NPS Report Year 1

Tanzania National Panel
Survey Report
Round 1, 2008-2009
United Republic National Bureau

of Tanzania of Statistics
Contents
Introduction ............................................................................................................................................5
CLUSTER 1:
GROWTH AND POVERTY REDUCTION ............................................................................................. 11
Goal 1: Ensuring sound economic management ..................................................................................20
Goal 2: Promoting sustainable, broad-based growth .............................................................................1
Goal 4: Reducing income poverty of both men and women in rural areas..........................................33
Goal 6: Provision of reliable and affordable energy to consumers ......................................................37
CLUSTER 2:
IMPROVEMENT OF QUALITY OF LIFE AND SOCIAL WELL-BEING ..................................................... 40
Goal 1: Ensure equitable access to quality primary and secondary education for boys and girls,
universal literacy and expansion of higher, technical and vocational education.................................42
Goal 2: Improved survival, health and well-being of all children and women and especially vulnerable
groups ................................................................................................................................................... 44
Goal 3: Increased access to clean, affordable and safe water, sanitation, decent shelter and a safe
and sustainable environment ...............................................................................................................47
Goal 4: Adequate social protection and rights of the vulnerable and needy groups with basic needs
and services...........................................................................................................................................52
CLUSTER 3:
GOVERNANCE AND ACCOUNTABILITY ............................................................................................ 54
Goal 6: Improved personal and material security, reduced crime, eliminate sexual abuse and
domestic violence .................................................................................................................................56
Appendix A. ‘Projecting’ poverty rates for comparison with the HBS .................................................59
Appendix B. Creating the price deflators .............................................................................................75
2
List of Tables
Table 1. National Panel Survey 2008/09 Sample Design .......................................................................1

Table 2. Cluster 1 MKUKUTA indicators available in the NPS ..............................................................15
Table 3. Comparing the NPS to the HBS based on “projected” poverty headcounts rates (%).............1
Table 4. Price inflation, annual percentage change in price indices, by strata and year a ....................1
Table 5. Unemployment rates in %: by age category and sex ................................................................1
Table 6. Labour force participation rates in %: ILFS 2006 and NPS 2008/09.........................................1
Table 7. Wage employment: Employment shares and wages by sector, ILFS and NPSa .......................1
Table 8. % of households with erosion problems.................................... Error! Bookmark not defined.
Table 9. % of households with irrigation ................................................. Error! Bookmark not defined.
Table 10. Use of fertilizers and pesticides/ herbicides ...........................................................................1
Table 11. Use of improved seeds............................................................................................................1
Table 12. % of households that received extensions..............................................................................1
Table 13. Ownership of farm implements (multiple responses allowed)...............................................1
Table 14. % increase in number of customers connected to the national grid and off-grid sources of
electricity ................................................................................................................................................1
Table 15. Main fuel used for cooking (%) by stratum and year ..............................................................1
Table 16. Cluster 2 MKUKUTA indicators available in the NPS ............................................................40
Table 17. Net primary school enrolment rates, by year, sex and stratum ............................................1
Table 18. Net secondary school enrolment rates, by year, sex and stratum ........................................1
Table 19. Proportion of children under 5 years-old with low height-for-age (stunted), weight-for-
height (wasted), or weight-for age (underweight), % ............................................................................1
Table 20. Proportion of births attended by various service providers ..................................................1
Table 21. Primary source of drinking water (%), by stratum and year/season ......................................1
Table 22. Toilet type (%), by stratum and year.......................................................................................1
Table 23. Household tenure status by stratum and year .......................................................................1
Table 24. Building materials (%), by stratum and year ........................................................................51
Table 25. Net primary school enrolment rate among disabled .............................................................1
Table 26. Net primary school enrolment rate among orphans .............................................................1
Table 27. Cluster 3 MKUKUTA indicators available in the NPS ............................................................54
Table 28. % of women 15-50 yrs old who agree with the following statements ..................................1
Table 29. Self-reported incidence of domestic violence (% of women 15-50 yrs old) ...........................1
Table 30. % of domestic violence victims who report abuse to various institutions ............................1
Table 31. Gender-based violence indicators, by respondent’s characteristics (women 15-50 yrs old) 1
Table 32. Average values for independent variables in the model of log consumption, by strata and
year .......................................................................................................................................................66
Table 33. Regression of log consumption per adult equivalent on household characteristics in 2007
HBS........................................................................................................................................................70
Table 34. Projected poverty rates based on regression model of log consumption ...........................74
Table 35. Imputation of missing price data from the raw CPI data files................................................1
3
Table 36. Price inflation, annual percentage change in price indices, by strata and year a ..................1
Table 37. Average budget shares and prices underlying the price indices, 2007 and 2009. (Dar es
Salaam).................................................................................................................................................. 82
Table 38. Average budget shares and prices underlying the price indices, 2007 and 2009. (Other
Urban) ................................................................................................................................................... 83
Table 39. Average budget shares and prices underlying the price indices, 2007 and 2009. (Rural) ...84
Table 40. Average budget shares for non-food items, 2007 and 2009. ...............................................85
4
Introduction
over time. As such, the NPS is intended to
provide a key benchmark for tracking progress
What is the National Panel on poverty reduction and a wide range of
Survey? other development indicators.1
The NPS is nationally-representative
household survey which provides measures of The second goal of the NPS is to facilitate
poverty, agricultural yields, and other key better understanding of the determinants of
development indicators. The NPS is an poverty reduction in Tanzania. The NPS will
enable detailed study of poverty dynamics at
“integrated” household survey, in that it
covers a broad range of topics in the same two levels. In addition to tracking the
questionnaire – from education and health to evolution of aggregate poverty numbers at
crime, gender-based violence and a range of the national level in years between Household
other sections – to allow analysis of the links Budget Surveys, the NPS will enable analysis
between sectors and the determinants of of the micro-level determinants of poverty
development outcomes. reduction at the household level. Panel data
will provide the basis for analyzing the causal
Current plans are for the NPS to be repeated determinants of income growth, increasing or
biennially, i.e., every 2 years. Thus round 2 decreasing yields, improvements in
will begin in late 2010. The term “panel” in educational achievement, and changes in the
the NPS title refers to surveys that return to quality of public service provision over time
the same interviewee on multiple occasions by linking changes in these outcomes to
over time. The 2008/09 round is the first household and community characteristics.
round of the NPS. However, in future years
A third objective of the NPS is to provide data
the NPS will return to all of the households
interviewed in 2008/09 to track their to evaluate the impact of specific policies
outcomes over time. and programs. With its national coverage and
long time frame, the NPS will provide an ideal
Objectives platform to conduct rigorous impact
The National Panel Survey (NPS) was designed evaluations of government and non-
to meet three principle objectives. The first,
overarching goals was to monitor progress 1
In many cases, the NPS provides data on
toward the goals set out in the National
indicators already measured, often on a less
Strategy for Growth and Poverty Reduction frequent basis, in the official sources designated in
(aka, the MKUKUTA goals) and other national the MKUKUTA monitoring framework. Readers
development objectives (MDG, PAF, etc.). are referred to the MKUKUTA Monitoring
The NPS provides high-quality, annual data on Management System and Indicator Framework
a long list of MKUKUTA indicators that is both which lists the official source for each of the
nationally representative and comparable indicators measured here.
5
government development initiatives. To The guiding principle in the choice of sample
achieve this goal, the NBS will need to work in size, following standard practice for NBS
close collaboration with the relevant line surveys, was to produce estimates with a 95%
ministries to link administrative data on confidence interval no larger than 5% of the
relevant projects to changes in development mean for key indicators. In this case,
outcomes measured in the survey. household consumption and maize yields
were used as the basis for those calculations.
Sample design
In order to monitor progress toward the The NPS was based on a stratified, multi-stage
MKUKUTA goals, it was vital that the NPS cluster sample design. The principle strata
have a nationally-representative sample were Mainland versus Zanzibar, and within
design. As such, in 2008/09 the NPS these, rural versus urban areas, with a special
interviewed 3,280 households spanning all stratum set aside for Dar es Salaam. Within
regions and all districts of Tanzania, both each stratum, clusters were chosen at
mainland and Zanzibar. random, with the probability of selection
proportional to their
The sample size of 3,280
population size. In urban
households was The NPS is based on a nationally areas a ‘cluster’ was
calculated to be sufficient representative sample of 3,280 defined as a census
to produce national households across 410 clusters. enumeration area (from
estimates of poverty, This sample was designed to the 2002 Population and
agricultural production
produce national estimates and to Housing Census), while in
and other key indicators.
allow disaggregation between Dar rural areas an entire
It will also be possible in
es Salaam, other mainland urban village was taken as a
the final analysis to
areas, mainland rural areas, and cluster. This primary
produce disaggregated
Zanzibar. It is not possible to motivation for using an
poverty rates for 4
entire village in rural areas
different strata: Dar es produce regional or district level
was for consistency with
Salaam, other urban areas statistics.
the HBS 2007 sample
on mainland Tanzania,
which did likewise.
rural mainland Tanzania,
and Zanzibar. Alternatively, estimates of most Table 1 shows the break-down of the sample
key indicators can be produced at the zone by geographic stratum. Based on the 2002
level, as used for the Demographic and Health Population and Housing Census, rural
Survey (DHS) reports and other surveys. residents comprise roughly 77% of the
There are 7 of these zones in total on the population, compared with 63% of the NPS
mainland: North, Central, Eastern, South, sample. The NPS sample gives slighter greater
Southern Highlands, West and Lake. As with weight to urban areas due to the higher levels
any survey though, the confidence of the of inequality in these areas, and added
estimates declines as statistics are difficulty in estimating poverty rates and
disaggregated into smaller zones. other statistics. Similarly, Zanzibar comprised
roughly 3% of the Tanzanian population in the
Due to the limits of the sample size it is not
2002 census, but constitutes nearly 15% of
possible to produce reliable statistics at the
the NPS sample, so as to allow separate
regional or district level.
6
Zanzibar-specific estimates to be presented Table 1. National Panel Survey 2008/09
for most indicators. Sample Design
House
Finally, although it has been stressed that the Clusters -holds
2008/09 round is the first year of the NPS, the Mainland Total 350 2,800
sample design for year 1 was deliberately Dar es Salaam 70 560
linked to the 2007 HBS to facilitate Other urban areas 52 416
comparison between the surveys. On Rural areas 228 1,824
mainland Tanzania, 200 of the 350 in the NPS Zanzibar Total 60 480
were drawn from the 2007 HBS sample (this Urban areas 30 240
included all 140 rural HBS clusters). Within Rural areas 30 240
these 200 HBS clusters, a portion of the (8) Tanzania Total 410 3,280
households sampled for the NPS were taken
from the sample of (24) HBS households in
the cluster.2 Shinyanga; Central zone including Dodoma,
This design created a panel of approximately part of Iringa, Morogoro, Singida and Tabora;
1,200 HBS households – interviewed in both Southern zone including part of Iringa, Mbeya,
the HBS and NPS – within the total sample of Rukwa and Ruvuma; Eastern zone including
Lindi, Mtwara, and Pwani; the Dar es Salaam
3,280 NPS households.
zone and finally a separate zone for Zanzibar.
Timeline & organization of fieldwork
The first round of the NPS was collected over Within each zone, each district and each
a 12-month period between October 2008 region were visited at 3 separate (randomly
and September 2009. assigned) points during the year, so as to
account for seasonal fluctuations.
Seven mobile survey teams conducted
interviews year round, with each team The mobile teams spent roughly 4 to 5 days in
working year round in a specific “work zone” each cluster (village or urban enumeration
of the country. Note that in order to balance area). The first day was devoted to listing the
the workload and travel times across teams, cluster, i.e., compiling a list of the population
these work zones did not correspond of households in the cluster from which to
perfectly to the administrative zones of the draw a sample. The second and third days
country. (The work zones were divided as were devoted to interviews and the fourth to
follows: North-coast including Arusha, finalize data entry, call backs, etc. Median
Kilimanjaro, Mara, Manyara and Tanga; Lake interview time was approximately 2.5 hours
zone included Kagera, Kigoma, Mwanza and for the household questionnaire and 1.5
hours for the agricultural questionnaire.
2
The number of HBS households sampled varied Considerable additional time was spent on
from cluster to cluster, in proportion to the share anthropometric measurement of all
of the population, as measured through a household members and taking direct GPS
comprehensive household listing, that had measurement of a sub-sample of
remained stationary in the cluster since the time respondents’’ farm plots.
of the HBS. This was done to ensure that the NPS
sample remained nationally representative despite Each mobile team was overseen by a
possible non-random attrition of HBS households. supervisor from NBS and included a driver,
7
four enumerators, and a data entry operator of farm by-products, etc. For a sample of
equipped with a laptop. The data entry roughly 25% of the farming households,
operator was responsible for entering all enumerators used GPS devices to directly
questionnaires using the CsPro software measure the size of all farming plots.
package while in the field, conducting
consistency checks of the data and instructing Finally, apart from the questionnaires
enumerators to re-visit households when administered to households, a separate
problems were flagged by the software. Once community questionnaire collected
entered and validated in CsPro, the electronic information from village, kitongoji and/or
data was sent on a weekly basis from the field mtaa leaders. The community questionnaire
covered topics including local administration
teams to NBS headquarters by email using 3G
modems. and governance and access to basic services.
Rather than simply tabulating the data from

Contents of the survey and outline of this
report the hundreds of questions asked in these
The main survey instrument of the NPS was various questionnaires, this report is
the household questionnaire. This was organized around the monitoring framework
administered to all for the MKUKUTA goals.
households in the sample. It is hoped that this
General household The 2008/09 survey is the first, organization provides a
information – including baseline round of the NPS. The more intuitive
food consumption and survey is scheduled to be repeated presentation of the results
other household that links directly to the
on an annual basis for at least the
expenditure, which is national policy dialogue.
next 5 years, providing regular,
central to poverty It is not intended,
comparable updates on key
measurement – was however, that the NPS
development indicators. report should in anyway
solicited from the
household head or supplant existing
another knowledgeable adult member of the MKUKUTA documents or
household. In addition, wherever possible, the Tanzania Poverty and Human
each individual member over 5 years of age Development Report. These latter
was interviewed directly for sections on publications draw on data from a wide variety
education, health, labour, and food eaten of sources to measure progress on MKUKUTA
outside the home. indicators. The NPS report focuses narrowly
on the goals and indicators collected through
In addition to the household questionnaire, a the NPS, drawing only occasionally on other
separate 46-page agricultural questionnaire datasets for the purpose of comparison.
was administered to all households with any
agricultural activities (including farming, In a number of places, the NPS questionnaires
fishing or livestock, or ownership of any provide extra detail relevant to MKUKUTA
shamba even if not under cultivation). The progress that goes beyond the specific
agricultural questionnaire included detailed indicators outlined in the MKUKUTA
sections on each plot and each crop under monitoring framework. In such cases,
cultivation, as well as information on farm additional tables and statistics have been
assets, extension services, use and marketing presented – in the relevant sections of the
8
report – as a way of providing a deeper possible, this report makes direct
understanding of the process at work comparisons to previous nationally-
underlying progress on the core indicators. representative surveys to put the NPS data in
Key examples here are the enormous detail context and highlight trends over time.
available on smallholder farming activities However, due to differences in methodology,
(presented under Cluster 1), which go far there are often limits to the detail and
beyond the basic MKUKUTA indicators on reliability of these comparisons.
technology usage and food production, and
the in depth questions in the NPS on gender- Going forward, the NPS is intended to be
based violence (data from which is presented repeated every two years. Round 2 in 2010-
2011 will return to all 3,280 households from
under Cluster 3).
round 1 with a nearly identical questionnaire.
The future of the NPS As a result, in 2011 and every second year
The 2008-09 survey is the baseline round of afterward, the NPS will provide a detailed and
the NPS. Thus it provides a snapshot of rigorously comparable picture of changes in
development and household welfare in household welfare and economic activities
Tanzania at a given point in time. Wherever over time.
9
10
Summary of Key Findings
poverty in urban areas outside Dar es Salaam
(from a basic needs poverty rate of 24.1%
Poverty down to XXX%), and a small increase in rural
The headcount poverty rate measures the areas (from 37.6% to XXX%).
percentage of the population living below the
national poverty lines. In 2007, Tanzania’s
food poverty line (or the cost of acquiring Inflation
enough food for subsistence) was set at T.Sh. The NPS data on household expenditure
13,098 in Dar es Salaam, T.Sh. 10,875 in other provides an independent source of
urban areas, and T.Sh. 9,574 in rural areas, information about price levels in Tanzania.
while the basic needs poverty line (which The NPS price index computed for this report
includes the cost of other, non-food differs from the official CPI in a number of
expenditure) was set at T.Sh. 17,941, T.Sh. ways: it relies on household survey data
14,896 and T.Sh. 13,114 in these same three rather than market prices; it focuses on the
strata. consumption habits of the poor and includes
rural areas in the analysis; and it uses newer
The best available estimates suggest there data on budget shares to weight the
was virtually zero change in the proportion of individual prices in the index. However, the
Tanzanians living below these poverty line index is not intended to serve as a substitute
from the 2007 HBS to the 2008/09 NPS.3 The for the CPI, but rather as a complementary
food poverty rate rose marginally from 16.6%
data source.
nationally to XXX%, while basic needs poverty
rose from 33.6% to XXX%. Given the margin Comparing price levels in the 2007 HBS to the
of error in the survey, this is consistent with NPS, the inflation rate measured here was
the long term trend of a fall in the national, 13.9% from Jan. 2007 to Jan. 2008, 20.4%
basic-needs poverty headcount rate of about from Jan. 2008 to Jan 2009, and 8.2% per
0.3% per annum since 1991. Looking at the annum in the first 3 quarters of 2009. This
geographic strata separately, poverty in Dar pattern of inflation is consistent with the CPI
es Salaam appears unchanged. However, series in the following respects: moderately
there is evidence of a significant decline in high inflation levels in 2007 accelerated in
2008, driven primarily by food prices, but
3
The NPS collects data on household consumption peaked in 2008 and fell to a fairly low level in
which can be used to measure household welfare the first three quarters of 2009. However, the
and the headcount poverty rate. However, there inflation rates reported here are consistently
are key methodological differences from the
higher than rates derived from the official CPI
Household Budget Surveys (HBS), making any
– particularly in 2008 when official inflation
direct comparison of consumption data across the
two surveys impossible. Instead, this report uses
was 12.9% and the NPS index records a rate of
statistical techniques to produce estimates of 20.4%. The analysis below shows that part of
current consumption that are more comparable to this discrepancy is due to differences in
the HBS, based on indicators such as asset weighting of individual prices, however this
ownership, housing amenities, etc., that are cannot account for the full difference.
collected in identical ways in the two surveys.
Comparing inflation across rural and urban from 13.8% in other urban areas in 1991 up to
areas, the patterns are strikingly similar. 24.1% in 2008/09.
However, Dar es Salaam is an outlier, with
much higher inflation in 2008, driven primarily Looking at long term wage trends, there is a
stark difference between the 1990s and the
by non-food prices.
2000s. During the 1990s real, median
earnings grew steadily for all occupations –
Employment & Income but particularly for the public sector, where
Unemployment in 2008/09 remained low, at real wages grew at 12.9% per annum from
just 3.2% by the standard definition, and 1.8% 1991 to 2000 as measured by the HBS. Since
by the more restrictive ILO international 2000, real wage growth for both public- and
definition. Unemployment was highest private-sector wage workers has more or less
among the young (age 15-24), at 7.6% and stopped. Throughout the decade, median
4.0% by these two definitions. wages, in real 2007 Shillings, remained at
around T.Sh. 110,000 to T.Sh. 115,000 for
Looking at earnings levels – and comparing public sector workers, and around T.Sh.
public-sector wage works, private-sector wage 41,000 to T.Sh. 44,000 for private-sector wage
workers, and the non-farm self employed – employees. Measured real earnings for the
median, nominal, monthly earnings in Dar es non-farm self-employed have been more
Salaam were T.Sh. 230,000, T.Sh. 104,000 and volatile in the 2000s, but show some signs of
T.Sh. 160,000 for these three occupations, moderate growth during the 2000s.
respectively. In other urban areas the figures
were T.Sh. 170,000, T.Sh. 57,000, and T.Sh.
75,000; in rural areas T.Sh. 142,000, T.Sh. Agriculture
50,000 and T.Sh. 72,000, and in Zanzibar, T.Sh. The NPS contains an extensive agricultural
98,000, T.Sh. 87,000 and T.Sh. 75,000. Thus module, administered to all households
the gap between the public and private sector involved in farming, fishing, or livestock
was largest in Dar es Salaam, at over 120%, cultivation.
and lowest in Zanzibar at just 13%.
Survey results show modest gains in crop
The report also summarizes longer-term production during the 2008 long rainy-season
trends in employment and earnings, (masika) relative to the 2002 long rainy-
comparing various survey sources from 1991 season covered by the 2002/03 National
to 2009 and dividing the adult population (age Sample Census of Agriculture. The masika
15-65) by their main occupation. The main harvest of major cereal crops – maize, rice
trends that emerge in terms of employment paddy and sorghum – increased moderately
levels by occupation are (i) no significant between the two surveys: up by 31.1% for
trends in the size of public or private wage maize, up 34.8% for paddy, and up 40% for
employment, (ii) volatility and/or sorghum. However, production of cassava,
inconsistency across surveys in the share of another major food crop, declined by 44.8%
the population in farming, but no sign of a between the surveys.
steady decline, and (iii) steady growth in the
share of urban workers who are self- One clear reason for the lack of large gains in
employed, from 18.4% in Dar es Salaam in small-holder production has been the failure
1991 to 29.7% in the NPS in 2008/09; and to adopt improved farm technologies. The
12
number of households using irrigation rose education policy which saw net primary
modestly from 240,721 in 2002/03 to 276,958 enrolment increase on the mainland from
in 2008/09. However, the share of 58.7% in 2000 to 83.7% in 2007, as measured
households using inorganic fertilizer showed by the HBS.
no change – 12.0% in 2002/03 and 11.6% in
2008/09. (Use of organize fertilizer fell from Net secondary enrolment rates (for Form I to
26% to 19.2% over the same period.) IV) continued to rise rapidly: from 5.1% for
Similarly, the share of households using the mainland in 2001, to 15.2% in 2007 and
improved seed varieties grew only slightly 23.5% in 2008/09. In part this may reflect the
from 18.0% to 19.5%. delayed effect of primary enrolment increases
earlier in the decade. As with primary
enrolment, for the country as a whole
Public Services (including Zanzibar) the overall net secondary
The NPS covers household access to and use enrolment rate was higher for girls (27.3%)
of a range of public services, including than boys (23.2%).
electricity generation, water supply,
Health. The NPS collects information on
sanitation, education, health facilities, and so
health facility usage, and takes detailed height
on.
and weight measurements for all household
Electricity. Access to the electricity grid members to track nutritional status. One key
remained low, at 59.3% in Dar es Salaam, indicator on this front in the MKUKUTA
27.4% in other urban areas, just 2.1% in rural monitoring framework is the proportion of
areas, and 24.7% in Zanzibar. For the most births attended by a skilled health worker – an
part these levels reflect little or no change indicator with obvious links to both infant and
from recent rounds of the NPS. maternal mortality. Among women who
reported giving birth to a child in the last 24
Water. The proportion of households with months, 47.3% reported giving birth to their
access to piped water in 2008/09 was 74.1% most recent child in a hospital, 7.6% in a clinic,
in Dar es Salaam, 60% in other urban areas, 43.5% at home, and 17.8% elsewhere.
22.7% in rural areas, and 79.8% in Zanzibar.
For the mainland as a whole, the 2008/09 rate Looking at nutrition indicators, low height-for-
of roughly 40% reflects a modest increase age or “stunting” among children under 5
over previous years. Previous rates were years old provides an indicator of chronic
35.9% in 1991, 39.3% in 2001, and 33.9% in malnutrition. The rate of severe stunting for
2007. all children under 5 was 15.6% in rural areas,
5.4% in urban areas, 11.2% in Dar es Salaam
Education. The net primary school enrolment and 9.1% in Zanzibar.
rate calculated in the NPS was 81.9% for the
country as a whole (mainland and Zanzibar),
and significantly higher for girls (85.0%) than Gender-Based Violence
for boys (78.6%). On the mainland, primary The first round of the NPS contained an
enrolment rates showed a slight decline (by experimental module which is the first
approximately 1.4%) compared to 2007, but attempt to measure the incidence of actual
maintained most of the dramatic gains in gender-based violence at the national level (in
enrolment achieved under the free primary contrast to the opinions about GBV as
13
measured by the Demographic and Health are more progressive among younger women,
Surveys). Based on women’s own reports, but the incidence of abuse is not significantly
14.1% of women aged 15 to 50 years old said different. Patterns of abuse are quite similar
their partner had slapped them or thrown across rural and urban areas. Divorced or
something that could hurt them; 10.7% separated women are the most likely to
reported being pushed or shoved, 8.6% hit report being physically abused.
with a fist, and 6.5% reported being forced to
have sexual intercourse by their current The majority of women affected by GBV
partner. report the abuse only to family members
(53%). In 20.5% of cases, women reported
Looking at the correlation between GBV and going to community or village leaders, and in
women’s personal characteristics, women’s only 6.4% of cases to the police.
belief that abuse is justified falls dramatically
with education, but the actual incidence of
abuse does not. Similarly, views about GBV
14
CLUSTER 1:
GROWTH AND POVERTY REDUCTION
The National panel survey is, first and covered in the NPS questionnaire. It should
foremost, a socio-economic survey. A high be noted that primarily due to time
proportion of the questionnaire is devoted to constraints in the few weeks since fieldwork
measuring household consumption and ended, not all available indicators are
wealth, as well as farm production for small- presented in the current report. Additional
holder households. indicators will be computed prior the release
of official results in early 2010.
The following table provides a list of the
MKUKUTA indicators under Cluster 1,
denoting whether or not the indicator is
Table 2. Cluster 1 MKUKUTA indicators available in the NPS

