
B
r
i
b
e
y
-
n
.
0
4
9
.
0
4
3
.
0
1
6
.
1
3
6
*
*
.
0
0
8
.
0
0
4
(
.
0
4
8
)
(
.
0
6
4
)
(
.
0
5
1
)
(
.
0
6
5
)
(
.
0
0
6
)
(
.
0
1
3
)
641
(
S
m
a
l
l
)
#
B
r
i
b
e
y
-
n
.
0
2
1
.
0
2
9
.
0
4
3
.
1
1
8
*
.
0
0
3
.
0
0
6
(
.
0
4
5
)
(
.
0
7
0
)
(
.
0
5
5
)
(
.
0
6
3
)
(
.
0
0
6
)
(
.
0
1
4
)
(
M
e
d
i
u
m
)
#
B
r
i
b
e
y
-
n
.
0
5
9
.
0
7
2
.
0
2
6
.
2
5
2
*
*
*
.
0
0
7
.
0
0
5
(
.
0
6
3
)
(
.
0
8
7
)
(
.
0
6
9
)
(
.
0
8
6
)
(
.
0
0
8
)
(
.
0
1
5
)
(
L
a
r
g
e
)
#
B
r
i
b
e
y
-
n
.
0
3
0
.
0
4
5
.
0
1
9
.
1
7
8
*
*
.
0
1
0
.
0
0
8
(
.
0
6
3
)
(
.
0
7
6
)
(
.
0
6
4
)
(
.
0
7
8
)
(
.
0
1
0
)
(
.
0
1
7
)
D
a
y
s
-
n
o
p
o
w
e
r
.
0
0
3
.
0
1
2
.
0
0
0
.
0
2
4
.
0
0
0
.
0
1
8
(
.
0
1
0
)
(
.
0
1
5
)
(
.
0
1
0
)
(
.
0
2
0
)
(
.
0
1
0
)
(
.
0
1
7
)
(
S
m
a
l
l
)
#
D
a
y
s
-
n
o
p
o
w
e
r
.
0
0
7
.
0
0
4
.
0
0
0
.
0
1
1
.
0
0
1
.
0
1
1
(
.
0
0
6
)
(
.
0
1
2
)
(
.
0
0
7
)
(
.
0
1
8
)
(
.
0
0
7
)
(
.
0
1
2
)
(
M
e
d
i
u
m
)
#
D
a
y
s
-
n
o
p
o
w
e
r
.
0
1
5
*
.
0
0
2
.
0
0
5
.
0
4
3
*
.
0
0
5
.
0
3
0
*
(
.
0
0
9
)
(
.
0
1
5
)
(
.
0
0
9
)
(
.
0
2
2
)
(
.
0
0
9
)
(
.
0
1
7
)
(
L
a
r
g
e
)
#
D
a
y
s
-
n
o
p
o
w
e
r
.
0
0
7
.
0
0
5
.
0
0
1
.
0
2
5
.
0
0
9
.
0
2
4
(
.
0
1
0
)
(
.
0
1
4
)
(
.
0
1
0
)
(
.
0
2
4
)
(
.
0
1
1
)
(
.
0
1
7
)
F
i
r
m
c
o
n
t
r
o
l
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
S
u
r
v
e
y
#
S
e
c
t
o
r
x
e
d
e
f
f
e
c
t
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
O
b
s
e
r
v
a
t
i
o
n
s
3
1
,
8
5
4
2
3
,
1
5
7
4
1
,
5
7
5
1
5
,
2
0
5
3
3
,
1
0
4
2
3
,
6
6
7
R
2
.
1
6
.
1
1
.
1
6
.
1
2
.
1
7
.
1
2
N
o
t
e
.
F
i
r
m
c
o
n
t
r
o
l
s
:
d
u
m
m
i
e
s
f
o
r
r
m
s
i
z
e
a
n
d
a
g
e
(
t
3
)
,
f
o
r
e
i
g
n
o
r
g
o
v
e
r
n
m
e
n
t
o
w
n
e
d
,
e
x
p
o
r
t
e
r
s
t
a
t
u
s
,
i
f
s
m
a
l
l
c
i
t
y
,
e
m
p
g
b
a
s
e
y
e
a
r
s
,
a
v
g
.
e
m
p
l
o
y
m
e
n
t
g
r
o
w
t
h
i
n
c
e
l
l
,
a
n
d
a
c
o
n
s
t
a
n
t
t
e
r
m
.
A
l
l
I
C
v
a
r
a
r
e
c
e
l
l
a
v
g
,
m
a
t
c
h
e
d
t
o
i
n
i
t
i
a
l
s
i
z
e
;
c
e
l
l
s
b
a
s
e
d
o
n
s
u
r
v
e
y
-
s
e
c
t
o
r
-
l
o
c
.
-
a
v
g
-
s
i
z
e
(
e
x
c
l
u
d
i
n
g
o
w
n
)
.
S
t
a
n
d
a
r
d
e
r
r
o
r
s
a
r
e
c
l
u
s
t
e
r
e
d
o
n
s
u
r
v
e
y
-
l
o
c
a
t
i
o
n
-
s
e
c
t
o
r
-
s
i
z
e
.
*
S
i
g
n
i
c
a
n
t
a
t
1
0
%
.
*
*
S
i
g
n
i
c
a
n
t
a
t
5
%
.
*
*
*
S
i
g
n
i
c
a
n
t
a
t
1
%
.
However, if, in a stronger institutional environment, there is still corruption,
it provides incentives to keep operations small. The effects of corruption on
larger rms are signicantly negative in countries with better control of cor-
ruption. It thus appears that the distortions generated by corruption are more
important and distort rm growth more in countries where markets and
institutions are more developed and a higher share of resources is allocated
via markets. Finally, infrastructure bottlenecks appear as a constraint to the
growth of small, medium, and/or large rms in most samples of countries
although the levels of signicance vary across samples, with the stronger effects
for medium-sized rms.
All in all, the results above suggest that improving on some domains of
the investment climate constraints may be more important to the extent that
other aspects of the business climate and institutional development improve.
Thus, nancial market development brings higher development to medium
and large rms in regions where there is higher control of corruption and a
better rule of law. Similarly, improving business regulations may be more
important where institutional development otherwise lags. Cracking down on
corruption is still an issue even in more developed institutional countries,