Indicator Covered in NPS?
In
GDP growth per annum No
GDP growth of sectors per annum No
Gini coefficient Yes4
Headcount ratio, basic needs poverty line Yes
Goal 1: Ensuring sound economic management
Annual rate of inflation Yes
Central Government revenue as % of GDP No
Fiscal deficit as % of GDP (before and after grants) No
External Debt Service as % of Exports No
Export as % of GDP No
Goal 2: Promoting sustainable and broad-based growth
Unemployment Rate Yes
Domestic credit to private sector as % of GDP No
% increase in foreign direct investment No
Interest rate spread on lending and deposits No
% of rural population who live within 2 kms of an all-season passable road (Rural No
access indicator)
% of trunk and regional road network in good and fair condition No
Proportion of enterprises undertaking Environmental Impact Assessments No
Food self sufficiency ratio No
Proportion of districts reported to have food shortages No
4
Not calculated in this report. Further analysis of consumption data from the NPS is planned in the future,
which will include the Gini coefficient.
Indicator Covered in NPS?
In
% change in food crop production Yes
Proportion of households who take no more than one meal per day No5
% of small holders participating in contracting production and out-grower Yes*
schemes
% of small holders using modern methods of farming (irrigation, fertilizers and Yes
hybrid seeds)
% of small holders who accessed formal credit formal credits for agricultural Yes*
purposes
% of small holder households who have one or more off-farm income generating Yes*
activities
% of households whose main income is derived from the harvesting, processing Yes
and marketing of natural resource products
% increase in number of customers connected to the national grid and off-grid No6
sources of electricity
% of households in rural and urban areas using alternative sources of energy to Yes
wood fuel (including charcoal) as their main source of energy for cooking
Total electricity generating capacity and utilization No
*Indicators covered by the NPS questionnaire but which are not yet covered in this report.
5
While this specific indicator was not included in the year 1 questionnaire, much more detailed information on
food consumption is available, and it is possible to calculate daily caloric intake.
6
Household connectivity to the national grid is measured.
16
Cluster-wide indicators: Poverty &
Inequality
such as the HBS and NPS. Box 1 provides
preliminary estimates of “projected” poverty
Providing a benchmark for the rates based on these methods. It should be
next MKUKUTA targets stressed, however, that these results are very
Measuring poverty, through the collection of preliminary and subject to revision upon
detailed data on household consumption, is a release of the final 2008/09 NPS report in
central goal of the National Panel Survey. By early 2010.
using a consistent methodology and
questionnaire in each subsequent round of
the survey, the NPS will provide comparable A cautionary note on making
measures of household welfare on a routine comparisons between the HBS
basis. Because the completion of the first and the NPS
round of the NPS coincides with launch of the Household Budget Surveys conducted in
new MKUKUTA targets (as well as Kilimo 1991/92, 2001, and 2007 have traditionally
Kwanza goals), it is intended that the new provided the main source of information on
poverty line and headcount poverty rates household welfare in Tanzania. The National
derived from the NPS will serve as a Panel Survey, which will be repeated on a
benchmark for measuring progress toward much more frequent basis to track annual
these goals. progress on a variety of MKUKUTA indicators,
replicates a number of key features of the
The focus of the new poverty calculations for
HBS. However, there are important
the NPS will be to provide a starting point for
methodological differences that make any
comparisons going forward. However, it is
direct comparison of poverty rates across the
also necessary to place current results in an
surveys potentially misleading.
historical context. To do so requires
comparison to the poverty figures published The principal methodological differences are
in the various HBS reports since 1991/92. outlined below. Many of these changes were
Unfortunately, a simple direct comparison of driven by an attempt to ensure strict control
consumption figures between the HBS and over data quality during fieldwork, including
NPS is not feasible, for reasons documented in the use of a much smaller and more closely
the following sub-section. supervised corps of enumerators in the NPS.
These enumerator teams were ‘mobile’,
A partial bridge linking the NPS poverty results
spending only a few days in each enumeration
to the HBS can be provided by using
area or village, in contrast to the locally-
techniques developed for “Small Area
recruited enumerators for the HBS who
Estimation” or poverty mapping. These
resided in their respective survey areas year
techniques are less reliable than direct
round.
measurement of consumption, however, they
are specifically intended to overcome
methodological differences between surveys
1.1 “Recall” versus “diary” of home-produced food to measure food
measurement of food consumption. However, these subjective
consumption valuations are not solicited in the NPS.
First, the NPS collects data on food Rather, home-produced food will be assigned
consumption by asking the head of the a value by relying on prevailing prices in the
household or their spouse to recall how much geographic stratum as reported by other
they have eaten of various food items in the households who reported purchases of the
past seven days. (See Section K of the item. This approach, relying on locally
“Household and Individual Questionnaire”.) reported ‘unit values’ is increasingly common
In contrast, the HBS requests households to in poverty analysis, and will provide the basis
keep a running diary of their food intake (and for the new annual, NPS poverty series.
other expenditures) for 30 days.
1.2 A shorter list of food items

A second difference between the HBS and
NPS questionnaires themselves is in the list of
food and non-food items for which
consumption data is collected. See the
appendix for a detailed list of the
correspondence between HBS and NPS items
that was used in construction of the price
indices presented in the next chapter.
1.3 Different treatment of home-

produced food
Most Tanzanians, particularly in rural areas,
get a majority of their calories from food they
produce themselves. This food has never
been purchased and thus has no directly
observable market price. To construct a
monetary measure of welfare, it is necessary
to assign a monetary value to this home-
produced food. There are a number of
methodologies commonly used
internationally to do this, none of which is
perfect.
The HBS questionnaire asks respondents to

report a Shilling value for all food consumed,
whether it is purchased or produced at home.
For purchased food this should be the
purchase price, but for home-produced food
it is a subjective assessment of the food’s
value. The consumption aggregate in the HBS
reports relies on these subjective assessments
18
Box 1. Preliminary “projections” of poverty in 2008/09
As noted in the main text, the official poverty Disaggregating the results by geographic area
numbers from the NPS, based on household suggests that there has been a slight
consumption data, are still in process. These divergence between rural and urban areas in
numbers will not be directly comparable to the last 18 months. Both food poverty and
the HBS. However, a preliminary estimate of basic needs poverty declined (by 4.1% an
changes in welfare 6.0% respectively) in urban
between the 2007 HBS areas outside Dar es
and the 2008/09 NPS can Preliminary projections show no Salaam, while they rose by
be derived by modeling significant change in the a small margin (2.0% and
consumption using other headcount poverty rate between 2.5%) in rural areas. In Dar
household characteristics the 2007 HBS and the 2008/09 NPS es Salaam both figures rose
that are common across for Mainland Tanzania. However, by negligible amounts.
the surveys. These there is evidence of a decline in Further analysis of the NPS
include household urban poverty outside Dar es consumption data is
demographics, schooling required to lend more
Salaam and a slight increase in
and human capital, confidence to these results.
rural areas.
building materials and
household amenities, and asset ownership.
This box presents preliminary estimates based

on this methodology. Full details of the Table 3. Comparing the NPS to the HBS
underlying methods can be found in the based on “projected” poverty headcounts
appendix. rates (%)
Food
Table 3 shows the poverty headcount ratios Poverty Basic Needs
for each stratum (Dar es Salaam, other urban Rate Poverty Rate
areas, and rural areas) in each round of the Dar es ‘91/92 13.6 28.1
Salaam ‘01 7.5 17.6
HBS from 1991/92 to 2007, together with
estimated rates for 2008/09 using NPS data. ‘07 7.4 16.4
‘08/09 7.6 15.9
Results from the NPS 2008/09 suggest that Other ‘91/92 15 28.7
poverty has remained virtually unchanged in urban ‘01 13.2 25.8
Mainland Tanzania as a whole – with food
‘07 12.9 24.1
poverty rising (insignificantly) from 16.6% in
‘08/09 8.8 18.1
2007 to 17.4% in 2008/09 and basic needs
Rural ‘91/92 23.1 40.8
poverty from 33.6% to 34.0%. This is in line
‘01 20.5 38.7
with the long-term trajectory of poverty
‘07 18.4 37.6
documented in previous HBS reports, which
‘08/09 20.4 40.1
show that both food and basic needs poverty
have declined on average by about 0.3 Main- ‘91/92 21.6 38.6
land ‘01 18.7 35.7
percentage points per annum over the last 17
years, well within the margin of error used ‘07 16.6 33.6
here. ‘08/09 17.4 34.0
Goal 1: Ensuring sound economic
management
specific purpose of comparing household
welfare over time in the HBS and NPS
Indicator: Annual rate of datasets. The numbers presented here differ
inflation from the official CPI in three key ways: (a)
The annual rate of inflation measures the they cover rural as well as urban areas, (b)
percentage change in they give greater weight
prices over the course of a to the items consumed by
year. Maintaining low, Data on food expenditure can be the poor, and (c) they use
stable inflation is a key used to analyze price increases newer data on
indicator of economic between the HBS and NPS. consumption habits,
management in the otherwise known as
However, any attempt to compare
MKUKUTA framework. ‘budget shares’, to weight
these inflation numbers to the
However, both the prices in the
international and
official CPI should take careful note
consumption basket.
domestic events made of methodological differences
achieving this goal outlined in the appendix. The appendix at the end
especially difficult during of the report provides
the period covered here: details on the
from the beginning of 2007 through the third methodology used to measure prices in this
report, and explains the reasons why this
quarter of 2009. As noted in the Bank of
Tanzania’s June 2009 Monetary Policy methodology – and thus the results – differ
Statement, “the economy continued to somewhat from the official CPI series.
experience inflationary pressures that had
Figure 1 summarizes the pattern of inflation
arisen from the lagged effects of soaring
during 2007, 2008 and 2009 (ending in
world commodity prices earlier in the year,
September). The rates shown measure the
compounded by severe food supply shocks in
annual percentage change in a weighted
the region, and poor short rains in some areas
basket of prices, measured from January to
of Tanzania in the last quarter of 2008.”7
January (Jan 2007 to Jan 2008, Jan 2008 to Jan
This section of the report presents measures 2009, and Jan 2009 to Sept 2009). The items
of inflation based on the prices paid for goods in the consumption basket are weighted by
and services by households in the HBS 2007 the consumption habits (budget shares) of the
and the NPS 2008/09. The inflation rates poorest 50% of the population within each
presented in this section are not intended to stratum of the survey: Dar es Salaam, other
be directly comparable to the official CPI for a urban areas, and rural areas.
variety of reasons. Rather, the inflation
The left side of Figure 1 shows the level of
numbers presented here are designed for the
food price inflation. Food inflation jumps
7
Governor, Bank of Tanzania, Monetary Policy from 11.7% in 2007 to a high of 25.5% in
Statement, June 2009, p. vii. 2008, but falls rapidly again to a current rate
Figure 1. Price inflation, annual percentage change in price indices, by year
All prices Food prices Non-food prices
20.4 20.8 20
18.2
Annual Rate in Percent
16.3
15
13.9
12.8
10
9.5
8.2
5
5.1
0
2007 2008 2009 2007 2008 2009 2007 2008 2009
Food prices are based on costs reported by households in the 2007 HBS and 2008/09 NPS (i.e., ‘unit values’).
Disaggregated, raw data for non-food prices is taken from the CPI database and re-weighted using 2007 HBS and
2008/09 NPS budget shares.
of 6.5% per annum in the first three quarters outstanding question for policymakers is
of 2009. whether the inflation captured through the
CPI accurately measures price changes
The right side of Figure 1 shows non-food experience by the majority of Tanzanians who
price inflation. As seen, non-food price live in rural areas? Data from the NPS and
inflation has steadily declined over the period HBS can help to answer this question,
covered, from a high of 19.9% in 2007, to particularly for differences in food prices.9
13.4% in 2008 and just 1.7% per annum in the
first three quarters of 2009.8
9
Rural-urban differences All food prices presented in the main text are
As noted already, one key limitation to the based on “unit values” reported by households in
existing, official CPI series is that all data is the HBS and NPS data. However, the NPS does not
collected in major urban centres. It is an collect price information on non-food items. Thus
the non-food price indices rely on raw price data
urban price series. Thus an obvious
collected for the purposes of the official CPI series
in urban areas only. Nevertheless, separate non-
8
A cautionary note is important here: the basket food inflation rates are presented for urban and
of non-food items with comparable measurement rural areas here. The difference between them
across the HBS and NPS is relatively small, and is stems entirely from differences in consumption
disproportionately weighted toward energy and patterns: i.e., the weight given to different items.
fuel expenditures. The full list of items included in For instance, an increase in maize prices will have
the index can be found in the appendix. a greater impact on rural households, and rice
prices will have a greater impact on urban
21
Figure 2. Food and non-food price indices, by strata and month (Jan 2007 to Sept 2009)
Food Prices Non-Food Prices All Prices
Mainland DSM Other Urban Rural

2.5
Index based on Jan 07
1.5
1
07
08
09
09
07
08
09
09
07
08
09
09
07
08
09
09
l0
l0
l0
l0
l0
l0
l0
l0
n
p
Ju
Ju
Ju
Ju
Ju
Ju
Ju
Ju
Ja
Ja
Ja
Ja
Ja
Ja
Ja
Ja
Ja
Ja
Ja
Ja
Se
Se
Se
Se
All price indices begin at 1 in Jan 2007, and figures in the graph show the price level relative to this date.
Food price data and budget shares are not available from Jan to Sept. 2008, when neither the HBS nor the
NPS was in the field. The NPS also paused in April 2009, creating a second break. Price levels for these
periods are shown with dashed lines, connecting the nearest available data points.
Table 4 shows a break-down of food price The second point where a notable discrepancy
inflation, non-food price inflation, and the between inflation across strata is seen is the
overall index for the three Mainland strata relatively high rate of non-food price inflation
used in the surveys, and Figure 2 gives a in Dar es Salaam in 2007. (Note that this
visual depiction of the same information. difference represents differences in both
underlying prices and budget shares, as the
Two key points emerge from these figures. CPI collects separate price data for Dar es
First, food price inflation – while high across Salaam and each other regional capital.)
the board, particularly in 2008 – has been While note shown here, this increase was
fairly even across rural and urban areas. If largely attributable to a rapid surge in the
anything, urban areas have fared somewhat price of charcoal in Dar es Salaam in early
worse, with food price inflation peaking at 2007.
32.1% in 2008 (excluding Dar es Salaam)
compared to 23.8% in rural areas. While not
shown here, in welfare terms it is likely that
rural households are further cushioned from
these food price increases due to the larger
share of their food consumption which is
derived from home production.
households – as rural households eat relatively

more maize and less rice.
22
Table 4. Price inflation, annual percentage change in price indices, by strata and year a
2007 2008 2009 b

Dar es Salaam Food prices 10.4% 26.4% 1.8%
Non-food prices 41.3% 54.3% 8.9%
All prices 18.8% 38.6% 2.9%
Other urban Food prices 12.6% 24.1% 16.6%
All prices 12.7% 21.3% 13.3%
23
Box 2. Why do the NPS inflation figures differ from the official CPI?
Attentive readers will note that the inflation Up until 2009, the official CPI weighted each
figures cited in this report are significantly good according to its share of expenditure in
different from NBS’s Consumer Price Index – the 2000 HBS (this was recently updated to
the official measure of price inflation in use budget shares from the 2007 HBS). In
Mainland Tanzania. Why are the figures so contrast, the NPS index relies on data from
different? the 2007 HBS and the 2008/09 NPS. In
addition, because a primary focus of the NPS
To understand the differences, note that is on poverty measurement, the price index
there are two main ingredients in any price reflects the average prices paid by the poorest
index: (i) prices for individual goods at various 50% of Tanzanians.
locations for each month, and (ii) a set of
weights used to aggregate the prices of To provide an example, the total budget share
various goods and to average across the for food in the CPI is XX%, and this share is
prices paid by different people in different revised only every several years. In the NPS
places. The weight assigned to each good index however, the share of food in total
should reflect that good’s share of total expenditure is re-calculated every month, to
expenditure for a representative consumer, or reflect seasonal fluctuations and responses to
the “budget share”. The CPI and the NPS changes in food prices. The average food
price index use both different price data and share for the index was XXX% in 2007, XXX%
different budget shares. in 2008 and XXX% in the first three quarters of
2009.
First, consider the price data. The official CPI
is based on market prices: these are collected Which of these two factors is more important
by NBS staff during monthly visits to in explaining the difference between the NPS
marketplaces in the main city of each region and the CPI? Figure XXX helps to illustrate the
of the country. The NPS price index uses this relative importance of (i) different underlying
same price data for non-food expenditure. price data, and (ii) different weights in
However, for food prices, the NPS index is explaining the gap between the CPI and the
based on households’ reports of how much NPS index. The figure shows the inflation rate
they paid (per kilogram, litre or item) for the for each index in 2007, ’08 and ’09. The dark
food items they purchased in the past week – blue columns show the NPS index, using NPS
the technical term for these household prices household survey data for food prices and
is a “unit value”. Because the NPS visits NPS weights to create the index. The light
households all across the country, the unit blue columns show the official CPI, using
values it collects represent both urban and market price data and official weights from
rural consumers, while the CPI collects price the 2000 HBS. In between, the grey columns
data only in urban centres. show a hybrid index, using the market price
data from the CPI, but weighted using the NPS
Second, the NPS uses newer data on budget
weights. This hybrid measure highlights the
shares, and allows the spending habits role of different data sources and different
captured by these shares to vary each month. weights in explain the discrepancies between
24
the two main indices. The figure is differences in weights. (This is true for overall
reproduced for the overall price level on the prices and food prices, though for food it
left, and exclusively for food prices on the seems the differences in inflation for 2007
right, since this is the area where the NPS and 2009 are explained mostly by the
index and the official CPI differ in terms of different data sources: applying NPS weights
data sources. to the CPI market price data gives results
relatively similar to the official CPI.)
The results show that under ‘normal’
conditions, such as 2007 and 2009, the hybrid In 2008 however, the situation is quite
index gives an inflation rate in between the different. The hybrid index, combining CPI
(higher) NPS index and the (lower) official CPI market price data with NPS weights, gives a
inflation rate. Thus the difference is explained much higher inflation rate than either of the
partially by differences in data and partially by other indices. [Explanation???]
Figure 3. Inflation rates using different price data & different budget shares
NPS prices, NPS weights CPI prices, NPS weights CPI prices, CPI weights
All prices Food prices only
31.9 30
Annual Inflation Rate in Percent
29.5
20.8 20
20.4
18.6
13.9
12.9 12.8
11.7 10
10.5 10.1
9.5
8.6 8.2
3.9 3.7
3.0
2.1
0
2007 2008 2009 2007 2008 2009
Inflation rates are measured from Jan of the year listed to Jan of the following year. (This choice of periods, rather than
Dec to Dec, is due to the lack of survey data for Dec 2006). Due to the duration of the survey and limitations on data
availability at the time of writing, for 2009 the inflation rates reflect changes in the price indices from Jan 2009 to July
2009.
25
Goal 2: Promoting sustainable, broad-
based growth
Table 5. Unemployment rates in %: by age
category and sex
Indicator: Unemployment rate Int’l. Standard
Under the internationally recommended By age: Definition Definition
definition, a person is classified as 15-24 4.0 7.6
unemployed if she/he meets all the following 25-34 2.3 3.4
three conditions during a specified period 35-64 0.4 0.9
(one week), that he/she is: (a) without work, 65+ 0.1 0.4
(b) available for work, and (c) seeking work. Total 1.8 3.2
these two components of unemployment
The international recommendations allows gives the overall unemployment rate based on
the relaxation of condition (c) above, i.e. the relaxed standard definition, hereafter
seeking work, especially in countries where a referred to as the standard definition.10
large proportion of the population is engaged
in agriculture and informal activities with Table 5 shows unemployment rates by both
generally low knowledge of labour market of these definitions in the 2008/09 NPS,
developments in the rest of the economy. broken down by age and sex. Overall, total
unemployment in Mainland Tanzania in this
Tanzania is characterized by the above- period was 1.7% by the international
mentioned conditions, and therefore uses the definition and 3.1% by the standard
relaxed international definition of definition. By both definitions unemployment
unemployment, while at the same time is highest for the youngest members of the
presenting results according to the stricter labour force: 4% and 7.7% by the respective
Table 6. Labour force participation rates definitions for individuals 15 to 24 years of
in %: ILFS 2006 and NPS 2008/09 age. Looking at the gender breakdown, there
are no differences between males and
Women
females by the standard definition. However,

Men
All
men show slightly higher rates of

DSM 2006 87.3 86.2 88.5
2008/09 88.6 89.9 87.0
Other 2006 89.0 88.4 89.8 10
The Integrated Labour Force Survey, 2006, also
urban 2008/09 87.8 89.1 86.4 computes a third definition of unemployment,
international definition for comparison with referred to as the “national definition”, which
other countries. Those persons who were includes individuals with “extreme marginal
without work, available but have not taken attachment to employment”. Unfortunately it is
active steps to find work, thus satisfying not possible to calculate unemployment by this
conditions (a) and (b) above, are referred to definition in a comparable way in the 2008/09
as unemployment category B. The sum of NPS.
26
unemployment by the international definition between the ILFS and the NPS. However, it is
(2.1% versus 1.4% for women). possible to compare another important labor
market indicator, the participation rate.
Overall, unemployment as defined here is
extremely low in Tanzania as a whole, Table 6 shows a comparison of participation
reflecting the tendency for poorer individuals rates in the NPS with the ILFS. The overall
to find free-entry, low-income work in small- trend shows a slight decline in the labour
holder agriculture or informal self- force participation rate for Mainland Tanzania
employment rather. as a whole, from 91.6% to 90.9%.
Disaggregating by gender, this decline is
The most authoritative source for labour
concentrated among women, whose
market indicators in Tanzania is the Integrated
participation right declines from 92% to 90%
Labour Force Survey (ILFS), which was last
between the two surveys. In all cases though,
conducted over the calendar year of 2006.
these changes are not sufficiently large
Unfortunately, due to changes in the wording
relative to the sample sizes to provide a
for questions related to availability for work,
confident signal an economically significant
job search, and treatment of unpaid
shift in the structure of the labour market.
household labor, it is not informative to make
The overall picture is of relative stability
a direct comparison of unemployment rates
Table 7. Nominal and real median monthly earnings, by occupation, 1991 – 2008/09a
Self-employed earnings Private sector wages Public sector wages

Nominal Real b Nominal Real b Nominal Real b
1991 HBS 4,167 18,260 6,000 25,364 7,167 32,924
2000 HBS 16,667 30,616 30,250 53,846 52,000 98,087
2001 ILFS 26,000 44,082 30,000 50,278 65,000 115,442
2006 ILFS 45,500 46,511 45,000 44,771 110,000 114,697
2007 HBS 29,167 27,432 45,000 41,231 116,000 110,431
2009 NPS 90,000 56,081 70,000 42,163 170,390 111,341
a
The public sector here includes both local and central government employees, as well as
individuals working for religious organizations, political parties, NGOs or international
organizations. The private sector includes parastatal enterprises as well. “Farming” includes
crop-farming as well as livestock keeping and forestry. “None” includes individuals who are not
in the labour force (including the disabled and students), the unemployed and unpaid family
workers. Precise definitions for each category in each survey are listed in Appendix C.
b
Real earnings are expressed in (June) 2007 Tanzanian Shillings. Nominal and real figures from
2007 do not perfectly coincide because real figures take account of geographic price differences
as well as differences across time. Thus wages from Dar es Salaam are deflated to reflect
national price levels, and similarly for urban and rural areas. All wages are deflated using the
poverty lines published in the HBS reports. For intermediate years (2001 and 2006), levels of the
poverty line are estimated using a logarithmic interpolation. For the 2008/09 NPS, wages are
deflated using the price index presented in the chapter above on inflation. This price index was
constructed to be a continuation of the deflators underlying HBS poverty line.
27
rather than change. Figure 4. Median Monthly Wages, Public and
Private Sector Wage Employees
Given the attention paid in recent years on
the pace of job creation, Table 7 provides a
breakdown of the wage employment rates Median monthly earnings
and wage rates by three different sectors: Public Sector Private Sector Self-Employed
public employment, which includes both 230
central and local government, the private DSM 104
160
sector, and “other” categories which include
170
political parties, coops, NGOs, international Other Urban 57
organizations and religious organizations. The 75
table is restricted to non-agricultural wage 142

Rural 50
employment. 75
98
Looking at the first two columns which deal Zanzibar 87
with the distribution of employment, overall 75
there is a sign of a small increase in the size of 0 50 100 150 200 250
Nominal TSh in '000s
the private sector between the two surveys.
Further rounds of data collection will be
needed to establish whether this is a genuine
trend rather than a cyclical fluctuation or
and 27.3% respectively. In assessing these
sampling fluke. Nevertheless, the available
changes it is important to bear in mind that
evidence points to an increase from 5.1% to
these are nominal changes and should be
6.7% of the mainland adult population (15
compared to changes in the CPI over the
years or older) in private sector wage
2006-2009 period to appreciate the
employment. In proportional terms, this
significance of these movements in real
increase has been largest in rural areas, where
terms.11
the share nearly doubled from 2.1 to 4.1% of
the population working in private sector wage
employment. Nevertheless, this overall level
remains quite low, and even in Dar es Salaam,
the proportion of the adult population with a
private sector wage job is just 18.3%.
Turning to wage levels, the public sector

continues to pay dramatically higher wages
than the private sector: an average of
TSh.312,130 per month in the former and
TSh.215,262 in the latter.
11
Because the price indices computed for the
However, nominal wage increases in the poverty analysis were only done for the 2007-2009
public sector have lagged behind the private period covered by the HBS and NPS, there is no
sector slightly, particularly in rural areas. attempt to present real wage changes here, as the
Median nominal wages in the public and price series used in this report is not available for
private sector rose by an annual rate of 24.4 2006.
28
Box 2. Comparing labour market indicators over time
Box 1 addressed the challenges of estimating divide workers between jobs and sectors
changes over time in poverty – defined by low differ between surveys. The analysis
levels of household consumption. To presented here groups workers into a small
understand why poverty is or is not falling, it list of broad categories, to ensure that the
is equally important is to examine trends in groupings are as consistent as possible across
household income. This box examines labour years. 12
market earnings, from
both wage employment Sector of employment
From 1991 to 2009, the fastest Figure 4 divides all adults,
and non-farm self-
growing sector of employment has age 15 to 65, by their
employment. (Farm
been non-farm, self-employment in main occupation: (i)
production is dealt with in
urban areas. The share of adults in farming, (ii) public-sector
the next section.)
wage employment, (iii)
farming and wage employment has
How do the wage and self- private-sector wage
not changed dramatically over
employment earnings employment, (iv) non-
time. Real earnings for both wage-
levels recorded in the NPS agricultural self-
compare to earlier earners and the self-employed employment, (v) or none
surveys? Have significant grew rapidly during the 1990s, with of the above.13 The top
changes taken place in the panel shows this
distribution of the labour force across sectors breakdown for Dar es Salaam, the middle
in recent years? This box attempts create panel for other urban areas on the Mainland,
trends in earnings and employment levels and the bottom panel for rural Mainland
over the eighteen years from 1991 to 2009, by Tanzania. The figure includes data from six
comparing three sets of surveys: the HBS points in time covered by the HBS, ILFS and
(1991, 2000, 2007), the ILFS (2001, 2006), and NPS, spanning the period 1991 to 2009.
the NPS (2008/09). The figures presented
12
here are based on new analysis of the micro- See Appendix C for a detailed comparison of the
data from each survey, attempting to align question phrasing used for occupational
definitions and categories, so they may differ categories, wages, and self-employed earnings in
from earlier official reports. the HBS, ILFS, and NPS.
13
Readers should note that key labour market The public sector here includes both local and
central government employees, as well as
questions are asked in different ways in each
individuals working for religious organizations,
set of surveys. Thus direct comparisons of the
political parties, NGOs or international
HBS, ILFS and NPS labour data must be organizations. The private sector includes
treated with extreme caution. In particular, parastatal enterprises as well. “Farming” includes
while the ILFS and NPS ask relatively similar crop-farming as well as livestock keeping and
questions, based on current activities in the forestry. “None” includes individuals who are not
past seven days and earnings in the most in the labour force (including the disabled and
recent payment period, the HBS asks students), the unemployed and unpaid family
questions over a 12-month recall period. workers. Precise definitions for each category in
Similarly, the occupation categories used to each survey are listed in Appendix XXX.
29
Figure 5. of
In terms Comparing employment
the comparability of the levels by
different Three key trends in employment levels seen in
occupation andthe
data sources, strata acrosssuggest
estimates 6 surveys.
that the Figure 4 include:
ILFS data - relative to both the HBS and the
NPS - finds fewershares
people over
out oftime
work in all First, there has been a sustained increase in
Employment
geographic strata. This may be due the ILFS's the share of the population involved in non-
Dar es Salaam
exclusive, detailed focus on labour issues, agricultural self-employment, particularly in
100 urban areas. This rate rises fairly steadily
enabling the survey to tease out activities that
the HBS or NPS miss. Beyond the jump up in from 18.4% in Dar es Salaam in 1991 to 29.7%
80
overall employment in the ILFS rounds in the NPS in 2008/09; similarly in other urban
however, the different surveys appear to areas the rate rises 13.8% to 24.1% over the
60
same period.
present fairly consistent trends.
40
Second, the share of the workforce in
20
agriculture has declined significantly in urban
areas outside Dar es Salaam (from 42.8% in
0 1991 to just 29.2% in 2008/09) but only
91-HBS . 00-HBS 06-ILFS 09-NPS
01-ILFS .. 07-HBS modestly in rural areas (from 83.1% to 80.0%
over the same period). However, this share is
Other Urban
highly volatile across surveys, likely reflecting
100
differences in measurement techniques as
much as genuine economic fluctuations.
% of Adults 15-65 Yrs Old
80
60
40
20
0
01-ILFS .. 07-HBS
Rural
100
80
60
40
20
0
01-ILFS .. 07-HBS
None Farm Wage-Pub W age-Pri Self
30
of the occupation categories
Figure 1. 6.
Figure Comparing earnings
Comparing by by
earnings occupation across
occupation 6 6
across analyzed above: public sector
surveys, 1991-2009
surveys, 1991-2009 wages, private sector wages,
and non-agricultural self-
employed earnings (farm
Real
Realmedian
medianmonthly
monthlyearnings
earnings income is not included here).
Once again, individuals are
Self
Self Wage-Priv
Wage-Priv W age-Pub
W age-Pub grouped by their main
120,000
Real 2007 Shillings per month
120,000
Real 2007 Shillings per month
occupation, and earnings are

100,000
100,000 presented only for that main
occupation category.
80,000
80,000
Once again, three key trends
60,000
60,000 in real wage levels stand out in
40,000
40,000 Figure 5:
20,000
20,000 First, during the 1990s,
median real earnings grew
1991-HBS
2000-HBS
2001-ILFS
2006-ILFS
2007-HBS
2009-NPS
steadily across all occupation
1991-HBS
2000-HBS
2001-ILFS
2006-ILFS
2007-HBS
2009-NPS
categories, and earnings
inequality between sectors
grew much wider. Real
earnings grew at a rate of
5.9% for the self-employed,
Third, the relative size of the wage- 8.7% for private-sector wage employees, and
employment sector (both public and private) 12.9% for public-sector wage employees.
has not grown dramatically over the period
Second, in the 2000s, median real earnings for
covered here. The public sector has
wage workers (both public and private) have
employed just over 20% of the adult
stagnated. Median earnings in the public and
population in Dar es Salaam throughout this
private sectors were Tsh. 115,441/- and TSh.
period, at or below 10% in other urban areas,
44,770/- respectively in the 2001 ILFS, and
and approximately 2% in rural areas.
had fallen slightly to TSh. 110,071/- and TSh.
Meanwhile, the private sector has employed
41,764/- in the 2008/09 NPS.
roughly 5% of adults in Dar es Salaam and
other urban areas, and between 1 to 2% in Third, earnings of the non-farm self-employed
rural areas. (The private sector employment appear to have surpassed private-sector
figures in Dar es Salaam have been wages during the 2000s. Earnings in the HBS
particularly volatile, fluctuating between 2.9% years are consistently much lower than the
and 8.7%, but show no sustained trend ILFS and NPS rounds. However, a general
upward or downward.) trend is discernible of increased earnings for
the self-employed despite declining private-
sector wages. Interestingly, these increased
Earnings levels earnings for the self-employed come while
Figure 5 examines median, monthly, real
earnings in 2007 Tanzanian Shillings in three
31
the share of workforce in self-employment is
increasing steadily in urban areas.
Table 8. Share of adults (age 15-65) by sector of employment, 1991-2009
Self- Private Wage Public Wage
Year Farming Employed Employee Employee None
Dar es Salaam 1991-HBS 2.3% 18.5% 22.1% 8.7% 48.4%
2000-HBS 3.6% 23.2% 17.3% 5.3% 50.6%
2001-ILFS 4.8% 27.6% 20.4% 2.9% 44.3%
2006-ILFS 8.4% 31.3% 19.7% 5.3% 35.4%
2007-HBS 4.0% 29.4% 20.7% 6.1% 39.7%
2009-NPS 1.7% 29.7% 21.8% 5.4% 41.5%
32
Goal 4: Reducing income poverty of
both men and women in rural areas
As noted in the introduction to the report, the The overall picture which emerges from the
NPS contains a wealth of information on NPS agriculture data is unsurprising: the vast
smallholder farming activities. A separate majority of smallholders rely on traditional
one-and-half hour interview was conducted farming techniques and employ relatively few
specific to agriculture, livestock and fishing new ‘modern’ technologies such as hybrid
activities for all households engaged in these seeds, inorganic fertilizer, etc.
sectors. As such, the NPS has the potential to
track virtually all of the agriculture-related However, these results are relatively
MKUKUTA indicators, as well as various other uninformative without a benchmark for
indicators related to the Agricultural Sector comparison. In future years the NPS will
Development Program (ASDP) and the provide this time series and show the
emerging Kilimo Kwanza goals. evolution of technology adoption, crop
choice, farm output and other indicators over
time. In this baseline year
Table 9. Production of Major Crops (in tons), Ag Census however, an attempt is made
2002/03 and NPS 2008/09 below to put the NPS figures
in context by making direct
2002/03 2008/09 %
NSCA NPS Change comparisons, wherever
Masika only (2002, 2008) possible, to the results of the
Maize 2,004,388 2,628,430 31.1% 2003 National Sample Census
Paddy 513,652 692,506 34.8% of Agricultural, referred to
Beans 207,571 152,175 -26.7% here as the “Agricultural
Sorghum 173,174 242,426 40.0% Census”.
Groundnuts 138,847 407,515 193.5%
Sweet potatoes 131,123 291,840 122.6% By linking these two datasets,
Sunflower 52,925 151,767 186.8% it is possible to see how some
Full year of the NPS key indicators have
Maize 2,617,115 2,993,055 14.4% developed over time. These
Cassava 2,102,838 1,159,987 -44.8%
include crop output numbers,
Paddy 604,978 956,767 58.1%
asset holdings and various
Mango 336,028 929,099 176.5%
Beans 333,312 217,733 -34.7% indicators of farming
Sweet potatoes 216,478 385,755 78.2% technologies. As the
Sorghum 216,435 251,989 16.4% Agricultural Census and the
Orange 186,695 345,043 84.8% NPS use different
Cashew 183,419 73,066 -60.2% questionnaires, it is often the
Groundnuts 160,257 434,273 171.0% case that the indicators
Coconut 102,458 348,747 240.4% produced from the two will
Coffee 61,602 69,417 12.7%
differ qualitatively on various
Sunflower 55,070 153,799 179.3%
points. Below is a description
33
Table 10. % of households with erosion 2008. However, respondents interviewed
problems during roughly the first half of fieldwork were
also asked about farming during the short
Ag
rainy season in 2007-08, while those
Census NPS
Having Erosion Problem 23.8% interviewed in the second half were asked
Using Erosion Control 10.0% 25.6% about the 2008-09 short rainy season. As a
Type of Erosion Control result, the crop year covered by the NPS
Terraces 13.3% 43.2% differs by respondent. The most consistent
Erosion Control Bunds 53.0% 36.5% figures are those restricted to the long rainy-
of how the NPS indicators are produced and season only.
what are the limitations in terms of
comparability to the Ag Census numbers. Table 9 reports total crop production for
major food crops in the 2002/03 Ag Census
and the 2008/09 NPS. The figures for the long
% change in food crop rainy season cover 2002 and 2008. The
production figures are nationally representative, including
The NPS collects data on production of all Zanzibar, but include only smallholders. All
major crops in a way that is intended to be production figures are measured in metric
comparable to the 2002/03 Ag Census. tons.
However, the Ag Census sample is much As seen in the table, the masika harvest of
larger than the NPS sample so some
major cereal crops – maize, rice paddy and
differences in the figures derived from these sorghum – increased moderately between the
surveys may be due to the greater sampling
two surveys: up by 31.1% for maize, up 34.8%
error in the NPS. The timing of the questions
for paddy, and up 40% for sorghum. These
during the crop cycle also differed between increases represent annual percentage
the surveys. growth rates of between 4.6% and 5.8% over
In the 2002/03 Ag Census, all respondents six years. Also noteworthy is the change in
were asked about the 2002 long rainy season cassava production. This figure is calculated
(masika) and the 2002/03 short rainy season on an annual basis, and declined by 44.8%
(vuli). In contrast, the NPS fieldwork began in when comparing the 2002/03 agricultural
October 2008 and ended in October 2009. year to the full calendar year covered by the
Thus all respondents were asked about their 2008/09 NPS.
farming activities for the long rainy season
34
% of small holders using methods and water sources, pesticide and
fertilizer type used, etc. In addition to this,
modern methods of farming
there is information on the fraction of all
(irrigation, fertilizers and households growing
hybrid seeds) annual crops using
The statistics on modern Preliminary comparisons with the improved seeds. The
farming methods include information on improved
2002/03 agricultural census show
the fractions of all seeds is only available in
that use of key modern
households cultivating at the section on annual
technologies by smallholders has
least one plot of land who crops, and so any usage of
are using erosion
remained remarkably constant improved seeds for
controls, irrigation, over the past six years: roughly permanent crops/fruits is
organic and inorganic 12% use any inorganic fertilizer not recognized.
fertilizers, and pesticides. and just less than 20% use any
With regard to organic
There is also more improved seeds.
fertilizer, the closest
detailed information on
reference in the Ag census is on the usage of
type of erosion controls, irrigation types,
farm yard manure as a fertilizer. As usage of
compost as fertilizer is reported separately, it
is not possible to see from the reported
Table 11. % of households with irrigation statistics what fraction are using at least one
Ag of them, which would have been a closer
Census NPS match to the NPS version of the question.
Number of HH Using
Irrigation 240,721 276,958 For the types of erosion controls used, the Ag
Percent Using Irrigation 4.7% Census reports the percentage of the
Type of Irrigation: structures that belong in different categories.
Flooding 56% 66.0% While the categories match well between the
Sprinkler 2% 3.3% Ag Census and the NPS, the latter only contain
Drip Irrigation n/a 3.2% information on whether a structure exists, so
Bucket/Watering Can 41% 24.6% it is only possible to construct statistics on the
Water Hose 1% 5.2% fraction of the households using a specific
Other n/a 0.0% control structure.
Method of Obtaining Water
Gravity 58% 73.0% For the irrigation water source statistics, the
Hand Bucket 39% 24.5% Ag Census and NPS categories do not perfectly
Hand/Foot Pump 1% 2.7% match align; non-matching codes have been
Motor Pump 1% 10.2% re-assigned as follows: NPS references to
Other 1% 1.3% “pond/tank” were matched to ag census
Source of Water references to “dam”, NPS responses stating
Well 15% 14.3%
“river/stream” to ag census responses of
Borehole 1% 12.3%
“canal or river” and NPS “other” responses to
Pond/Tank 3% 2.8%
ag census “lake or pipe water” responses.
River/Stream 75% 73.6%
Other 4% 5.1% Ownership of farm implements
35
The NPS includes detailed information on the Table 12. Use of fertilizers and pesticides/
farm implements owned by a household. The herbicides
indicators based on this information are Ag
presented as the fraction of the households Census NPS
owning a specific implement (such as a hand Using Organic Fertilizer 26% 19.2%
hoe, a tractor, etc.). It is important to note Using Inorganic Fertilizer 12% 11.6%
that this refers to the fraction of the total Type of Inorganic Fertilizer
number of households involved in agriculture, Di-Ammonium Phosphate 13.1%
i.e. either owns or cultivates some land or (DAP)
keeps some livestock. UREA
following sources were not included in65.8%
the Ag
Census: Radio, Publication and Neighbor.
Extension services
The NPS contains information on whether Finally, additional agricultural indicators that
households have received any extension can be extracted from the NPS but which have
services regarding agricultural activities or not been analyzed to date include: the % of
prices, and from what source this extension small holders who accessed formal credit
came (e.g. the government, an NGO or the formal credits for agricultural purpose; % of
radio). As in the case of farm implements, the small holder households who have one or
indicators presented are expressed as the more off-farm income generating activities;
percentage of people receiving an extension and % of households whose main income is
from a specific source. Note that also in this derived from the harvesting, processing and
case do the percentage indicators refer to the marketing of natural resources products.
fraction of people involved in agriculture that
received some extension.
Table 13. Use of improved seeds