with particularly deleterious effects on medium and large rms.
VIII. Conclusions
This study has provided new evidence of the role of the investment climate
in employment growth. The results indicate signicant differences across size
categories of rmsboth in terms of differences in objective conditions faced
by rms and in terms of nonlinearities in the impact of these conditions. The
study also devotes substantial effort to surmount endogeneity concerns between
investment climate measures and business performance and nds that ac-
counting for endogeneity reduces the differential effects of business climate
on the performance of different types of rms. Yet the results indicate that a
poor business climate tends to hurt the growth of small, medium, and large
rms and to benet in relativebut sometimes also in absolutecases, micro
rms.
It also nds that increased access to nance, particularly external nance to
fund investment, has a larger positive effect in the growth of small, medium,
and large rms, particularly in economies with more developed nancial sys-
tems or a better rule of law. Thus, unlike previous research, this study does
not nd that the impact on employment growth of an extra unit of external
nance is highest for micro rms. We interpret these results as evidence that
micro rms are not necessarily more credit constrained. Instead, many micro
entrepreneurs may have little willingness or capabilities to grow their rms.
As access to nance increases, micro rms may endure input price increases
and production displacement caused by the growth of large, more productive
rms that attain higher benets of increased access to nance. The results also
reinforce the importance of differentiating the impact across size classes of
rms and allowing micro rms (fewer than 10 employees) to be different from
small rms.
Another important nding of this study is that business regulations mea-
sured as a higher share of management time spent dealing with regulations
or inspections may stunt the growth of small rms. Thus, as management
time dealing with authorities as well as inspections increases substantially with
rm size, the marginal increase in enforcement can act as a strong disincentive
to grow a rm.
What are the aggregate implications of these ndings for the size, efciency,
and dynamism of the business sector in developing countries? The results
found in this study suggest that a weak business environment displaces activity
from large, medium, and small rms to the benet of micro ones. To the
extent that larger rms are more productiveas suggested by many theoretical
models of rm dynamicsthis would imply substantial resource reallocation
from more productive to less productive ones, potentially leading to important
losses in aggregate productivity. In addition, given that enforcement of taxes
and regulations tends to be smaller in micro rms, a shift of economic activity
toward these rms is likely to imply that a larger share of rms remain informal
or semi-informal, reducing the capacity of the state to collect taxes and pay
for fundamental inputs for development such as education. Finally, if there
are substantial xed costs in activities like innovation or worker training, a
higher share of activity in micro rms may imply more rms and workers
locked in activities with limited innovation and growth opportunities.
644
Appendix
TABLE A1
DATA SET
Country No. Obs. % Country No. Obs. % Country No. Obs. %
Albania2002 166 .29 Ghana2007 249 .44 Panama2006 550 .97
Albania2005 202 .35 Greece2005 539 .95 Paraguay2006 561 .99