Ag
Census NPS
Using Improved Seeds 18.0% 19.5%
Type of Improved Seeds
Certified Seeds 86.2%
Quality Declared
Seeds 17.5%
Table 14. Use of modern farming methods

When comparing this measure to the Ag by gender of household head
census measure, there is a discrepancy
between the interview questions. While the Male- Female-
headed headed
Ag census indicator refers to “crop extension Using Irrigation 4.5% 4.3%
services” the NPS refers to both “advice for Using Erosion Control 24.4% 24.3%
your agricultural/livestock activities” and Using Organic Fertilizer 19.7% 12.6%
“agricultural prices”. In addition to this, the Using Inorganic Fertilizer 11.8% 9.7%
NPS explicitly asks about extension received Using Pesticide/Herbicide 12.6% 8.8%
via more sources than did the Ag Census. The Using Improved Seeds 20.4% 14.9%
36
Table 16. % of households that received Table 17. Ownership of farm implements
extensions
Table 15. Possession of land titles among farming(multiple responses
households, allowed)
by gender of household head
Ag
Census NPS All Farm Ag
Male- Female-
Received Any Kind of Households Census
headed NPS
headed
Having a title for at least one plot
Extension 34% 58.8% Hand Hoe 11.1% 12.0%
97.7% 98.9%8.1%
Type of Title Extension
Received (Multiple Responses
From Allowed) Hand-Powered Sprayer 14% 6.1%
(Share of
Granted All of
Right Households)
Occupancy Ox Plough 18.7% 18.1%
23.0% 22.0%
8.0%
Government
Cert. Of CustomaryExtension 32.3% 21.1%
Right of Occupancy 10.2%
Ox Seed Planter 11.2% 9.1%5.5%
NGO
Residential License 1.0% 2.2% Ox Cart 0.5% 0.5%
5.0% 0.1%1.0%
Cooperative/Farmer's
Village-Government-Witnessed Purchase AgreementTractor 25.6% 28.3%
3.0% 12.3%
2.2%
Association
Local-Court-Certified Purchase0.3% 6.0%
Agreement 2.4% 1.7%
Tractor Plough 2% 0.1%5.4%
Large Scale Farmer 0.3% 2.4%
Tractor Harrow 0.4%
Radio n/a 27.1%
Sheller/Thresher 2% 0.1%
Publication n/a 4.6%
Hand Mill/Grinder 0.4%
Neighbour n/a 37.4%
Watering Can 2.0%
Other 0.0% 9.2%
Farm Buildings 6.4%
Geri Cans/Drums 10.5%
Goal 6: Provision of reliable and

affordable energy to consumers
37
% of households in rural and
% increase in number of urban areas using alternative
customers connected to the sources of energy to wood fuel
national grid and off-grid (including charcoal) as their
sources of electricity main source of energy for
Comparison with the Household Budget cooking.
Survey series provides a picture of the Table 19 shows trends across the HBS and
evolution of electricity usage by Tanzanian NPS in terms of fuel-type for cooking, broken
household over time. Table 18 shows the down by geographic strata. Looking across
proportion of households in each geographic strata, the overwhelming majority of
stratum of the survey who have access to (a) Tanzanians in Mainland rural areas (and in
electricity through the grid or (b) solar power. Zanzibar) rely on farm residuals for cooking
The results show a significant increase in Dar fuel. In urban areas, and in particular in Dar
es Salaam (from 55% in 2007 to 62.4% in es Salaam however, the primary cooking fuel
2008/09). Urban areas show a much smaller is charcoal.
increase (from 25.9% to 27.2%) and in rural
The subsequent rounds of the HBS have
areas – where electricity access is extremely
shown a steady increase in the proportion of
sparse – there was a slight decline in the
households reporting reliance on charcoal.
estimated figures, from 2.5% to 2.3%. While
Results here suggest that this trend may be
no time-series comparison is presented here,
stabilizing – as charcoal usage on the
the overall rate of electricity access in
Mainland as a whole fell slightly from 22.7 to
Zanzibar is roughly equivalent to urban areas
20.2% from 2007 to 2008/09 – though more
outside Dar es Salaam on the mainland at
data from future rounds is necessary to
23.5%/ In all strata the usage of solar power
determine whether this is a permanent
has been and remains extremely low.
change in the trend.
Table 18. % increase in number of

customers connected to the national grid
and off-grid sources of electricity
Electricity Solar
grid
Dar es 2001 58.9 1.3
Salaam
2007 55.0 0.7
2008/09 59.3 0.1
Table 19. Main fuel used for cooking (%) by stratum and year
Elect. Gas Biogas Paraffin Char- Wood Other
coal
Dar es 91/92 9.7 1.2 . 33.7 52.1 1.2 2.1
Salaam 2001 4.8 0.4 0.2 43.0 46.2 4.6 0.9
2007 2.2 0.9 0.1 12.4 74.9 8.0 1.5
08/09 1.5 3.1 0.0 11.5 78.4 3.1 2.3
Other Urban 91/92 4.8 0.6 . 13.3 36.6 43.4 1.3
39
CLUSTER 2:
IMPROVEMENT OF QUALITY OF LIFE
AND SOCIAL WELL-BEING
In addition to collecting a wide-variety of information on recourse to public versus
household characteristics, the NPS private facilities for service delivery, money
questionnaire also includes several detailed spent on health and education for each
modules which are administered at the member of the household, and so on. This
individual level. These provide disaggregated extensive individual-level data is central to
information on personal demographics, tracking progress on various indicators under
education, health, and nutrition. The survey Cluster 2.
combines not only outcome indicators – e.g.,
height and weight for children, years of
education achieved, etc. – but also provides
Indicator NPS
coverage
Goal 1: Ensure equitable access to quality primary and secondary education for boys and girls,
universal literacy and expansion of higher, technical and vocational education
Literacy rate of population aged 15+ No
Net enrolment at pre-primary level Yes
Net primary school enrolment rate Yes
% of cohort completing Standard VII Yes*
% of students passing the Primary School Leavers’ Exam No
Pupil/Teacher ratio No
% of teachers with relevant qualifications No
Pupil/text book ratio No
Transition rate from Standard VII to Form 1 Yes*
Net secondary enrolment Yes
% of students passing the form four examination No
Enrolment in higher education Institutions No
Goal 2: Improved survival, health and well-being of all children and women and especially
vulnerable groups
Infant mortality rate Yes
Under-five mortality rate No
% change in mortality attributable to malaria among children under-five No
DPTHb3 coverage No
Proportion of under-fives moderately or severely stunted (height for age) Yes
Maternal mortality ratio No
40
Indicator NPS
coverage
Proportion of births attended by a skilled health worker Yes
% of persons with advanced HIV infection receiving ARV combination therapy No
HIV prevalence amongst 15 - 24 yrs No
TB treatment completion rate No
Goal 3: Increased access to clean, affordable and safe water, sanitation, decent shelter and a safe
and sustainable environment
Proportion of population with access to piped or protected water as their main Yes
No. of reported cholera cases No
% of households with basic sanitation facilities Yes
% of schools having adequate sanitation facilities (as per Policy) No
Total area under community based natural resources management No
Goal 4: Adequate social protection and rights of the vulnerable and needy groups with basic needs
and services
Goal 5: Systems are in place to ensure effective universal access to quality public services that are
affordable and available
Proportion of children in child labour Yes
Proportion of children with disability attending primary school Yes
Proportion of orphaned children attending primary school Yes
Proportion of elderly accessing medical exemptions at public health facilities Yes*
Proportion of population reporting to be satisfied with health services No
*Indicators covered by the NPS questionnaire but which are not yet covered in this report.
41
Goal 1: Ensure equitable access to
quality primary and secondary
education for boys and girls, universal
literacy and expansion of higher,
technical and vocational education.
2002, the NER rose quite dramatically, as
documented in shift between the 2000/01
Net primary school enrolment and 2007 rounds of the HBS. The total NER on
rate the mainland rose from 58.7 to 83.7% over
The net enrolment rate (NER) at the primary this period, with similar increases for both
level is defined as the share of children age 7 boys and girls.
to 13 who are enrolled. Table 21 reports the
In the short period between the2007 HBS and
primary school NER for boys and girls in each
the 2008/09 NPS the table shows that there
geographic stratum of the survey sample,
has been a slight reversal of this trend. The
together with comparable figures from the
decline in enrolment is most pronounced in
HBS.
Dar es Salaam, where the primary NER fell
After the elimination of primary school fees in from 90.8% to 81.8%. However, enrolment
was relatively stable in rural areas, which
drives the national results. (Also, it is worth
Table 21. Net primary school enrolment noting that the largest changes are witnessed
rates, by year, sex and stratum in the urban strata with the smallest sample
and hence largest potential sampling errors in
All Boys Girls the NPS.)
Dar es 2000/01 71 68.3 73.4
Salaam 2007 90.8 91 90.7
2008/09 81.2 82.8 79.5 Net secondary enrolment
Other 2000/01 71.4 72 70.7
urban 2007 91.3 91 91.6 Following the practice in the HBS reports, net
areas 2008/09 88.3 84.8 91.5 enrolment rates for secondary school are
Rural areas 2000/01 56 53.9 58.4 divided between the rate for Forms 1-IV,
which uses the population of individuals age
2007 81.5 78.7 84.4
14 to 17 in the denominator, and the NER for
2008/09 81.1 77.3 84.6
Forms I-VI, which uses the population of
Mainland 2000/01 58.7 56.7 60.8
Tanzania individuals age 14 to 19 in the denominator.
2007 83.7 81.4 86.1
2008/09 82.3 78.9 85.5 Table 22 shows the results for these
Zanzibar 2008/09 78.9 76.0 81.7 secondary school indicators. The general
URT 2008/09 81.9 78.6 85.0 pattern, unlike at the primary level, is of a
42
continued rapid rise in enrolment. Across Table 22. Net secondary school
mainland Tanzania the proportion of enrolment rates, by year, sex and stratum
secondary-school aged children enrolled in
Forms I-IV: All Boys Girls
Forms 1-4 rose from just 5.1% in 2000/01 to
Dar es 2000/01 18.9 17.2 20.4
15.2% in 2007 and to 22.1% in 2008/09.
Salaam 2007 31.5 33.7 29.8
Rural areas continue to lag behind urban 2008/09 44.9 51.3 39.0
areas, unsurprisingly, with a secondary-school Other 2000/01 15.2 12.7 17.1
NER of 14.9% compared to 46.8% in urban urban 2007 28.1 25.7 30.2
areas 2008/09 46.4 44.7 47.9
areas outside Dar es Salaam. Indeed, girls in
rural areas are the one group shown in the Rural areas 2000/01 2 1.5 2.5
Table with little or no improvement in 2007 10.4 9.7 11.2
secondary NER over the past year and a half. 2008/09 15.6 16.8 14.2
Mainland 2000/01 5.1 4 6.1
Across the board, gender differentials remain Tanzania 2007 15.2 13.9 16.5
considerable. Contrary to the situation in 2008/09 23.5 23.8 23.2
primary school, boys are now more likely to Zanzibar 2008/09 37.7 18.9 55.7
be enrolled in secondary school than girls. URT 2008/09 25.2 23.2 27.3
This phenomenon is new, as increased Forms I-VI
enrolment by boys has pushed the male Dar es 2000/01 16 14.5 17.4
secondary NER above the female rate only Salaam 2007 31.7 35.8 28.5
within the past year and half across Mainland 2008/09 44.8 54.5 36.1
Tanzania. Other 2000/01 15.3 14.9 15.6
urban 2007 29.2 29 29.4
areas 2008/09 44.9 46.0 43.8
Rural areas 2000/01 2.3 1.8 2.8
2007 11.4 11.7 11.1
2008/09 18.3 18.7 17.8
Mainland 2000/01 5.3 4.6 5.9
Tanzania 2007 16.4 16.5 16.3
2008/09 25.8 26.6 24.9
Zanzibar 2008/09 39.7 28.3 50.6
URT 2008/09 27.6 26.8 28.4
43
Goal 2: Improved survival, health and
well-being of all children and women
and especially vulnerable groups
on these latter indicators are generally
interpreted as an indication of short-term
Proportion of under-fives malnutrition, while low height-for-age is seen
moderately or severely stunted as a measure of the cumulative effects of
(height for age) chronic malnutrition.
Height-for-age provides an indication of
The analysis here was conducted in Epi-Info
children’s nutritional status. While the height
based on the sex-specific 1978 CDC/WHO
of individual children may vary for a variety of
reasons, including genetic differences, the
existence of a large proportion of children
Figure 7. Stunting: Proportion of children
with extremely low height-for-age in a given
under 5-years moderately and severely
population is a strong indication of food stunted, by stratum
security deficiencies.
“Moderate stunting” measures the proportion Moderate Severe

of children whose height is more than two 0 10 20 30 40
standard deviations below the mean height 41.5
for their age in an international reference Rural
15.6
population. Similarly, “severe stunting”
measures the proportion of children more 21.4
than three standard deviations below the Urban
5.4
average for their reference group.
35.1
The NPS collected anthropometric data on all DSM
11.2
members of the sampled households, from
age zero upwards. However, the data
24.1
analyzed here is restricted to children under Zanzibar
9.1
five, in line with the scope of the MKUKUTA
indicator.
To complement the data on stunting, this normalized version of the 1977 NCHS
section also reports the proportion of children reference curves for height-for-age, weight-
who are moderately or severely wasted (low
weight-for-height) and moderately or severely
underweight (low weight for age). Low scores
44
Figure 8. Cumulative distributions of Height- most and right-most vertical lines indicate the
for-Age Z-Scores by stratum share of children moderately and severely
stunted, respectively. What is perhaps
somewhat surprising about this graph is the
similarity of the distribution of height –
particularly at the lowest levels – between Dar
es Salaam and rural areas.
for-age, and weight-for-height.14 The

reference curves provide the basis for the
calculation of Z-scores (percentile ranks on a
standardized normal distribution of
anthropometrics in the reference population)
which underlie the classification of children as
moderately or severely stunted, wasted, or
underweight.
Figure 7 gives a more precise look at the

distribution of height underlying the simple
proportions reported in stunting and wasting
statistics. The figure shows the cumulative
Table 23. Proportion of children under 5 years-old with low height-for-age (stunted),
weight-for-height (wasted), or weight-for age (underweight), %
Severely Moderately Severely Moderately Severely Moderately
stunted Stunted wasted wasted underweight underweight
Total: 13.4 36.9 0.8 3.0 3.8 21.0
By age: (months)
0 to 6 7.0 21.5 3.2 4.6 2.8 5.8
7 to 11 14.6 35.8 1.7 4.5 3.6 17.3
12 to 23 17.7 46.3 0.8 3.7 4.8 25.5
distribution of the height-for-age Z-scores
relative to the 1978 CDC/WHO reference
population. The two vertical lines are at two
and three standard deviations below the
average. The points (on the y-axis) at which
the sample distribution intersects the left-
14
Dibley MJ, Goldsby JB, Staehling NW,
Trowbridge FL. Development of normalized curves
for the international growth reference: historical
and technical considerations. Am JClinNutr
1987;46:736-48
45
Proportion of births attended
by a skilled health worker
Among women who reported giving birth to a
child in the last 24 months, 47.3% reported
giving birth to their most recent child in a
hospital, 7.6% in a clinic, 43.5% at home, and
17.8% elsewhere.
Table 24 shows the proportion of births

attended by various service providers.
Nationwide, 60.8% of births were attended by
a skilled health worker – either a doctor,
nurse or midwife – and that rate was above
90% in both Dar es Salaam and other urban
areas.
Table 24. Proportion of births attended by various service providers

Traditional Birth
Friend/Relative
Total Unskilled
Doctor/Clinical
Total Skilled
Attendant
Midwife
Officer
Nurse
Other
Self
Total 15.8 32.2 12.8 60.8 15.7 18.7 3.9 0.9 39.2
Dar es Salaam 35.7 54.0 5.4 95.1 3.4 1.5 0.0 0.0 4.9
Urban 34.9 44.2 12.3 91.5 2.1 4.9 1.6 0.0 8.5
Rural 11.3 28.6 12.8 52.8 15.8 24.9 5.2 1.3 47.2
Zanzibar
12.0 30.2 17.8 60.0 37.8 2.2 0.0 0.0 40.0
46
Goal 3: Increased access to clean,
affordable and safe water, sanitation,
decent shelter and a safe and
sustainable environment
Looking across time, the overall picture is one
of continuity rather than change. The overall
Proportion of population with proportion of households with access to piped
access to piped or protected water has been relatively stable since the
water as their main drinking 1991/92 HBS. The notable exception to this
trend is Dar es Salaam, where access to piped
water source
water has actually declined, from 93.1% in
Table 25 reports the primary source of
1991/92 to the current rate of between 72.8
drinking water for households in the each
and 76.3%. This decline appears to be related
geographic stratum of the survey. An attempt
to the phenomenon of rural to urban
is made to provide a direct comparison to
migration. Access to piped water has declined
figures from the multiple rounds of the HBS.
at least marginally in all other Mainland strata
However, one important difference between
as well, while the Mainland average has
the HBS and the NPS relates to the reference
remained constant – reflecting the larger
period of the question: whereas the HBS asks
proportion of households in urban areas,
about a household’s primary source of
including Dar es Salaam, in later survey
drinking water in general, the NPS asks
rounds. In short, it appears that Tanzanian
separate questions for the ‘rainy’ and ‘dry’
households have maintained access to
seasons.
improved sources of drinking water by moving
Within the NPS, sources of drinking water are to water infrastructure (i.e., cities) rather than
relatively constant across seasons. On the the water infrastructure moving to them.
mainland as a whole, between 34 and 35% of
households have access to piped water of
some sort in each season. The rate is
significantly lower, roughly 22% for the
majority of Tanzanians in rural areas.
Table 25. Primary source of drinking water (%), by stratum and year/season
91/92 2001 2007 08/09 08/'09
Rainy Dry
Dar es Salaam Any piped water 93.1 85.8 61.5 72.1 76.1
Piped water inside dwelling 22.1 13.7 8.0 11.5 11.1
Private outside standpipe 52.6 19.1 11.8 25.0 25.8
tap
Piped water neighbor . 46.4 37.6 27.7 29.1
Public standpipe tap 18.4 6.6 4.1 2.4 2.9
91/92 2001 2007 08/09 08/'09
Rainy Dry
Mainland Any piped water 35.9 39.3 33.9 34.7 35.4
Tanzania Piped water inside dwelling 5.2 3.8 3.8 4.1 3.9
Private outside standpipe tap 9.2 5.5 4.8 6.4 5.7
Piped water neighbor . 10.2 10.1 9.3 9.7
Public standpipe tap 21.5 19.8 15.2 13.7 13.6
Any well 34.0 34.3 38.3 28.5 31.6
Rainwater . . 0.7 8.9 0.4
% of households with basic
sanitation facilities
Table 26 shows households’ access to various increased steadily over time, to a current rate
types of toilet facilities, by stratum and year. of 18.1% in Dar es Salaam and 13.7% in other
Across all strata, the vast majority of urban areas. However, the large jump
households have access only to pit latrines – upward in VIP access in just roughly 18
80.9% report using a pit latrine on the months in Dar es Salaam suggest something
Mainland in the NPS. In urban areas, access of a statistical fluke here, and results should
to ventilated improved pit latrines (VIP) has be treated with caution.
Table 26. Toilet type (%), by stratum and year