Angola2006 274 .48 Guatemala2003 451 .79 Peru2006 603 1.06
Argentina2006 950 1.67 Guatemala2006 494 .87 Philippines2003 614 1.08
Armenia2002 170 .3 Guinea2006 181 .32 Poland2002 495 .87
Armenia2005 351 .62 GuineaBissau2006 129 .23 Poland2003 105 .18
Azerbaijan2002 162 .28 Guyana2004 152 .27 Poland2005 969 1.7
Azerbaijan2005 348 .61 Honduras2003 436 .77 Portugal2005 501 .88
Bangladesh2002 964 1.69 Honduras2006 419 .74 Romania2002 254 .45
Belarus2002 250 .44 Hungary2002 245 .43 Romania2005 586 1.03
Belarus2005 322 .57 Hungary2005 602 1.06 Russia2002 488 .86
Belarus2008 232 .41 India2006 2,919 5.13 Russia2005 592 1.04
Benin2004 182 .32 Indonesia2003 707 1.24 Rwanda2006 154 .27
BiH2002 171 .3 Ireland2005 499 .88 Senegal2007 473 .83
BiH2005 194 .34 Jamaica2005 70 .12 Serbia2003 399 .7
Bolivia2006 556 .98 Jordan2006 451 .79 Slovakia2002 163 .29
Botswana2006 253 .44 Kazakhstan2002 249 .44 Slovakia2005 208 .37
Bulgaria2002 245 .43 Kazakhstan2005 578 1.02 Slovenia2002 188 .33
Bulgaria2005 291 .51 Kenya2003 211 .37 Slovenia2005 221 .39
BurkinaFaso2006 131 .23 Kyrgyzstan2002 166 .29 SouthAfrica2003 551 .97
Burundi2006 215 .38 Kyrgyzstan2003 102 .18 SouthKorea2005 594 1.04
Cambodia2003 464 .82 Kyrgyzstan2005 199 .35 Spain2005 601 1.06
Cameroon2006 165 .29 Laos2005 225 .4 Swaziland2006 208 .37
CapeVerde2006 93 .16 Latvia2002 170 .3 Tajikistan2002 170 .3
Chile2004 942 1.66 Latvia2005 200 .35 Tajikistan2003 107 .19
Chile2006 962 1.69 Lebanon2006 345 .61 Tajikistan2005 199 .35
Colombia2006 936 1.64 Lesotho2003 44 .08 Tajikistan2008 316 .56
CostaRica2005 333 .59 Lithuania2002 193 .34 Tanzania2003 245 .43
Croatia2002 174 .31 Lithuania2004 230 .4 Tanzania2006 360 .63
Croatia2005 229 .4 Lithuania2005 196 .34 Turkey-b2005 1,238 2.18
Czech2002 261 .46 Madagascar2005 269 .47 Turkey2002 513 .9
Czech2005 327 .57 Malawi2005 150 .26 Turkey2005 544 .96
DRC2006 271 .48 Mali2003 123 .22 Turkey2008 900 1.58
Ecuador2003 411 .72 Mauritania2006 204 .36 Uganda2003 291 .51
Ecuador2006 592 1.04 Mauritius2005 177 .31 Uganda2006 509 .89
ElSalvador2003 459 .81 Mexico2006 1,343 2.36 Ukraine2002 463 .81
ElSalvador2006 620 1.09 Moldova2002 173 .3 Ukraine2005 586 1.03
Estonia2002 160 .28 Moldova2003 103 .18 Ukraine2008 733 1.29
Estonia2005 214 .38 Moldova2005 346 .61 Uruguay2006 540 .95
FYROM2002 165 .29 Montenegro2003 100 .18 Uzbekistan2002 255 .45
FYROM2005 194 .34 Mozambique2007 429 .75 Uzbekistan2003 100 .18
Gambia2006 126 .22 Namibia2006 242 .43 Uzbekistan2005 298 .52
Georgia2002 173 .3 Nicaragua2003 448 .79 Uzbekistan2008 361 .63
Georgia2005 199 .35 Nicaragua2006 461 .81 Vietnam-b2005 497 .87
Georgia2008 325 .57 Nigeria2007 2,005 3.52 Vietnam2005 1,109 1.95
Germany2005 1,195 2.1 Pakistan2002 937 1.65 Zambia2002 187 .33
TABLE A2
SAMPLE COMPOSITION
Percent Percent
Region: Sector:
Sub-Saharan Africa 16.01 Textiles 5.3
East Asia & Pacic 14.45 Leather 1.15
East Europe & Central Asia 27.21 Garments 8.46
Western Europe 4.57 Agroindustry 6.65
Latin America 20.87 Beverages 7.17
Middle East & North Africa 6.72 Metals and machinery 9.42
South Asia 10.16 Electronics 1.55
Chemicals and pharmaceuticals 4.9
Income level: Construction 5.07
Low 31.43 Wood and furniture 5.02
Lower middle 44.59 Nonmetallic and plastic materials 4.37
Upper middle 16.36 Paper 1.98
High 7.62 IT services 3.85
Other manufacturing 1.39
Initial size: Advertising and marketing 1.73
Micro (110 employees) 39.36 Other services 7.28
Small (1150 employees) 33.64 Retail and wholesale trade 17.67
Medium (51200 employees) 16.39 Hotels and restaurants 3.41