No toilet Flush toilet Pit latrine VIP Other toilet
Dar es Salaam 1991/92 1.3 9.3 89.1 0.2 0.1
2001 5.7 10.6 82.0 1.7 0.0
2007 1.1 10.3 80.5 7.8 0.2
2008/09 0.1 18.9 62.4 18.4 0.0
Other Urban 1991/92 1.8 3.4 94.6 0.2 0.0
2001 2.3 7.5 87.5 2.5 0.2
2007 2.7 5.6 78.6 12.9 0.1
49
Table 27. Household tenure status by stratum and year
Owned by the Occupying Rented Subsidized
household w/out by employer
paying rent
Dar es Salaam 1991/92 31.2 3.8 62.3 2.7
2001 32.4 3.6 62.7 1.2
2007 38.8 5.3 55.2 0.6
2008/09 39.0 6.4 54.3 0.2
Other urban 1991/92 55.8 2.3 41.0 0.8
2001 54.2 4.1 40.5 1.1
2007 55.0 4.6 39.9 0.5
2008/09 51.6 6.7 39.7 0.4
Rural 1991/92 95.3 1.6 2.7 0.3
2001 94.5 2.3 2.9 0.2
2007 92.3 3.1 4.5 0.0
2008/09 91.3 5.5 3.1 0.0
Total Mainland 1991/92 85.4 1.8 12.1 0.5
2001 84.3 2.7 12.5 0.4
2007 79.9 3.6 16.3 0.2
2008/09 78.3 5.9 15.4 0.1
Zanzibar 2008/09 86.0 10.4 3.6 0.0
URT 2008/09 79.2 6.4 14.0 0.1
50
Table 28. Building materials (%), by stratum and year
Dar es Salaam Other Urban Rural Mainland Tanzania Zanz URT
91/92 2001 2007 08/09 91/92 2001 2007 08/09 91/92 2001 2007 08/09 91/92 2001 2007 08/09 08/09 08/09
Floor:
Earth 14.5 6.7 8.7 5.6 44.6 38.3 37.1 33.0 90.8 86.6 83.1 83.3 79.2 74.0 67.0 65.3 30.1 61.3
Cement, tiles 84.3 92.4 90.4 94.2 54.2 61.1 61.9 66.8 8.0 12.5 15.6 15.2 19.6 25.2 31.8 33.6 69.4 37.7
Other 1.2 0.9 1.0 0.0 1.2 0.5 0.9 0.2 1.2 0.9 1.3 1.5 1.2 0.8 1.2 1.1 0.5 1.0
Walls:
Poles, branches, 3.4 0.9 1.5 0.8 5.7 5.3 4.6 3.6 23.7 19.3 16.9 12.4 19.8 16.0 13.0 9.5 1.8 8.6
etc.
Mud & poles / 15.1 5.2 4.7 2.7 16.3 13.1 10.9 18.1 27.7 21.8 22.0 26.4 25.3 19.4 18.2 22.4 49.9 25.6
stones
Mud only 2.0 2.2 1.9 0.5 11.1 12.1 10.3 1.6 14.6 18.1 12.0 4.6 13.3 16.1 10.7 3.5 0.5 3.2
Mud bricks 12.0 3.2 1.3 0.2 37.6 30.8 22.6 30.7 24.2 23.5 26.4 30.3 25.4 23.3 23.2 27.6 2.3 24.6
Baked / burnt 4.8 1.3 1.6 0.2 11.9 15.9 29.9 23.3 8.1 13.7 18.8 22.3 8.5 13.2 19.3 20.5 0.3 18.1
bricks
Concrete, cement, 62.1 87.2 88.3 95.0 17.1 22.4 20.7 22.7 1.5 3.0 3.1 3.5 7.6 11.5 14.8 16.1 41.0 18.9
stone
Other 0.7 0.0 0.5 0.5 0.2 0.4 1.0 0.1 0.1 0.6 0.9 0.6 0.2 0.5 0.9 0.5 4.3 0.9
Roof:
Grass, leaves, 1.1 1.1 2.1 1.5 21.7 14.3 12.3 13.0 63.1 55.7 48.2 46.1 53.1 45.8 36.8 34.9 24.1 33.6
bamboo
Mud & grass 0.2 0.7 0.4 0.2 1.7 1.5 2.6 0.8 12.8 12.5 9.2 5.6 10.4 10.1 7.1 4.1 0.8 3.7
Concrete, cement 3.4 3.6 1.2 1.7 0.7 0.5 0.0 0.4 0.1 0.0 0.0 0.1 0.4 0.3 0.1 0.3 0.3 0.3
Galvanised metal 91.5 91.7 94.4 94.5 74.2 81.9 84.1 84.6 23.8 31.1 41.8 46.3 35.4 42.8 55.1 59.0 73.5 60.6
sheets
Asbestos sheets 0.1 0.5 0.3 0.4 0.0 0.3 0.3 0.6 0.1 0.0 0.2 0.1 0.1 0.1 0.2 0.3 1.3 0.4
Tiles 3.8 2.4 1.2 1.5 0.5 1.0 0.2 0.1 0.0 0.1 0.0 0.1 0.3 0.4 0.2 0.2 0.1 0.2
Other 0.0 0.0 0.3 0.0 1.3 0.5 0.5 0.4 0.1 0.5 0.6 1.6 0.3 0.5 0.5 1.2 0.0 1.1
Goal 4: Adequate social protection and
rights of the vulnerable and needy
groups with basic needs and services
Table 29. Net primary school enrolment

Proportion of children with rate among disabled
%
disability attending primary Enrolled
school Total 81.8
Disability rates in the NPS are calculated on With disability 57.6
the basis of the response to the single, Without disability 82.3
straightforward questions “Are you physically
handicapped?” On this basis, 3.34% of
Tanzanians are classified as handicapped.
Figure 8 shows a breakdown of the type of
disability among this slice of the population
classified as disabled, with blindness and
partial paralysis as the largest categories.
Among children in the age-range used to

calculate net enrolment rates for primary
schooling, i.e., 7 to 13 years old, the disability
rate 1.40% by this definition.
This provides a quite small sub-sample o f

disabled children in the NPS sample with Figure 9. Types of disability among those
which to calculate enrolment rates, of just 48 classified as disabled
children. Table 29 presents the results from
POOR EYESIGHT /BLIND
this small-sample calculation, showing that
PARALYSED/LAME/CRIPPLED
just 57.6% of disabled children of primary- POOR HEARING/DEAF
school age are enrolled, as compared with MENTALLY DISABLED
OTHER
82.3% of Tanzanian children who are not
MISSING LIMB(ARM/LEG/HAND/FOOT)
reported as disabled. While the sample size
demands caution in the interpretation of
these results, this disparity suggests a
potentially serious gap in access to education
for the disabled.
Figure 10. Orphanhood status among children
of primary school age (7-13 yrs)
Proportion of orphaned Father alive, mother alive

children attending primary Father died, mother alive
Father alive, mother died
school Father died, mother died
Figure 9 shows the proportion of Tanzanian
children of primary-school age who are
1.3%
orphaned. Among children age 7 to 13, 1.3% 3.5%
have lost both their father and mother, an 72.3%, relative to 81.8% for children with
additional 3.5% have lost their mother, and an both parents alive.
additional 7.4% have lost their father.
Table 30 shows the primary school enrolment

rate among orphaned children, distinguishing
among children who have lost neither, one, or
both parents. As with the case of enrolment
among disabled children, it should be noted
that the sample sizes available to calculate
these enrolment rates are quite small.
Taking the results at face value, the enrolment

pattern implies little or no disadvantage in
educational access for children losing one
parent, but a modest enrolment gap for
children who have lost both parents. The
latter group shows an enrolment rate of
Table 30. Net primary school enrolment

rate among orphans
%
Enrolled
Total 81.8
Non-orphan 81.5
Father deceased 87.3
Mother deceased 83.6
Both parents deceased 72.3
53
CLUSTER 3:
GOVERNANCE AND ACCOUNTABILITY
The core of the NPS is a questionnaire related with the Cluster 3 indicators in mind, a
to households’ material conditions, their deliberate attempt was made in the design of
economic welfare and their means of earning the NPS to incorporate into the questionnaire
a living and accessing basic services and a specific modules related to governance
amenities. The majority of governance and indicators that are appropriately captured at
accountability indicators under Cluster 3 of the household level. This includes a module
MKUKUTA relate to system and institution on crime, citizens’ views about their local and
outcomes that are best measured through national leaders, and women’s experiences of
administrative data. However, specifically gender-based violence.
Indicator NPS
coverage
Goal 1: Structure and systems of governance as well as the rule of law are democratic,
participatory, representative, accountable and inclusive
% of population with birth certificates Yes
Proportion of women among senior civil servants No
% of women representatives (elected) to district council No
Proportion of women among Members of Parliament No
Proportion of villages assemblies holding quarterly meeting with public minutes Yes
Proportion of LGAs posting public budgets, revenue and actual expenditures on easily No
accessible public notice boards
% of female from small holder households with land ownership or customary land Yes
rights
Goal 2: Equitable allocation of public resources with corruption effectively addressed
Total revenue collected as % of revenue due at national level No
% of procuring entities complying with the public procurement act and procedures No
% of government entities awarded clean audit certificate from the National Audit No
Office
Number of corruption cases convicted as % of number of investigated cases No
sanctioned for prosecution by the Director of Public Prosecutions
% of LGAs that receive the full calculated amount of their annual formula-based No
budget allocation
Total value of revenue received from concessions and licenses for mining, forestry, No
fishing and wildlife as % of their estimated economic value
54
Indicator NPS
coverage
Goal 3: Effective public service framework in place to provide foundation for service delivery
improvements and poverty reduction
% of population reporting satisfaction with Government Services No
% of population who found key service providers to be absent when they needed a No
service
Goal 4: Rights of the poor and vulnerable groups are protected and promoted in the justice system
% of court cases outstanding for two or more years No
% of prisoners in remand for two or more years compared to all prisoners in a given No
year
% of detained juveniles accommodated in juvenile remand homes No
% of districts with a team of trained Paralegals No
Goal 5: Reduction of political and social exclusion and intolerance
Number of cases filed on infringement of human rights No
Goal 6: Improved personal and material security, reduced crime, eliminate sexual abuse and
domestic violence
Average no. of inmates per facility as % of authorized capacity No
Number of cases of crimes reported Yes
% of cases of sexual abuse reported that resulted in a conviction No
% who agree that a husband is justified in hitting or beating his wife for a specific Yes
reason
Goal 7: National cultural identities enhanced and promoted
At present, process indicators will be used which follow the MKUKUTA Annex.
Outcome indicators should evolve from this process.
55
Goal 6: Improved personal and material
security, reduced crime, eliminate
sexual abuse and domestic violence
she was free to stop the interview at any time
or request that the enumerator move on to
% who agree that a husband is another question.
justified in hitting or beating his
Attitudes
wife for a specific reason Roughly half (51%) of Tanzanian women feel
The first round of the NPS included a free-
that a husband is justified in hitting or beating
standing module in the household
his wife in at least some circumstances.
questionnaire devoted specifically to the issue
of gender-based violence (GBV). The inclusion Table 32 shows women’s response to a
of this module – though well outside the question – now standard in surveys around
normal scope of integrated household surveys the world – asking whether such abuse is
– was deemed important by the NPS technical justified in specific contexts. For the four
committee given the increased focus on circumstances commonly included in this
violence against women among civil society battery of questions – going out without
groups and international organizations, paired telling him, neglecting the children, arguing,
with the relative lack of concrete information or refusing sex – between 1/4th and 1/3rd of
on its prevalence, causes and effects. women feel that physical violence is justified.
The GBV module in the NPS focused on two Interestingly however, when presented with
dimensions of the issue: (i) attitudes toward
GBV, which are the basis for the MKUKUTA
indicator and are currently being measured Table 32. % of women 15-50 yrs old who
again through the DH, and (ii) prevalence of agree with the following statements
GBV – i.e., actual, self-reported experience of
“Sometimes a husband is annoyed or angered by
victimization.
things that his wife does. In your opinion, is a
husband justified in hitting or beating his wife in
The questionnaire was administered to all
the following situations:”
women in the sampled households aged 15 to
% who agree
50, regardless of marital status. Questions on
If she goes out without telling him 35.8
prevalence focused on intimate-partner
violence and referred to either a husband, or If she neglects the children 38.1
boyfriend where applicable. If she argues with him 31.3
If she refuses to have sex with him 33.7
Enumerators administering the GBV module
If there are problems with his or her
were instructed to take particular caution in 3.5
family
ensuring that the interview was conducted in
If there are money problems 1.7
private, that the respondent understood that
If there is no food at home 6.4
all responses would be confidential, and that
Table 33. Self-reported incidence of domestic violence (% of women 15-50 yrs old)
in the
Has your current partner, or any partner... past 12
ever... months...
Slapped or thrown something at you that could hurt you? 18.0 7.9
Pushed you or shoved you? 13.1 6.6
Hit you with his fist or with something else that could hurt you? 10.9 4.8
Kicked you, dragged you, or beaten you up? 8.1 3.6
Choked or burnt you on purpose? 1.2 0.5
Threatened to use or actually used a gun, knife or other weapon against you? 2.3 1.0
Physically forced you to have sexual intercourse when you did not want to? 8.0 4.9
Did you ever have sexual intercourse you did not want because you were 6.1 3.5
afraid of what he might do?
alternative scenarios which are less closely As alarming as these statistics on prevalence
related to any specific action by the women are, there is reason to believe that the
(family problems, money problems, lack of underlying truth is much worse, and that
food), the vast majority of respondents felt women significantly under-report actual
that felt that violence was not justified. violence. One suggestive piece of evidence is
the significant relationship between the
Prevalence gender of the interviewer and the prevalence
Table 33 shows the pattern of women’s of violence reported. As seen in the bottom
responses to a series of questions about their
panel of Table 35, the overall rate of
own experiences with GBV. The structure of
reported physical abuse is more than double
the table mimics the structure of the
(31.8% versus 15.3%) when a woman is
questionnaire itself: respondents were asked
speaking to another woman. These results
whether their current partner or any other
provide a clear guideline for the conduct of
partner had ever committed one of the acts
future rounds of the survey.
listed. For affirmative responses, follow-up
questions asked about the frequency of this Correlates
abuse and whether it had occurred in the last This section examines personal characteristics
12 months. that correlate with women’s attitudes toward
and experiences of GBV. Correlation is not
Overall, 14.1% of women reported that they causation. Nevertheless, much can be
had ever been slapped or had something learned about risk-factors associated with
thrown at them by a partner, and 5.8% of
GBV by examining the profile of victims.
women had experienced this in the past year.
6.5% of women reported having been raped Table 35 disaggregates the responses from
by a partner, and 3.9% in the past. In total, previous tables along several dimensions:
19% of women reported that they had ever location, age, marital status, education, and
experienced one of the forms of violence employment. The implicit hypothesis behind
listed. this disaggregation is that more autonomous
individuals may be less subject to condone or
fall victim of violence.
57
Table 34. % of domestic violence victims Table 35. Gender-based violence indicators,
who report abuse to various institutions by respondent’s characteristics (women 15-50
yrs old)
After any of the incidents of physical
% who have been abused

violence, did you ever go to […] for help?
% of those abused who

% who have ever been
justified in some cases
% who agree hitting is
Family 49.6
Hospital/health centre 7.3
reported abuse
in past 12 mos.
Village/community leaders 21.7
NGO 0.7
abused
Religious leader 2.8
Police 4.4
No one 0.4 Total 53.5 24.9 13.4 56.7
By location:
Dar es Salaam 47.0 23.8 13.4 52.3
Taking each of these dimensions in turn,
Other urban areas 49.8 19.6 10.2 54.9
results show that while attitudes are more
Rural areas 55.6 26.9 14.5 57.7
permissive toward GBV in rural areas, and the
By age:
prevalence is higher, these differences are not
15-19 yrs 42.8 6.0 4.6 48.3
large. GBV is ubiquitous. Similarly, there are
20-29 yrs 54.3 28.3 17.7 55.4
not particularly marked differences in
30-39 yrs 58.8 31.3 17.1 58.5
attitudes toward or prevalence of GBV across
age groups, with the exception of the 40-50 yrs 57.1 31.9 11.2 58.3
youngest bracket (15-24 years) where both By marital status:
are much lower. Married - mono 57.2 27.5 15.9 57.5
Married - poly 58.3 45.1 21.5 55.3
Looking at marital status, single, never- Living together 70.8 35.4 25.7 50.4
married women are the least likely to Divorced 54.2 44.6 16.9 72.3
condone violence, and women living together Separated 65.7 36.6 14.2 80.9
with a partner but not married are the most Never married 40.7 7.3 4.4 44.6
likely to condone it. The highest rates of Widowed 53.4 21.8 5.4 33.2
prevalence are seen among women in
By education:
polygamous marriages and divorcées.
None 61.8 29.9 15.7 52.8
Results are fairly strong in suggesting that Some primary 58.7 31.2 16.9 59.6
education reduces women’s acceptance and Some secondary 36.5 7.9 4.3 43.1
experience of GBV, in line with the autonomy By employment status:
hypothesis. However, dividing women No income 53.4 20.4 11.3 58.1
between those who have and do not have a Non-farm income 53.6 35.1 18.2 54.9
non-farm cash income, shows the opposite Alcohol consumption
result. Women with an earned-income are No 52.2 23.4 12.5 55.2
more accepting of GBV, and more than twice Yes 63.0 35.8 20.3 63.8
as likely to report having experienced violence Gender of interviewer
themselves. Male 49.5 21.0 11.6 52.4
Female 62.3 33.8 17.5 62.5
58
Appendix A. ‘Projecting’ poverty rates
for comparison with the HBS
This appendix presents the methodology used large body of applied work on poverty
to estimate preliminary, “predicted” poverty mapping.18
rates for the NPS, as presented in the section
of the report on “Cluster 1: Growth and At the core of all of these approaches is a
Poverty Reduction”. The goal of the analysis linear regression model of household
presented here is to enable comparison of consumption. And the basic underlying logic
poverty over time even when there is doubt is closely analogous to that of poverty
about the comparability of the measures of mapping. In the latter context, the
household consumption between surveys (as relationship between consumption and
in the case of the HBS and NPS, where the widely-available household and community
methodology for collecting consumption data characteristics is estimated in survey data,
changed dramatically between surveys). and the results are used to estimate
consumption (and hence poverty) in
The methodology draws on Mathiassen disaggregated geographic areas with census
(2009) who presents a regression-based data. In the current application, the challenge
model for estimating the headcount poverty is to extrapolate over time rather than space,
rate without relying on consumption or using the relationship between consumption
expenditure data.15 A similar approach has and household characteristics in the 2007 HBS
been presented by Stifel and Christiaensen to predict poverty rates in 2008/09 based on
(2007) to estimate the evolution of poverty changes in those same household
over time in Kenya.16 Christiaensen, et al characteristics. The following sections detail
(2008) extend this approach with further the steps in this process.
applications to Russia and Vietnam.17 The
latter authors build directly on Small-Area-
Estimation techniques formalized by Elbers et Step 1. Comparing HBS (2007)
al. (2003), which laid the groundwork for a and NPS (2008/09) variables
The first task in a survey-to-survey
15 comparison of consumption levels is to
A. Mathiassen (2009). “A model based approach
for predicting annual poverty rates without identify a set of plausible correlates of
expenditure data.” Journal of Economic Inequality, consumption that are measured in a
7(2), pp. 117 – 136. comparable way across surveys. The HBS and
NPS questionnaires differ enormously in
16
D. Stifel and L. Christiaensen (2007). “Tracking terms of structure and content. However,
poverty over time in the absence of comparable
precisely for the purpose of maintaining a
consumption data.” World Bank Economic Review,
consistent set of welfare indicators across
21(2), pp. 317 – 341.
surveys, virtually all NBS social surveys employ
17
L. Christiaensen, P. Lanjouw, J. Luoto, and D.
18
Stifel (2008). “The reliability of small area C. Elbers, J. Lanjouw and P. Lanjouw (2003).
estimation prediction methods to track poverty.” “Micro-level estimation of poverty and inequality.”
World Bank, mimeo. Econometrica, 71(1), pp. 335 – 364.
a common set of core questions that provide the discrepancies in response options
the basis for the modelling exercise here. rendered these variables unusable for
consumption modelling.
In particular, variables are available in five
categories that are commonly used in Small • Assets. The HBS and NPS ask about
Area Estimation of poverty. These include: household ownership of a common set of
52 assets, including the number of each
• Demographic characteristics, including the item owned.
number of males and females in the
household in various age categories, as Given that the point of the consumption
well as indicators of marital status and modelling exercise conducted here is to
relationship to the household head. measure changes in poverty over time, rather
than to map poverty rates across space is in
• Education levels for each member of the standard Small Area Estimation, it is
household. We also interact education important to highlight the variables in the
with the household-head variable to allow model that can reasonably be expected to
for the potentially greater significance of vary over time and thus explain genuine
the head’s human capital in explaining the changes in welfare between 2007 and
socioeconomic status of the household. 2008/09.19 These variables include
Also, the two surveys contain comparable occupational status, school enrolment, and a
questions about the current enrolment number of the less-durable (and therefore
status of children, and the level of more frequently purchased) asset variables.
schooling attended: primary, secondary or
tertiary. In total, after inspection of the questionnaires
and summary statistics for each survey, a total
• Occupational status. Unfortunately, the of 86 variables were deemed comparable
labour market questions in the HBS and across years. These 86 variables comprise the
the NPS are not identical. However, basic set of ‘candidate’ variables to be used in the
indicators can be constructed in a regression modelling in the next section. The
comparable way. These include an full set of candidate variables used in the
indicator variable for whether or not each analysis is given in Table 36, separated by
household member is working at all, and
stratum and listed side-by-side for the two
whether or not their primary activity is
survey samples.
agriculture.
• Housing and amenities. The HBS and NPS

contain nearly identical questions on the
materials used to construct the roof,
walls, and floor of the dwelling, and
questions about the tenure of the house
19
(owned, rented, occupied without rent, Note that the average levels of all variables in
the model are likely to change between surveys,
etc.). An attempt has been made to line
due to sampling error and measurement error in
up responses to questions about energy
both surveys. Our focus here is on variables that
and fuel usage for the section of the
might genuinely change for a given household over
report above that deals specifically with time, ignoring measurement error.
these topics. However, it was judged that
60
Step 2. Modelling consumption
in the 2007 HBS
The model of consumption
At the core of step 2 is a regression model of
consumption which is estimated with the
2007 HBS data and used to construct
“predicted” consumption in the 2008/09 NPS
data. Denote total nominal household where denotes the standard cumulative
consumption per adult equivalent in normal distribution function. The final
household i and cluster j at time t by . equation in this expression shows how
Assuming a log-linear relationship between estimates of the model parameters from (1)
combined with household characteristics yield
consumption and the household
characteristics described above, contained in an estimate of poverty rates. Averaging these
individual poverty probabilities over the
the matrix , we can write:
sample of individuals gives our empirical
estimator of headcount poverty:
(1)
where is an error term with total variance