Large (200 employees) 10.61 Transport 2.52
Mining and quarrying .46
Initial age: Other transport equipment .67
Young 15 30.71
Mature 615 38.34 Labor intensive 31.69
Older 16 30.95 Capital intensive 31.89
Services 36.42
Nonexporter 77.35
Exporter 22.65 Capital or city 1 million 35.27
City 500,0001 million 13.22
Domestic owned 88.82 City 250,000500,000 15.43
Foreign owned 11.18 City 50,000250,000 17.86
City ! 50,000 18.22
References
Ahsan, Ahmad, and Carmen Pages. 2009. Are All Labor Regulations Equal? Evi-
dence from Indian Manufacturing. Journal of Comparative Economics 37, no. 1:62
75.
Almeida, Rita, and Pedro Carneiro. 2009. Enforcement of Labor Regulation and
Firm Size. Journal of Comparative Economics 37 (March): 2846.
Autor, David H., William R. Kerr, and Adriana D. Kugler. 2007. Do Employment
Protections Reduce Productivity? Evidence from U.S. States. Economic Journal 117
( June): 189217.
Ayyagari, Meghana, Asli Demirgu c-Kunt, and Vojislav Maksimovic. 2008. How
Important Are Financing Constraints? The Role of Finance in the Business En-
vironment. World Bank Economic Review 22, no. 3:483516.
Batra, Geeta, Daniel Kaufmann, and Andrew H. W. Stone. 2003. Investment Climate
around the World. Washington, DC: World Bank.
Beck, Thorsten, Asli Demirgu c-Kunt, Luc Laeven, and Ross Levine. 2008. Finance,
Firm Size and Growth. Journal of Money, Credit and Banking 40, no. 7:1379
1405.
Beck, Thorsten, Asli Demirgu c-Kunt, and Vojislav Maksimovic. 2005. Financial
and Legal Constraints to Growth: Does Firm Size Matter? Journal of Finance 60,
no. 1:13177.
Besley, Timothy, and Robin Burgess. 2004. Can Labor Regulation Hinder Economic
Performance? Evidence from India. Quarterly Journal of Economics 119, no. 1:91
134.
Bigsten, Ane, and Mans Soderbom. 2006. What Have We Learned from a Decade
of Manufacturing Enterprise Surveys in Africa? World Bank Research Observer 21,
no. 2:24165.
Botero, Juan, Simeon Djankov, Rafael La Porta, Florencio Lopez de Silanes, and
Andrei Shleifer. 2004. The Regulation of Labor. Quarterly Journal of Economics
119, no. 4:133982.
Clark, Ximena, David Dollar, and Alejandro Micco. 2004. Port Efciency, Maritime
Transport Costs, and Bilateral Trade. Journal of Development Economics 75, no. 2:
41750.
Davis, Steven J., and John Haltiwanger. 1992. Gross Job Creation, Gross Job
Destruction, and Employment Reallocation. Quarterly Journal of Economics 107,
no. 3:81963.
. 1999. On the Driving Forces behind Cyclical Movements in Employment
and Job Reallocation. American Economic Review 89, no. 5:123458.
Demirgu c-Kunt, Asli, and Vojislav Maksimovic. 1998. Law, Finance and Firm
Growth. Journal of Finance 53:210737.
Djankov, S., R. La Porta, F. Lopez De Silanes, and A. Shleifer. 2002. The Regulation
of Entry. Quarterly Journal of Economics 117, no. 1:137.
Dollar, David, Mary Hallward-Driemeier, and Taye Mengistae. 2005. Investment
Climate and Firm Performance in Developing Countries. Economic Development
and Cultural Change 54, no. 1:131.
Galindo, Arturo Jose, and Alejandro Micco. 2007. Creditor Protection and Credit
Response to Shocks. World Bank Economic Review 21, no. 3:41338.
Fisman, Raymond, and Jakob Svensson. 2007. Are Corruption and Taxation Really
Harmful to Growth? Firm Level Evidence. Journal of Development Economics 83,
no. 1:6375.
Hallward-Driemeier, Mary, and Reyes Aterido. 2007. Comparing Apples with . . .
Apples: How to Make (More) Sense of Subjective Rankings of Constraints to
Business. Photocopy, World Bank, Washington, DC.
Haltiwanger, John, Stefano Scarpetta, and Helena Schweiger. 2006. Assessing Job