. The headcount poverty rate is
defined as the proportion of the population (3)
whose consumption per adult equivalent lies
below some poverty line, which we denote z.
Based on our model of consumption, the
probability that a given individual is poor can
be rewritten as follows:
(2)
In practice (3) is estimated with household

rather than individual level data, and this
difference is accounted for in the sampling
weights denoted .
Finally, because estimates of the headcount

poverty rate from equation (3) will be highly
sensitive to the variance of the residuals, it is
important to account for possible
heteroskedasticity in the consumption model.
61
Figure 11. Cumulative distribution of log of the residual for each data point as a
consumption by stratum function of the exogenous characteristics in
the consumption model:
CDF of Actual & Predicted Values

Dar es Salaam (4)
1
As with the main consumption model, the

.5
variance model in equation (4) is estimated

0
Other urban
using the 2007 HBS data, and the estimated γ
1
parameters are used to predict error
Cumulative Probability
variances for the 2009 NPS sample.
.5 Estimating the model

The regression model in equation (1) is
estimated by OLS using data from the 2007
HBS. The model is estimated separately for
0
each stratum of the survey, to allow for
Rural parameter heterogeneity, i.e., differences in
1
the relationship between household
characteristics and consumption across strata.
All regressions are weighted with individual
.5 sampling weights.
In many applications of Small Area Estimation,

the set of potential regressors is limited by
0 the small number of variables common to the
6 8 10 12 14
Log of monthly consumption per adult equivalent household survey and the census. In the case
Actual 2007 Predicted 2007 of comparing the HBS and the NPS the
Predicted 2009 opposite is the case: as already noted, the
number of variables available is quite large
relative to the sample size in particular strata
of the NPS: 86 variables compared to just XXX
observations in “other urban areas”.
Lacking firm theoretical grounds on which to

prioritize certain variables over others, we opt
instead to use a stepwise regression
approach, dropping variables from the model
To do so, we estimate the predicted variance whose regression coefficients have a p-value
62
of less than 0.2. The stepwise procedure is Figure 12. Predicted food poverty and basic
repeated for each strata, allowing the model needs poverty rates, HBS 2007 and NPS
specification – like the individual regression 2008/09, based on alternative samples
coefficients – to vary across strata, as it is
likely that the set of characteristics explaining
consumption in rural and urban areas may Full Sample
differ dramatically.
Food Basic Needs
The estimates of the regression model are 50
given in Table 37. In addition to the log consumption. This lower dispersion will
independent variables shown, all equations lead to the calculation of lower poverty rates.
include a full set of dummies for region and The inclusion of the simulated error terms is
month of interview (these variables were not necessary in the statistical calculation of
‘locked’ into the model during the stepwise headcount poverty rates, as the estimated
procedure). The R-squared across various variance of the residuals is accounted for in
strata ranges from .4 to .45, a reasonably high equation 3.)
number by the standards of SAE. Particular
care should be taken in not over-interpreting The final “predicted” poverty rates based on
individual parameter estimates. For instance, our model are given in Table 38. These
the fact that the dummy variable for whether results correspond to, and provide greater
the head of household is married has a detail underlying the numbers presented in
significant negative coefficient in Dar es the main text in Table 3.
Salaam should not be interpreted, in isolation,
to imply that such households are generally Note that in reporting the poverty rates in Box
poorer. This statement is only true controlling 1 in the main part of the report, only the
for a large, highly correlated, and somewhat changes are used (column 3 of Table 38).
arbitrary set of other household Rather than revise previous poverty estimates
characteristics. for the HBS – which would be unwarranted
and lead to potential confusion – the Box puts
As a check on the estimates provided by the the NPS figures in context simply by adding or
model, Figure 10 shows a graphical subtracting any estimated percentage-point
comparison of the raw consumption data (log changes to the poverty rates estimated for the
consumption per adult equivalent) and the 2007 HBS and already published in the official
fitted values from the regression model for report.
both 2007 and 2009. For the sake of graphical
comparison, the fitted values include a The table shows the regression of the
simulated error component. This is created predicted probability of poverty for each
by drawing a random number from a normal household, in both the HBS and NPS datasets
distribution with mean zero and variance as (each row corresponds to a separate
predicted by the variance model described regression). The first column “HBS
above. (Without including theses simulated headcount” shows the weighted-average
errors, the variance of predicted log value of these probabilities in the HBS, and
consumption will be, by construction, the third column “Change in the NPS” shows
significantly lower than the variance of actual the difference, if any, in the poverty rate in
2008/09.
63
The overall pattern of results in the top panel underlying consumption model using the full
of the table – using the full sample of sample of HBS households, but varying the
households – is that poverty has not changed specification by dropping one regressor in
significantly (in either an economic or each iteration. (Note: the stepwise procedure
statistical sense) in Tanzania between the is not repeated. In each iteration one
2007 HBS and the 2008/09 NPS. However, regressor is dropped from the baseline model
breaking this down by stratum, there is a hint presented in Table 37 and subsequently
of a decline in poverty in urban areas outside replaced in the next iteration.)
Dar es Salaam (4.34% decline for food
poverty; 6.5% decline for basic needs The results from this exercise are summarized
poverty), but nationally this decline is offset in Figure 12, showing the predicted change in
by a smaller but statistically significant food and basic needs poverty from each
increase in poverty in rural areas (1.77% iteration of the model. In the box plots in the
increase for food poverty; 2.09% increase for figure each observation corresponds to one
basic needs poverty). The following section iteration of the model. The boundaries of the
examines the robustness of this result. box – which are barely distinguishable given
the relatively stability of the results – show
the 25th and 75th percentiles. As seen, these
Step 3. Checking the robustness results seem to indicate that the predicted
of the results poverty rates in Table 38 are fairly stable,
with an average increase in food poverty of
Robustness of the consumption model around 1/4th of one percentage point, and no
specification change whatsoever in basic needs poverty.
The first technique used to check the
robustness of the results in Table 38 is to Using the panel dimension of the survey
vary the model specification. The underlying to remove sampling error from the
concern is that the model employed in Step 2 comparison of the HBS and NPS
– and in particular the set of X characteristics Out of the 410 clusters in the NPS, 200 were
included in the regression specification – were taken from the HBS sample of clusters.
quite arbitrary. If the overall results for Furthermore, a sub-sample of the NPS
predicting poverty are sensitive to the households in these clusters were drawn from
particular specification chosen, this would the HBS sample, creating a panel of
undermine confidence in the results. households. This section uses the unique
features of panel data to provide an
To test for robustness, we re-estimate the alternative estimate of the change in poverty
between the two survey rounds. Returning to
Figure 13. Box plots showing the the same clusters or same households
sensitivity of poverty projections to dramatically reduces sampling error in
alternative specifications measuring a change in any indicator.
However, there is a trade-off here. The
Food poverty Basic needs sample size is much smaller for the panel
1
element of the NPS, in particular for the Dar
es Salaam and ‘other urban areas’ strata.
.5
rty
64
Any sample survey is subject to sampling Salaam and increasing in very modestly in
error. There are known probabilities that a rural areas.
sample statistic will deviate from the true
population statistic by some degree. Two
surveys using independently drawn samples
of the same population will, then, potentially
diverge, even if the underlying reality has not
changed. One way to avoid this sampling
error in making comparisons over time is to
return to the identical sample. Doing so
ensures that the sampling error in the
estimates – whatever they may be – will be
identical in both surveys. This allows us to
ensure – without adding any additional
variables to the model – that there are no
observed or unobserved time-invariant
differences between the samples used in the
two surveys; i.e., that apparent changes are
not due to sampling error.
The center and bottom panels of Table 38

show the results from repeating the
estimation of predicted poverty described
above, but limiting the sample to (a) only
households in enumeration areas or villages
that were surveyed in both the HBS and NPS,
or (b) specific household which appear in both
samples. In both cases, the regression
coefficients ( parameters) used are identical
to those used in the main estimates for the
full sample. This is to highlight any
differences resulting from changing the
sample, without muddying the waters by also
allowing regression coefficients to differ.
In brief, the results in the bottom parts of

Table 38 confirm the earlier results from the
main sample. Nationally, poverty appears to
be if anything lower in 2008/09 than in 2007,
but any decline is not particularly robust.
Looking across strata, we find some
corroboration for the result that poverty has
been declining in urban areas outside Dar es
65
Table 36. Average values for independent variables in the model of log consumption, by strata and year
Dar es Salaam Other urban areas Rural areas
2007 HBS 2008/09 NPS 2007 HBS 2008/09 NPS 2007 HBS 2008/09 NPS
Full Panel Panel Full Panel Panel Full Panel Panel Full Panel Panel Full Panel Panel Full Panel Panel
sample EAs HHs sample EAs HHs sample EAs HHs sample EAs HHs sample EAs HHs sample EAs HHs
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18)
Log 10.53 10.58 10.53 11.35 11.37 11.28 10.17 10.09 10.03 10.71 10.60 10.58 9.789 9.788 9.780 10.22 10.23 10.23
consumption (0.619) (0.695) (0.607) (0.637) (0.576) (0.584) (0.663) (0.665) (0.666) (0.676) (0.669) (0.652) (0.620) (0.617) (0.599) (0.575) (0.561) (0.562)
Members < 1 yr 0.110 0.0870 0.0637 0.0937 0.0822 0.0701 0.158 0.161 0.172 0.178 0.190 0.190
old (0.322) (0.293) (0.245) (0.313) (0.275) (0.256) (0.389) (0.392) (0.406) (0.417) (0.430) (0.430)
Members 1 - 5 0.428 0.422 0.496 0.441 0.402 0.424 0.587 0.525 0.490 0.471 0.379 0.382 0.836 0.846 0.847 0.732 0.727 0.727
yrs old (0.657) (0.677) (0.829) (0.624) (0.572) (0.599) (0.770) (0.733) (0.694) (0.648) (0.540) (0.561) (0.937) (0.943) (0.937) (0.837) (0.824) (0.825)
Members 5 - 9 0.319 0.260 0.296 0.501 0.491 0.480 0.512 0.502 0.490 0.577 0.571 0.611 0.661 0.669 0.640 0.842 0.850 0.849
yrs old (0.595) (0.533) (0.554) (0.733) (0.754) (0.736) (0.759) (0.776) (0.756) (0.762) (0.747) (0.790) (0.850) (0.855) (0.835) (0.902) (0.913) (0.916)
Members 10 - 14 0.374 0.355 0.400 0.417 0.372 0.384 0.576 0.559 0.599 0.588 0.580 0.605 0.679 0.687 0.708 0.758 0.764 0.764
yrs old (0.677) (0.649) (0.741) (0.686) (0.617) (0.632) (0.833) (0.808) (0.854) (0.764) (0.764) (0.758) (0.881) (0.888) (0.913) (0.935) (0.943) (0.944)
Members 15 - 24 0.801 0.731 0.944 1.138 1.179 1.328 0.851 0.791 0.879 0.991 0.858 0.879 0.864 0.863 0.879 0.935 0.954 0.957
yrs old (1.052) (1.020) (1.214) (1.247) (1.271) (1.419) (1.129) (1.078) (1.365) (1.222) (1.081) (1.134) (1.126) (1.127) (1.148) (1.160) (1.212) (1.217)
Members 25 - 34 0.790 0.762 0.792 0.857 0.855 0.784 0.696 0.574 0.529 0.691 0.644 0.611 0.638 0.640 0.641 0.601 0.597 0.602
yrs old (0.757) (0.731) (0.733) (0.859) (0.821) (0.867) (0.756) (0.702) (0.721) (0.788) (0.802) (0.829) (0.772) (0.775) (0.763) (0.768) (0.780) (0.782)
Members 35 - 64 0.863 0.855 0.944 0.890 0.897 1.040 0.890 0.925 0.987 0.865 0.881 0.936 1.007 1.016 1.045 1.024 1.079 1.075
yrs old (0.836) (0.834) (0.826) (0.849) (0.892) (0.911) (0.828) (0.806) (0.840) (0.825) (0.843) (0.860) (0.856) (0.858) (0.847) (0.839) (0.839) (0.838)
Members 65 + 0.0843 0.0928 0.160 0.115 0.103 0.112 0.150 0.165 0.217 0.187 0.242 0.306 0.249 0.250 0.257 0.284 0.293 0.291
yrs old (0.309) (0.326) (0.429) (0.379) (0.344) (0.386) (0.418) (0.450) (0.523) (0.476) (0.543) (0.606) (0.528) (0.528) (0.520) (0.560) (0.554) (0.555)
Male hh 2.487 2.507 2.528 2.578 2.663 2.658
members (1.708) (1.714) (1.667) (1.801) (1.901) (1.900)
Head male 0.691 0.676 0.662 0.680 0.671 0.650
(0.462) (0.468) (0.474) (0.467) (0.471) (0.479)
Head completed 0.259 0.281 0.272 0.240 0.274 0.248 0.164 0.137 0.127 0.159 0.128 0.108 0.0410 0.0383 0.0384 0.0351 0.0358 0.0351
sec. (0.438) (0.450) (0.447) (0.428) (0.447) (0.434) (0.371) (0.345) (0.334) (0.366) (0.335) (0.312) (0.198) (0.192) (0.192) (0.184) (0.186) (0.184)
Head completed 0.0719 0.0623 0.0800 0.0838 0.0812 0.104
tert. (0.258) (0.242) (0.272) (0.277) (0.274) (0.306)
Head working 0.944 0.939 0.920 0.924 0.923 0.920 0.942 0.930 0.917 0.932 0.936 0.930 0.940 0.942 0.940 0.966 0.965 0.965
(0.231) (0.239) (0.272) (0.266) (0.267) (0.272) (0.234) (0.256) (0.276) (0.251) (0.245) (0.256) (0.238) (0.233) (0.238) (0.181) (0.183) (0.184)
Head agric 0.0464 0.0762 0.0720 0.0745 0.0342 0.0400 0.341 0.391 0.363 0.318 0.393 0.433 0.790 0.802 0.797 0.878 0.878 0.878
(0.210) (0.265) (0.260) (0.263) (0.182) (0.197) (0.474) (0.488) (0.482) (0.466) (0.489) (0.497) (0.407) (0.398) (0.403) (0.327) (0.327) (0.327)
Head never 0.181 0.191 0.144 0.143 0.137 0.104 0.0401 0.0374 0.0340 0.0173 0.0119 0.0121
married (0.385) (0.393) (0.353) (0.351) (0.344) (0.306) (0.196) (0.190) (0.181) (0.130) (0.109) (0.109)
Head married 0.584 0.575 0.640 0.533 0.517 0.544 0.602 0.569 0.548 0.488 0.457 0.490
(0.493) (0.495) (0.482) (0.499) (0.501) (0.500) (0.490) (0.496) (0.499) (0.500) (0.499) (0.502)
Head divorced 0.0526 0.0779 0.0637 0.113 0.128 0.108
(0.223) (0.268) (0.245) (0.317) (0.335) (0.312)
Head separated 0.0556 0.0609 0.0640 0.0298 0.0299 0.0480
(0.229) (0.239) (0.246) (0.170) (0.171) (0.215)
66
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18)
Head widowed 0.0763 0.0748 0.0800 0.0838 0.0812 0.0800
(0.265) (0.263) (0.272) (0.277) (0.274) (0.272)
Members 0.0340 0.0332 0.0240 0.110 0.0940 0.0960
attending tert. (0.224) (0.221) (0.199) (0.429) (0.382) (0.369)
Walls = mud 0.0103 0.0249 0.00800 0.00931 0 0
(0.101) (0.156) (0.0894) (0.0961) (0) (0)
Walls = mud 0.0160 0.0125 0 0.00186 0.00427 0
brick (0.125) (0.111) (0) (0.0432) (0.0654) (0)
Walls = brick 0.0192 0.0388 0.0560 0.00186 0.00427 0 0.296 0.263 0.242 0.294 0.288 0.306
(0.137) (0.193) (0.231) (0.0432) (0.0654) (0) (0.456) (0.440) (0.430) (0.456) (0.454) (0.462)
Walls = concrete 0.221 0.197 0.210 0.242 0.187 0.166
(0.415) (0.398) (0.409) (0.429) (0.391) (0.373)
Walls = other 0.00444 0.00693 0.00800 0.00559 0.0128 0.0160 0.0101 0 0 0.00218 0 0 0.0120 0.0110 0.0110 0.00773 0.0119 0.0121
(0.0665) (0.0830) (0.0894) (0.0746) (0.113) (0.126) (0.0999) (0) (0) (0.0467) (0) (0) (0.109) (0.105) (0.104) (0.0876) (0.109) (0.109)
Floor = concrete 0.930 0.898 0.968 0.916 0.953 0.920 0.619 0.573 0.561 0.662 0.589 0.567 0.152 0.140 0.148 0.132 0.127 0.128
(0.256) (0.304) (0.177) (0.277) (0.212) (0.272) (0.486) (0.495) (0.498) (0.473) (0.493) (0.497) (0.359) (0.347) (0.355) (0.339) (0.333) (0.335)
Floor = other 0.0133 0.0136 0.0132 0.0161 0.0163 0.0165
(0.114) (0.116) (0.114) (0.126) (0.127) (0.127)
Bicycles 0.549 0.558 0.562 0.565 0.557 0.555
(0.703) (0.710) (0.717) (0.804) (0.800) (0.802)
Boats 0.00654 0.0122 0 0.00871 0.00913 0.0127 0.0126 0.0123 0.0110 0.00714 0.00434 0.0044
(0.0902) (0.135) (0) (0.114) (0.135) (0.160) (0.134) (0.134) (0.124) (0.109) (0.0806) (0.081)
Carts 0.0006 0.00139 0 0.00559 0.00855 0 0.0271 0.0286 0.0351 0.0274 0.0337 0.0340
(0.0243) (0.0372) (0) (0.0964) (0.131) (0) (0.170) (0.174) (0.190) (0.187) (0.203) (0.204)
Books 2.785 2.668 2.552 5.361 4.598 4.144 1.189 1.134 1.313 1.136 1.128 1.136
(11.15) (19.16) (5.997) (19.33) (15.91) (11.34) (4.201) (3.745) (4.812) (5.442) (5.342) (5.371)
Livestock 1.087 1.113 1.338 1.174 1.370 1.548 20.96 22.00 61.88 6.077 7.018 7.046
(5.357) (5.375) (6.251) (4.692) (5.809) (6.631) (877.9) (901.4) (1657) (21.95) (22.50) (22.61)
Hoes 3.033 3.072 3.065 2.861 2.907 2.905
(2.206) (2.198) (2.112) (2.013) (2.033) (2.037)
Wheelbarrow 0.0188 0.0185 0.0165 0.0416 0.0456 0.0461
(0.163) (0.163) (0.127) (0.396) (0.415) (0.417)
Mosquito nets 1.806 1.869 2.146 1.978 1.735 1.815 1.214 1.203 1.243 1.171 1.188 1.182
(1.610) (2.110) (3.364) (1.800) (1.671) (1.690) (1.367) (1.364) (1.405) (1.419) (1.469) (1.472)
Irons 0.624 0.568 0.568 0.667 0.637 0.600 0.480 0.365 0.382 0.512 0.411 0.420 0.204 0.202 0.202 0.183 0.192 0.193
(0.606) (0.574) (0.544) (0.640) (0.594) (0.582) (0.614) (0.574) (0.572) (0.659) (0.602) (0.579) (0.453) (0.453) (0.441) (0.419) (0.429) (0.430)
Cars 0.0324 0.0229 0.0318 0.0741 0.0548 0.0510 0.00308 0.00325 0.00220 0.00773 0.0109 0.0110
(0.239) (0.217) (0.238) (0.461) (0.313) (0.316) (0.0554) (0.0569) (0.0468) (0.106) (0.131) (0.132)
Watches 0.907 0.832 0.792 0.741 0.726 0.720 0.579 0.579 0.599 0.345 0.380 0.384
(1.034) (1.123) (1.166) (0.919) (0.895) (0.839) (0.897) (0.900) (0.884) (0.690) (0.748) (0.751)
67
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18)
Radios 0.887 0.853 0.847 0.935 0.890 0.936 0.778 0.774 0.790 0.752 0.751 0.753
(0.701) (0.697) (0.744) (0.788) (0.817) (0.867) (0.746) (0.742) (0.754) (0.782) (0.787) (0.790)
Ploughs 0.00059 0.00139 0 0.00745 0.00855 0 0.0229 0.0122 0 0.0109 0.00913 0.00637 0.135 0.141 0.144 0.110 0.130 0.132
1
(0.0243) (0.0372) (0) (0.122) (0.131) (0) (0.190) (0.123) (0) (0.104) (0.0953) (0.0798) (0.449) (0.458) (0.455) (0.431) (0.478) (0.481)
Cell phones 1.052 1.003 1.080 1.613 1.611 1.640 0.606 0.446 0.478 1.017 0.900 0.847 0.161 0.156 0.159 0.315 0.318 0.319
(0.981) (0.955) (0.867) (1.305) (1.259) (1.253) (0.826) (0.760) (0.829) (1.112) (1.145) (1.122) (0.431) (0.424) (0.422) (0.616) (0.602) (0.604)
Donkeys 0.133 0.139 0.125 0.121 0.138 0.139
(1.042) (1.069) (0.690) (0.712) (0.760) (0.764)
Poultry 6.669 6.728 6.675 7.228 7.562 7.533
(8.609) (8.588) (7.955) (10.03) (9.364) (9.335)
Stoves 0.0828 0.0534 0.0446 0.122 0.0868 0.102
(0.340) (0.280) (0.236) (0.421) (0.313) (0.343)
Other stoves 1.421 1.273 1.336 1.708 1.709 1.648 1.152 0.925 0.803 1.102 0.831 0.726 0.314 0.301 0.285 0.284 0.305 0.305
(0.847) (0.837) (0.906) (0.784) (0.759) (0.806) (0.962) (1.015) (0.937) (1.026) (0.900) (0.896) (0.604) (0.590) (0.574) (0.582) (0.607) (0.607)
Harvesting 0.00027 0.00153 0 0 0 0
machines 2
(0.0165) (0.0391) (0) (0) (0) (0)
Spraying 0.0313 0.0412 0.00637 0.0501 0.0594 0.0510
machines (0.261) (0.362) (0.0798) (0.280) (0.347) (0.316)
Satellite dishes 0.0695 0.0473 0.0382 0.153 0.100 0.115
(0.275) (0.220) (0.192) (0.437) (0.394) (0.438)
Mopeds 0.0286 0.0183 0.00637 0.0697 0.0320 0.0382
(0.216) (0.199) (0.0798) (0.741) (0.201) (0.223)
Fields 0.214 0.220 0.240 0.410 0.346 0.384 1.354 1.699 1.739 1.279 1.511 1.726
(0.777) (0.542) (0.447) (0.699) (0.611) (0.565) (3.425) (3.224) (2.418) (2.292) (2.273) (2.536)
Milking 0.00148 0 0 0.00372 0.00855 0 0.00027 0 0 0 0 0
machines 2
(0.0516) (0) (0) (0.0863) (0.131) (0) (0.0165) (0) (0) (0) (0) (0)
Heaters 0.248 0.217 0.197 0.0632 0.0137 0.0191
(0.885) (0.727) (0.645) (0.353) (0.117) (0.137)
Telephones 0.0207 0.0153 0.0127 0.0283 0.00913 0.0127
(land0 (0.155) (0.123) (0.113) (0.202) (0.135) (0.160)
Water pumps 0.00654 0.00916 0 0.00871 0.00913 0.0127
(0.0839) (0.0953) (0) (0.0930) (0.0953) (0.113)
TVs 0.471 0.470 0.472 0.596 0.594 0.560 0.159 0.105 0.0764 0.303 0.237 0.223
(0.563) (0.562) (0.590) (0.645) (0.630) (0.627) (0.388) (0.331) (0.290) (0.570) (0.540) (0.526)
Couches 1.074 0.884 0.924 2.492 1.986 1.834
(1.624) (1.529) (1.466) (3.201) (3.342) (2.552)
Tables 1.563 1.553 1.656 1.721 1.676 1.694
68
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18)
(1.221) (1.364) (1.367) (1.324) (1.375) (1.319)
Motorcycles 0.0262 0.0214 0.0255 0.0349 0.0365 0.0510
(0.177) (0.145) (0.158) (0.216) (0.251) (0.295)
Beds 2.177 2.215 2.433 2.318 2.187 2.338
(1.928) (2.534) (3.359) (1.734) (1.744) (1.655)
Lamps 1.052 1.024 1.025 1.068 0.986 0.962
(0.982) (1.012) (1.209) (1.071) (1.038) (1.061)
Pots 22.68 20.15 18.52 56.67 44.32 46.02
(23.69) (20.74) (17.25) (67.84) (46.10) (47.12)
Fans 0.615 0.568 0.632 0.745 0.718 0.744
(1.031) (1.002) (0.963) (1.113) (1.047) (0.975)
Fridges 0.326 0.353 0.304 0.425 0.415 0.440
(0.553) (0.583) (0.512) (0.662) (0.617) (0.653)
Video players 0.255 0.269 0.248 0.454 0.479 0.432
(0.491) (0.497) (0.452) (0.654) (0.636) (0.614)
Tractors 0.00029 0.00139 0 0.00745 0.00855 0
6
(0.0172) (0.0372) (0) (0.122) (0.131) (0)
Harrowing 0.00118 0.00139 0 0.00931 0.0171 0.0160
machines (0.0344) (0.0372) (0) (0.114) (0.160) (0.126)
Cupboards 0.984 0.845 1 1.196 1.291 1.400
(1.427) (1.212) (1.631) (1.428) (1.542) (1.486)
Observations 3382 722 125 537 234 125 3671 655 157 459 219 157 3245 3078 911 1681 921 911
69
Table 37. Regression of log consumption per adult equivalent on household
characteristics in 2007 HBS
(1) (2) (3)
Members < 1 yr old -0.0730** -0.0710***
(0.0297) (0.0244)
Members 1 - 5 yrs old -0.0218 -0.0449*** -0.0437***
(0.0157) (0.0153) (0.0113)
Members 5 - 9 yrs old -0.0959*** -0.0981*** -0.0958***
(0.0175) (0.0124) (0.0116)
Members 10 - 14 yrs old -0.143*** -0.131*** -0.146***
(0.0172) (0.0139) (0.0113)
Members 15 - 24 yrs old -0.137*** -0.0882*** -0.111***
(0.0106) (0.00989) (0.00945)
Members 25 - 34 yrs old -0.0936*** -0.0777*** -0.0536***
(0.0149) (0.0167) (0.0164)
Members 35 - 64 yrs old -0.0646*** -0.114*** -0.0844***
(0.0174) (0.017) (0.0152)
Members 65 + yrs old -0.0780** -0.0710** -0.0827***
(0.0348) (0.0296) (0.0207)
Male hh members 0.015
(0.00948)
Head male 0.045
(0.0305)
Head completed 0.0933*** 0.0561* 0.117**
secondary school (0.0244) (0.0301) (0.0467)
Head completed 0.227***
tertiary school (0.0457)
Head working 0.139*** 0.157*** 0.200***
(0.0454) (0.049) (0.0436)
Head in agric -0.0851 -0.0728*** -0.168***
(0.053) (0.0257) (0.0276)
Head never married -0.0523 0.105*
(0.0384) (0.0592)
Head married -0.0854** -0.0713**
(0.0332) (0.0297)
Head divorced -0.127**
(0.0544)
Head separated -0.101*
(0.0545)
Head widowed -0.0953*
(0.0501)
Members attending 0.0562
tertiary (0.0385)
Walls = brick 0.139 0.0471*
(0.0909) (0.0263)
Walls = mud 0.159**
(0.0784)
(1) (2) (3)
Walls = mud brick -0.101
(0.0751)
Walls = concrete 0.0650**
(0.03)
Walls = other -0.336** 0.313*** 0.240***
(0.169) (0.0855) (0.0671)
Floor = concrete 0.0783* 0.127*** 0.0910***
(0.0435) (0.0266) (0.0272)
Floor = other 0.0681
(0.0519)
Cell phone 0.0406*** 0.0828*** 0.0902***
(0.0135) (0.015) (0.0218)
Other stove 0.0366*** 0.0485*** 0.0327**
(0.0127) (0.0142) (0.0151)
Plough 0.667*** 0.0817* 0.0809***
(0.0595) (0.0466) (0.0229)
Iron 0.0466** 0.0471** 0.0722***
(0.0211) (0.0205) (0.0202)
TV 0.0952*** 0.0595*
(0.0249) (0.0316)
Milking machines 0.531** 0.601***
(0.209) (0.067)
Fields 0.0156 -0.00447*
(0.0102) (0.00251)
Books -0.000871* 0.00469**
(0.00052) (0.00232)
Carts 0.456* 0.0786
(0.263) (0.0585)
Watches 0.0566*** 0.0217**
(0.013) (0.0102)
Cars 0.0679 0.606***
(0.0469) (0.224)
Radios 0.0389** 0.0574***
(0.0172) (0.0139)
Livestock 0.00392*** 5.13e-06***
(0.00151) (1.7E-06)
Boats 0.151 0.120**
(0.0927) (0.0596)
Mosquito nets 0.0245** 0.0454***
(0.0115) (0.00761)
Fridges 0.0980***
(0.0256)
Video players 0.037
(0.0254)
Tractors -2.085***
71
(1) (2) (3)
(0.321)
Harrows 0.289
(0.215)
Cupboards 0.0159**
(0.00733)
Fans 0.0343***
(0.0119)
Spraying machines 0.0437
(0.0314)
Satellite dishes 0.0862**
(0.0427)
Mopeds 0.149**
(0.0663)
Heaters 0.0152
(0.0106)
Telephones (land) -0.106
(0.069)
Water pumps 0.209*
(0.125)
Couches 0.0161*
(0.00912)
Tables 0.0324***
(0.0103)
Motorcycles 0.106**
(0.0502)
Beds -0.0104
(0.00771)
Lamps 0.0290***
(0.011)
Pots 0.00164***
(0.0005)
Stoves 0.0752**
(0.0305)
Harvesting machines -0.695***
(0.269)
Bicycles 0.0713***
(0.0174)
Hoes 0.0226***
(0.00534)
Wheelbarrows 0.154**
(0.0617)
Donkeys 0.0241***
(0.00608)
Poultry 0.00245**
(0.00119)
72
(1) (2) (3)
Constant 10.36*** 9.799*** 9.505***
(0.077) (0.0823) (0.0691)
Observations 3382 3671 3245
R-squared 0.467 0.466 0.448
73
Table 38. Projected poverty rates based on regression model of log consumption
HBS Change in
Poverty line Headcount S.E. NPS S.E. Observations
Full Sample
Mainland Food 0.175*** (0.00412) 0.00554 (0.00745) 12912
Basic needs 0.340*** (0.00522) 0.00449 (0.00947) 12912
DSM Food 0.0682*** (0.00359) 0.00188 (0.00908) 3906
Basic needs 0.167*** (0.00558) -0.00510 (0.0134) 3906
Other urban Food 0.127*** (0.00435) -0.0414*** (0.0117) 4109
Basic needs 0.254*** (0.00581) -0.0596*** (0.0180) 4109
Rural Food 0.196*** (0.00526) 0.0198** (0.00927) 4897
Basic needs 0.378*** (0.00656) 0.0251** (0.0113) 4897
Panel EAs
Mainland Food 0.192*** (0.00515) -0.00220 (0.0103) 5796
Basic needs 0.370*** (0.00640) -0.00924 (0.0125) 5796
DSM Food 0.0733*** (0.00737) -0.000340 (0.0123) 954
Basic needs 0.177*** (0.0122) -0.00185 (0.0210) 954
Other urban Food 0.137*** (0.0101) -0.0524*** (0.0147) 869
Basic needs 0.276*** (0.0140) -0.0781*** (0.0217) 869
Rural Food 0.197*** (0.00545) 0.0224* (0.0120) 3973
Basic needs 0.378*** (0.00675) 0.0287** (0.0141) 3973
Panel HHs
Mainland Food 0.189*** (0.00908) 0.0103 (0.0132) 2367
Basic needs 0.364*** (0.0110) 0.0110 (0.0158) 2367
DSM Food 0.116*** (0.0281) -0.0234 (0.0325) 250
Basic needs 0.241*** (0.0413) -0.0321 (0.0491) 250
Other urban Food 0.127*** (0.0152) -0.0373* (0.0206) 311
Basic needs 0.272*** (0.0238) -0.0695** (0.0313) 311
Rural Food 0.192*** (0.00947) 0.0266* (0.0143) 1806
Basic needs 0.369*** (0.0114) 0.0367** (0.0169) 1806
Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1
74
Appendix B. Creating the price deflators
As noted at the beginning of this appendix, the official CPI series, and prices reported by
the most common way of comparing households within the household surveys
consumption figures over time is to use a themselves (both HBS and NPS).
price index that adjusts nominal monetary
values for inflation. The price index is Where available, it is generally recognized as
intended to measure differences in the overall best practice to use prices from the household
cost of living, both over time and between survey itself, as these measure the actual cost
geographic areas. Constructing the price of goods in the specific locations where the
index requires data on prices for a large consumption data is collected (Deaton &
number of goods (ideally, all the goods in the Zaidi). Because the CPI is based exclusively on
consumption aggregate) and information on urban prices in Tanzania, it may not reflect
the budget shares for each item to weight its local price variation, especially in rural areas.
importance in the overall price index. Thus the preferred price deflator for food
prices in this report compares prices from
For the NPS poverty analysis, three separate within the HBS and NPS surveys. However,
price indices were created: for Dar es Salaam, survey data comparing the HBS and NPS is not
other urban areas, and rural areas available for non-food prices, as explained
respectively. Each index began with a value of below. Thus the price deflator used in the
1 in Jan 2007, the first month of HBS data NPS analysis for non-food prices relies on CPI
collection and ends in September 2009, the data, reweighted to reflect contemporary
final month of NPS 2008/09 data collection. A budget shares.
value for each index is computed for each
month. When the price for a given item in a Prices from within the household surveys
given stratum and given month was Within both the HBS and the NPS surveys,
unavailable (for instance, because no prices can be computed using ‘unit values’.
household in the sample purchased that item The unit value is simply the total expenditure
in that month), missing values were imputed on an item (say TZS 6,000 of maize) divided by
through interpolation as described below. In the quantity consumed (say 10 kg of maize,
addition, because the prices from survey data yielding a unit value of 600 TZS/kg) for a given
can be quite noisy due to sampling error with household and a given item. Unit values for a
small samples in each stratum-month, the consistent set of food items were constructed
price series were smoothed by taking a from the 2007 HBS and the NPS. (Unit values
moving average of prices for each item over were re-computed for the 2007 HBS, rather
time. than relying on the published figures from the
All “real” monetary values in this report are 2007 HBS report because of the need to align
deflated using these monthly, stratum-specific the item categories with the NPS as discussed
price indices. in detail below.)
Unit values for non-food consumption are not

Sources of price data
There are two principal sources of price data available in the NPS (Sections L and M of the
used in the NPS poverty analysis: raw prices household questionnaire, as well as sections C
collected by NBS regional staff as inputs into and D, respectively for education and health
Figure 14. Examples of ‘raw’ and ‘smoothed’ price series for Dar es Salaam
Maize grain Irish potatoes
800
expenditure) or for food consumption outside the unit values from two periods forward).
the household (e.g. restaurant meals, as Finally, the third step is to impute missing unit
measured in Section F). In order to measure values. This is done through log-linear
unit values, it is necessary to have both the interpolation/extrapolation. Interpolation
price paid in monetary units (current TZS) and and extrapolation are used for only a handful
the quantity purchased in comparable units of monthly-stratum-item cells, all of which are
across observations. In sections C, D, L, and M items that have budget shares less than 0.5%.
only the total TZS monetary value paid is
recorded. There are no quantities recorded to To provide a visual illustration of what the
use in the denominator for these sections, so moving average filter and interpolation do to
instead CPI price data is used. the unit value time series for each item,
For food consumption, unit values from Figure 13 shows two examples from Dar es
roughly 40,000 food purchases in the NPS and Salaam: the raw monthly unit values for maize
over 1.7 million food purchases in the HBS are grain and Irish potatoes respectively. As seen
used to construct monthly prices for each in the figure, the raw unit values are
item in each stratum. This is done in three somewhat volatile, and this volatility is
steps. First, within each month-item-stratum moderately smoothed by the blue moving
cell, the median unit value is found. This is average line. In addition, missing values for
referred to below as the ‘raw’ unit value. For the raw values are interpolated – though in
many items the raw unit value may be missing the example shown, interpolation is only used
in a given month, perhaps because no one for months when the survey was not running.
interviewed in, say, February 2009 in the rural
CPI Price data
NPS sample had purchased Irish potatoes in
the last 7 days. Even when unit values are Price data for the official NBS Consumer Price
available, monthly medians tend to be quite Index is collected on a monthly basis from all
volatile due to the small sample sizes in a regional capitals. A minimum of three
given month, particularly for the NPS. quotations is solicited for each price in each
location, and for certain staple goods actual
To overcome this volatility – which may be
purchases are made. The clear shortcoming
spurious due to sampling errors rather than
of the official CPI data for poverty
genuine price fluctuations – the second step
measurement is that it reflects urban prices
taken to convert unit values into usable prices
only. However, for the bulk of non-food
is to compute a moving average of the price
items, this is the only price data available
series for each item and stratum. The moving
which is comparable over time. While rural-
average uses a 1-4-6-4-1 symmetric envelop
urban price differentials clearly exist, this will
(i.e., a weight of 1 on the unit values from two
only pose a problem for the analysis if price
periods back, a weight of 4 on the unit values
changes (i.e. inflation) differ between rural
from the previous period, 6 on the current
period, 4 on the subsequent period, and 1 on
76
and urban areas as well.20 Furthermore, for imputed values on the overall price index.
many infrequently-purchased items, goods
are unlikely to be available within a given First, if prices for a given item are reported
village, and rural residents will likely purchase within the same region both before and after
these non-food items from urban centres. In the missing value, the price is interpolated
this case, regional prices provide a relevant based on a logarithmic trend.
estimate of non-food inflation even for rural Second, if at least one price observation exists
households. for a given item in a given region, but missing
To construct price deflators for the NPS, the values extend from the beginning or to the
raw CPI data (at the level of prices for end of the series, prices are extrapolated as
individual items in a given region and month) follows. Median inflation rates for the item
was re-analyzed. This re-analysis and re- are constructed across all regions with data
aggregation was necessary because (i) the available. These median inflation rates are
weights on various items used to compute the then used to cast prices backward or forward
official CPI may have changed since they were in time for regions with missing data.
last calibrated in 2001, and (ii) primarily for Third, if no price observations are reported for
consistency with earlier HBS poverty a given item in a given region, a subjective
numbers, the NPS relies on a Fisher ideal assessment is made. Items with very few
index rather than a Laspeyre’s index as used price observations nationally and which
in the published NBS CPI series. correspond to items in the survey data with a
Aggregating the raw CPI price data to very small budget share are dropped entirely
construct a new Fisher index poses a number from the index. (On this basis the following
of data challenges. Two key challenges are: (i) items are dropped from the raw CPI data:
dealing with missing values for a given price in domestic servants’ wages, 19- and 20-inch
a certain region and month, and (ii) matching televisions, and entry fees to football
and aggregating the specific prices recorded matches.) For the remaining items, prices for
for the CPI to the level of the broader regions with no data are assigned the median
categories of consumption recorded in the national price at the beginning of the series
(Jan 2007) and median, item-specific inflation
survey data.
rates are used to extrapolate prices forward,
To construct the Fisher index, missing values as described in (ii).
are imputed in three steps. These steps were
designed to minimize the influence of Raw CPI price data was used for 27 months
(Jan 2007 to March 2009) for 20 regional
Table 39. Imputation of missing price data from the raw CPI data files
Data source Observations %
Original data, not imputed 128,877 92.86%
(i) Interpolated using prices on the same item within the same region 2,290 1.65%
(ii)
20 Backward extrapolation using national inflation for the item 1,533 1.10%
Insert a comment on what the data shows in
(ii)
termsForward extrapolation
of rural-urban using
differences in national inflation
price levels and for the item 1,841 1.33%
(iii) Regional
changes values
for food, imputed
where we canfrom national
compare levels
with the 4,239 3.05%
survey data.
Total 138,780 100.00%
77
collection points and 257 items, giving an Having matched NPS items to HBS items, and
observation count of 138,780. Table 39 gives both of these to CPI information, it is possible
a breakdown of the sources of imputation. It to produce a weighted index of prices, using
is worth noting that imputed values are the budget shares from the household surveys
disproportionately used for items with a very to weight the price information from the CPI
small budget share, as reflected in NBS (in the case of non-food items) and from the
regional offices’ inability to locate the items surveys themselves (in the case of food
for sale. As a result, they will have a relatively items).
minor influence on the overall price index.
Methodology for measuring price
differences: Fisher’s ideal index
As shown in the main report, food prices in
Correspondence of consumption items Tanzania have risen dramatically in recent
across surveys years. However, underlying this index,
The first task in comparing prices over time is different prices have moved at different rates.
to align consumption items between the NPS Generally, as prices for an item go up
2008/09 and HBS 2007. consumers consume less of that item. This
presents the fundamental challenge of
In the calculations, unit values and budget
creating a price index: how to weight different
shares for each stratum and each month are
items in the consumption basket when their
used. However, to give a rough idea of how
relative importance is shifting over time?
well it is possible to match the items in the
two surveys, and the movements in prices and The standard approach used by statistical
budget shares between the two surveys, offices around the world in calculating a CPI is
Table 41Table 46 give median prices and to rely on a Laspeyres index. This is the
budget shares for each item in the procedure used by NBS for the official CPI
consumption basket, broken down by series. The Laspeyres index weights prices by
stratum. their budget shares in the baseline period. If
denotes the value spent on good i in
A full list of the correspondence between HBS
period 0, and is its price, then the
2007 and NPS 2008/09 items was also
Laspeyre’s index is calculated as follows:
prepared (due to space constraints, it is not
reproduced here).
The second task in comparing prices over time

is to line up budget shares from the surveys This approach – which does not require new
(HBS and NPS) to price data from the NBS CPI information on budget shares in period t – is
surveys conducted on a monthly basis in each particularly useful when price data is collected
regional capital. This involves careful and reported regularly, but information on
matching of each HBS codes to the consumer habits and budget shares is only
corresponding item in the CPI questionnaire collected infrequently (e.g., monthly CPI
employed by regional statistical managers. releases versus the HBS conducted every 5
(Due to the length of this list, it is not years). However, as consumers substitute
reproduced here.) away from goods which become more
expensive, the Laspeyre’s index will tend to
78
overstate the true inflation experienced by
households.
An alternative approach is to use a Paasche In specific economic models (where

index, which uses budget shares in the final consumers have quadratic utility functions)
period to weight the prices of different items. the Fisher index can be shown to provide the
This is given by: “true” cost of living index. On a more
practical level, the Fisher index avoids
extreme values which may be produced by
either of the alternatives.
The problem here is perfectly symmetrical to The main weakness of the Fisher index is that
the challenges with the Laspeyres index: if it requires data on both prices and budget
consumers substitute away from goods whose shares at multiple points in time. In the
price rises, the Paasche index will current analysis, however, such information is
underestimate the true level of inflation by available, and thus the Fisher index is
focusing only on consumers final consumption employed. This choice will explain, in part,
patterns. the divergence shown below between the
price deflators used here for poverty analysis,
On theoretical grounds, the preferred and the official CPI series. It should in no way
approach is to use a combination of the imply that one approach is better than the
Laspeyres and Paasche indices. One possible other: each is tailored to the particular task at
combination is the Fisher index, which is the hand.
geometric mean of the Laspeyres and Paasche
indices.
79
Comparing alternative methodologies in Combining these options in terms of
practice methodology and data source, and
Readers familiar with the pattern of inflation distinguishing food from non-food prices,
in Tanzania will be attentive to the fact that produces 5 different indices that are relevant
the numbers presented in the main text differ for the analysis:
from the official CPI series. This section
highlights and attempts to partially F1. A Fisher food price index using only
disentangle two reasons for that divergence: household survey data (HBS 2007 and NPS
methodology and data sources. 2009).
The previous section examined the choice of F2. A Fisher food price index using CPI
methodology for constructing a price index records.
for the NPS – a Fisher index – while Section 0
F3. The official NBS (Laspeyre’s) food CPI
outlined the various sources of price data
deflator.
available for analysis: household survey data
and CPI records, with differing availability for As explained above, for non-food
food and non-food items. consumption it is not possible to create a
price index relying exclusively on the survey
Table 40. Price inflation, annual percentage change in price indices, by strata and year a
2007 2008 2009 b