Flows across Countries: The Role of Industry, Firm Size, and Regulations. World
Bank Policy Research Working Paper 4070. World Bank, Washington, DC.
Haltiwanger, John, and Helena Schweiger. 2005. Allocative Efciency and the
Business Climate. Photocopy, University of Maryland.
Heckman, James, and Carmen Pages. 2004. Introductory Chapter. In Law and
Employment: Lessons from Latin America and the Caribbean, ed. J. Heckman and C.
Pages. Chicago: University of Chicago Press.
Hsieh, Chang-Tai, and Peter J. Klenow. 2009. Misallocation and Manufacturing
TFP in China and India. Quarterly Journal of Economics 124, no. 4:140348.
Jeong, Hyeok, and Robert Townsend. 2007. Sources of TFP Growth: Occupational
Choice and Financial Deepening. Economic Theory 32, no. 1:179221.
Johnson, Simon, John McMillan, and Christopher Woodruff. 2002. Property Rights
and Finance. American Economic Review 92, no. 5:133556.
Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2007. Governance Matters
VI: Governance Indicators for 19962006. World Bank Policy Research Working
Paper no. 4280, World Bank, Washington, DC.
Klapper, Leora F., Luc Laeven, and Raghuram G. Rajan. 2004. Business Regulations
as a Barrier to Entrepreneurship. Journal of Financial Economics 82, no. 3:591
629.
Loayza, Norman, Ana Maria Oviedo, and Luis Serven. 2006. The Impact of Reg-
ulation on Growth and Informality: Cross-Country Evidence. In Linking the Formal
and Informal Economy: Concepts and Policies, ed. Basuded Guha-Khasnobis, Ravi
Kanbur, and Elinor Ostrom. Oxford: Oxford University Press.
Love, Inessa, and Nataliya Mylenko. 2005. Credit Reporting and Financing Con-
straints. Credit Technology 50 (October): 733.
Micco, Alejandro, and Carmen Pages. 2006. The Economic Effects of Employment
Protection: Evidence from International Industry-Level Data. Discussion Paper
2433, Institute for the Study of Labor (IZA), Bonn.
Petrin, Amil, and Jagadeesh Sivadasan. 2006. Job Security Does Affect Economic
Efciency: Theory, a New Statistic, and Evidence from Chile. NBER Working
Paper 12757, National Bureau of Economic Research, Cambridge, MA.
Pierre, Gaelle, and Stefano Scarpetta. 2006. Employment Protection: Do Firms
Perceptions Match with Legislation? Economics Letters 90, no. 3: 32834.
Rajan, Raghuram G., and Luigi Zingales. 1998. Financial Dependence and
Growth. American Economic Review 88, no. 3:55986.
Restuccia, Diego, and Richard Rogerson. 2008. Policy Distortions and Aggregate
Productivity with Heterogeneous Establishments. Review of Economic Dynamics 11,
no. 4:70720.
Rienikka, Ritva, and Jakob Svensson. 2002. Coping with Poor Public Capital.
Journal of Development Economics 69, no. 1:5169.
Teal, Francis. 1999. The Ghanaian Manufacturing Sector, 19911995: Firm
Growth, Productivity and Convergence. Journal of Development Studies 36, no. 1:
10927.
Tybout, James R. 2000. Manufacturing Firms in Developing Countries: How Well
Do They Do, and Why? Journal of Economic Literature 38, no. 1:1144.
World Bank. 2004. World Development Report 2005: A Better Investment Climate for
Everyone. New York: Oxford University Press.
. 2008. The World Bank Enterprise Survey Database. World Bank, Wash-
ington, DC. www.enterprisesurveys.org.

The University of Chicago Press

Uploaded by

Copyright:

Available Formats

The University of Chicago Press

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The University of Chicago Press

Uploaded by

Copyright:

Available Formats

Big Constraints to Small Firms Growth?

Business Environment and Employment Growth

You might also like