Dar es Salaam Food prices (Survey data) 9.9% 29.0% -1.1%
Food prices (CPI data) 24.4% 40.5% 25.5%
All prices 16.4% 27.8% -1.1%
Other urban Food prices (Survey data) 11.6% 32.1% 9.1%
All prices 11.9% 28.0% 7.8%
Rural Food prices (Survey data) 11.8% 23.8% 6.0%
All prices 12.5% 22.8% 5.1%
Mainland Food prices (Survey data) 11.7% 25.5% 6.5%
All prices 12.6% 24.0% 5.1%
Official CPI – Food prices 10.1% 18.6% 2.1%
Official CPI – All prices 8.6% 12.9% 3.0%
a
Inflation rates are measured from Jan of the year listed to Jan of the following year. (This choice of periods, rather than
Dec to Dec, is due to the lack of survey data for Dec 2006).
b
For 2009, the inflation rates reflect changes in the price indices from Jan 2009 to Sept 2009 for survey-based prices, and
Jan 2009 to July 2009 for prices taken from the official CPI data. All figures are converted to annual rates of change.
80
data from the HBS and the NPS. Thus there Laspeyres index. However, the opposite is
are just two options to examine here: true here. It appears that other differences in
the methodology (possibly the use of new
NF1. A Fisher non-food price index using CPI budget shares from the 2007 HBS and the
records. NPS, as opposed to the 2001 weights used in
NF2. The official NBS (Laspeyre’s) non-food the official CPI) are pushing the survey-based
CPI. Fisher index up relative to the official CPI.
The second factor that appears to explain the

Table 40 presents each of these indices side differences in overall inflation rates is the
by side, separately for each geographic smaller non-food consumption basket used in
stratum in Mainland Tanzania. the NPS analysis. Again, the items in this
basket are limited to those which can be
As a starting point for comparing the indices, linked between the HBS and NPS.
notice that in the bottom panel the overall
inflation rate for Mainland Tanzania reported The table shows that the divergence in the
in the official CPI for 2008 was 12.9%, while in overall index (24.0% vs 12.9% for 2008) is
the current analysis it is 24.0%. The pattern of somewhat greater than the discrepancy in
results in the rest of the table suggest that food prices only (25.5% vs. 18.6%). While no
two things explain this gap. separate non-food aggregate index is
produced as part of the official CPI, the
First, with respect to food prices, the table implication here appears to be that the
shows that it is the difference in methodology narrower set of items included in the NPS
between the official CPI and the present basket (focused, as noted in the main text,
analysis that explains the different results – disproportionately on fuel prices) have had
not the different price data used. When more rapid price rises than the broader basket
applying the Fisher methodology and new used for the official CPI. An alternative
budget shares and weights to the raw, official, explanation is that the new budget shares
CPI food price data, the result is an even drawn from the NPS place higher weight on
higher rate of food price inflation: 30.8% in items with rapid price increases, however this
2008, compared to 25.5% in the survey data would be counter-intuitive in terms of the
and just 18.6% in the officially released underlying economics (with consumers
figures. substituting toward goods whose price is
rising).
This result is somewhat anomalous. It would
be expected that the Fisher index produce
somewhat lower inflation than the official
81
Table 41. Average budget shares and prices underlying the price indices, 2007 and 2009.
(Dar es Salaam)
Budget share Price Price ratio

Description NPS codes 2007 2009 2007 2009 2007/2009
rice 101-102 13.7% 14.4% 800 1,200 1.50
maize 103-105 11.1% 11.0% 403 702 1.74
millet 106-107 0.1% 0.6% 906 1,154 1.27
cassava 201-202 1.2% 1.1% 351 452 1.29
other carbs 203-207 2.5% 3.3% 469 655 1.40
meat 108-112 12.5% 8.2% 1,070 1,210 1.13
fish 401 5.9% 4.9% 821 1,300 1.58
vegetables 601-603 10.8% 12.1% 805 912 1.13
fruits 701-704 3.7% 3.8% 683 657 0.96
dairy 901-903 0.9% 2.0% 1,036 1,407 1.36
fats 1001-1002 4.7% 5.5% 1,837 2,731 1.49
eggs 807 0.4% 1.1% 4,000 4,000 1.00
bread, etc 801-806 7.6% 11.4% 2,567 3,906 1.52
pulses 808-810 8.0% 5.5% 1,886 2,882 1.53
sugar etc. 301-303 7.4% 5.6% 1,124 1,350 1.20
nuts and seeds 501-505 3.6% 4.3% 659 743 1.13
salt and spices 1003-1004 0.6% 0.8% 550 1,045 1.90
drinks 1101-1105 5.3% 4.2% 1,818 1,839 1.01
(Other Urban)

rice 101-102 10.5% 12.5% 720 988 1.37
maize 103-105 17.6% 16.7% 337 553 1.64
millet 106-107 0.9% 3.6% 664 891 1.34
cassava 201-202 3.3% 2.7% 268 330 1.23
other carbs 203-207 4.5% 5.2% 280 495 1.76
meat 108-112 4.9% 5.2% 1,086 1,563 1.44
fish 401 6.2% 4.9% 767 1,200 1.57
vegetables 601-603 11.5% 9.5% 631 877 1.39
fruits 701-704 2.9% 2.8% 496 665 1.34
dairy 901-903 1.2% 1.9% 502 670 1.34
fats 1001-1002 5.8% 4.8% 1,881 2,246 1.19
eggs 807 0.2% 0.6% 3,000 4,000 1.33
bread, etc 801-806 8.1% 9.6% 2,156 3,421 1.59
pulses 808-810 9.7% 5.8% 1,365 2,112 1.55
sugar etc. 301-303 7.0% 6.2% 1,252 1,345 1.07
nuts and 1.9% 4.3% 870 747 0.86
seeds 501-505
salt and 0.9% 0.9% 474 694 1.46
spices 1003-1004
drinks 1101-1105 2.8% 2.7% 2,081 2,335 1.12
83
(Rural)

rice 101-102 13.7% 14.4% 800 1,200 1.50
maize 103-105 11.1% 11.0% 403 702 1.74
millet 106-107 0.1% 0.6% 906 1,154 1.27
cassava 201-202 1.2% 1.1% 351 452 1.29
other carbs 203-207 2.5% 3.3% 469 655 1.40
meat 108-112 12.5% 8.2% 1,070 1,210 1.13
fish 401 5.9% 4.9% 821 1,300 1.58
vegetables 601-603 10.8% 12.1% 805 912 1.13
fruits 701-704 3.7% 3.8% 683 657 0.96
dairy 901-903 0.9% 2.0% 1,036 1,407 1.36
fats 1001-1002 4.7% 5.5% 1,837 2,731 1.49
eggs 807 0.4% 1.1% 4,000 4,000 1.00
bread, etc 801-806 7.6% 11.4% 2,567 3,906 1.52
pulses 808-810 8.0% 5.5% 1,886 2,882 1.53
sugar etc. 301-303 7.4% 5.6% 1,124 1,350 1.20
nuts and 3.6% 4.3% 659 743 1.13
seeds 501-505
salt and 0.6% 0.8% 550 1,045 1.90
spices 1003-1004
drinks 1101-1105 5.3% 4.2% 1,818 1,839 1.01
84
Table 44. Average budget shares for non-food items, 2007 and 2009.
(Dar es Salaam)
Budget share
NPS Code Description 2007 2009
6 Medical expenses 1.9% 6.3%
7 Educational expenses 2.5% 9.5%
8 Food outside home 16.3% 30.4%
101 Cigarettes or tobacco 1.1% 0.6%
102 Matches 0.4% 0.3%
103 Public transport 14.7% 14.0%
201 Kerosene 9.2% 3.5%
202 Electricity, including electricity vouchers 0.6% 2.6%
203 Gas (for lighting/cooking) 0.0% 0.3%
204 Water 7.1% 1.2%
205 Petrol or diesel 0.5% 1.0%
206 Cell phone voucher 5.1% 9.8%
207 Charcoal 31.8% 8.3%
209 Bar soap (body soap or clothes soap) 2.8% 1.5%
210 Clothes soap (powder) 2.7% 2.5%
211 Toothpaste, toothbrush 0.9% 0.9%
212 Toilet paper 0.0% 0.1%
213 Glycerin, Vaseline, skin creams 0.7% 0.6%
214 Other personal products (shampoo, razorblades, cosmetics, hair 0.6% 2.8%
products, etc.)
215 Household cleaning products (dish soap, toilet cleansers, etc.) 0.4% 0.2%
216 Light bulbs 0.0% 0.1%
217 Phone, internet, postage stamps or other postal fees 0.2% 0.5%
218 Donation - to church, charity, beggar, etc. 0.0% 1.5%
221 Wages paid to servants 0.0% 0.7%
302 Linen - towels, sheets, blankets 0.2% 0.4%
304 Mosquito net 0.1% 0.1%
306 Sports & hobby equipment, musical instruments, toys 0.1% 0.2%
307 Film, film processing, camera 0.0% 0.0%
85
(Other urban)
Budget share
102 Matches 1.6% 1.7%
201 Kerosene 20.6% 7.9%
204 Water 7.7% 0.8%
207 Charcoal 19.9% 8.5%
products, etc.)
86
(Rural)
Budget share
102 Matches 3.4% 3.1%
201 Kerosene 31.5% 9.6%
204 Water 2.7% 0.1%
207 Charcoal 1.1% 0.9%
products, etc.)
87
Appendix C. Comparing labour market
indicators across various household
surveys, 1991-2009
Box 2 in the main report compares labour below presents the precise question used
market indicators – including both sector of from each survey to create indicators of
employment and earnings levels – across occupational status (farmer, public-sector
three different sets of surveys: the Household wage employee, private-sector wage
Budget Surveys conducted in 1991, 2000 and employee, self-employed or none) and to
2007; the Integrated Labour Force Surveys, measure earnings for both categories of wage
conducted in 2001 and 2006, and the NPS employees as well as the non-farm self-
round 1, 2008/09. employed. Additionally, to resolve any
ambiguity about which occupation codes
This appendix provides details on the were assigned to which of these five broad
definitions of the indicators used for the occupational categories, the STATA code used
analysis in Box 2, as well as Tables XXX and to create the indicators in the 2007 HBS, 2006
XXX. As noted in the main text, the question ILFS and the NPS are included in the table as
phrasing and answer categories differ across
well.
the three surveys. For convenience, the table
88
Table 47. Question phrasing for key labour market indicators in the HBS, ILFS and NPS
HBS 2007 ILFS 2006 NPS 2008/09
(Phrasing is nearly identical for 1991 (Phrasing is nearly identical for 2001)
and 2000)
Sector of Section 2, Question 17: Form 2, Question 18(a), section on Section B, Question 10:
employment: What type of work did [NAME] do last “main (only) economic activity”: What was [NAME]’s main occupation for the past 12 months?
Question week? (l2q18a) (sbq10)
phrasing 1. Government employee Are you working as: 1. Agriculture/livestock
2. Parastatal / private employee 1. A paid employee. 2. Fishing
A self-employed (non-agric): 3. Mining
3. NGO employee 2. …With employees 4. Tourism
4. Religious organization 3. …Without employees Employed
employee 4. Unpaid family helper (non- 5. …Government
5. Other sector employee agric.) 6. …Parastatal
6. Self-employed (with others) 5. Unpaid family helper (agric.) 7. …Private sector
6. On your own farm or shamba. 8. …NGO/religious
7. Self-employed (alone) Self-employed (not agriculture):
8. Household duties Form 2, Question 22: 9. …With employees
9. Fishing Is this enterprise: (l2q22) 10. …Without employees
10. Livestock keeping 1. Central government 11. Unpaid family work
11. Crop farming 2. Local government 12. Job seekers
12. Forests keeping 3. Parastatal 13. Student
13. Unpaid family helper 4. Political party 14. Disabled
5. Partnership registered 15. No job
6. NGO 16. <5 years old
7. Religious organization
8. Co-operative - registered *Note: responses to section B were generally given by a
9. International/regional single respondent for all household members. Thus
organization responses from section E, administered one-to-one were
10. Household - fetching water & used to override Section B in cases where (a) an individual
HBS 2007 ILFS 2006 NPS 2008/09
and 2000)
collection of firewood for home was classified as not working in section B, but (b) reported
use wage or self-employed work in section E through the
11. Household - other economic following questions:
activities
12. Cooperative unregistered Section E, Question 9 (seq9):
13. Private own account Did you do any wage work during the last 7 days? (i.e. work
14. Private sector - employed for someone else for pay)
15. Partnership un-registered
16. Other private Section E, Question 10 (seq10):
Did you do any wage work during the last 12 months? (i.e.
work for someone else for pay)
Section E, Question 22 (seq22):

Did you operate any business or do any self-employed activity
during the last week, other than agriculture?

Did you operate any business or do any self-employed activity
during the last 12 months, other than agriculture?

In the last 7 days, how many hours did [NAME] spend on
household agricultural activities (including livestock or
fishing, whether for sale or for household food)?
Responses > 20 hrs were coded as farmers if otherwise not
working.
Sector of . generat occup = 4 if . generat occup =. . generat occup =.
q17b11_moccup == 1 . replace occup = 1 if l2q18a==5 | . replace occup = 1 if sbq10==1 |sbq10==2
employment: . replace occup = 3 if l2q18a==6 . replace occup = 2 if sbq10==9 |sbq10==10
STATA code to q17b11_moccup == 2 . replace occup = 2 if . replace occup = 3 if sbq10==3 |sbq10==4 |sbq10==6
90
HBS 2007 ILFS 2006 NPS 2008/09
and 2000)
create key . replace occup = 4 if l2q18a==2|l2q18a==3 |sbq10==7 |sbq10==11 |sbq10==12
q17b11_moccup == 3 . replace occup = 3 if l2q18a==1 & . replace occup = 4 if sbq10==5 |sbq10==8
indicators in . replace occup = 4 if (l2q22 ==5 | l2q22 ==14 | l2q22 ==15 . replace occup = 5 if sbq10==11 |sbq10==13 |sbq10==14
report q17b11_moccup == 4 | l2q22 ==96) |sbq10==15 |sbq10==16
. replace occup = 3 if . replace occup = 4 if l2q18a==1 & . gen occup =.
q17b11_moccup == 5 (l2q22 ==1 | l2q22 ==2 | l2q22 ==4 | . replace occup = 1 if sbq10==1 |sbq10==2
. replace occup = 2 if l2q22 ==7 | l2q22 ==8 | l2q22 ==9) . replace occup = 2 if sbq10==9 |sbq10==10
q17b11_moccup == 6 . replace occup = 5 if l2q08==1 | . replace occup = 3 if sbq10==3 |sbq10==4 |sbq10==6
. replace occup = 2 if l2q08==2 |sbq10==7 |sbq10==11 |sbq10==12
q17b11_moccup == 7 . replace occup = 4 if sbq10==5 |sbq10==8
. replace occup = 5 if . label define occup 1 "self employed . replace occup = 5 if sbq10==11 |sbq10==13 |sbq10==14
q17b11_moccup == 8 - farming" 2 "self-employed - other" |sbq10==15 |sbq10==16
. replace occup = 1 if 3 "wage employed - private" 4 "wage
q17b11_moccup == 9 employed - non-private" 5 . replace occup = 1 if occup==5 & (seq46>20 &
. replace occup = 1 if "unemployed/not active" seq46!=.)
q17b11_moccup == 10 . label values occup occup . replace occup = 2 if occup==5 & (seq22==1 |
. replace occup = 1 if . tab occup, g(occupd) seq23==1)
q17b11_moccup == 11 . replace occup = 3 if occup==5 & (seq12==3 |
. replace occup = 1 if seq12==9)
q17b11_moccup == 12 . replace occup = 4 if /*occup==5 &*/ (seq12==1 |
. replace occup = 5 if seq12==2 | seq12==4 | seq12==5 | seq12==6 | seq12==7 |
q17b11_moccup == 13 . replace occup seq12==8)
= 5 if q17a1_mtasco == 9998 . replace occup = . if seq3==1 & occup==5
. label define occup 1 "self . replace occup = . if seq4==1 & occup==5
employed - farming" 2 "self-
employed - other" 3 "wage employed . label define occup 1 "self employed - farming" 2 "self-
- private" 4 "wage employed - non- employed - other" 3 "wage employed - private" 4 "wage
private" 5 “unemployed/not active" employed - non-private" 5 "unemployed/not active"
. label values occup occup . label values occup occup
. tab occup, g(occupd) . tab occup, g(occupd)
Wage Section 7, Annual household income: Form 2, Question 58(b): (l2q58b) Section E, Question 18
earnings: Give details of household income What was your gross cash income from (seq18_1, _2)
Question during the previous 12 months your paid employment last month (all How much was your last payment? IF RESPONDENT HAS NOT
phrasing excluding the survey month. paid)? YET BEEN PAID, ASK: What payment do you expect? What
period of time did this payment cover?
[Section is administered at the 1. Hour
household level, but disaggregates 2. Day
income sources by member and type 3. Week
of income as follows] 4. Fortnight
91
HBS 2007 ILFS 2006 NPS 2008/09
and 2000)
5. Month
Income from employment (in cash) 6. Quarter
• Line/member number 7. Half year
• Name of household member 8. Year
• Amount in Shillings
Income from employment (in kind)
[same questions]
Self-employed Same as above: Form 2, Question 59: Section E
earnings: (l2q59b, c, d) (seq35, 36, 38, 39 40):
Question Income from non-farm self-
phrasing employment: (b) What gross income/takings did you 35. What gross income/takings did you get from your
• Line/member number get from your business or businesses business or businesses last week/month?
• Name of household member last week/month? Period: Week 1, Month 2
• Amount in Shillings Period: Week 1, Month 2
36. What was your net income (profit) from your business or
(c) What were your expenses to earn businesses last week/month?
this money? Period: Week 1, Month 2
Period: Week 1, Month 2
38. What is/was your total expenditure on wages last
(d) Your net income from your business month?
or businesses is thus: (b) – (c) in the last
week/month 39. What was your total expenditure on raw materials last
Period: Week 1, Month 2 month?
*Note, during entry and data cleaning, 40. How much were your other operating expenses (for this
the response to (d) was forced to business) such as fuel, kerosene, electricity, etc. last month?
correspond to the difference between
(b) and (c). *Note: for consistency with the ILFS, net profit was calculated
as the difference between gross revenue and expenses.
92

NPS Report Year 1

Uploaded by

Copyright:

Available Formats

NPS Report Year 1

Uploaded by

Document Information

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

NPS Report Year 1

Uploaded by

Copyright:

Available Formats

Tanzania National Panel

United Republic National Bureau

Table 1. National Panel Survey 2008/09 Sample Design .......................................................................1

Rather than simply tabulating the data from

Table 2. Cluster 1 MKUKUTA indicators available in the NPS

1.2 A shorter list of food items

1.3 Different treatment of home-

The HBS questionnaire asks respondents to

This box presents preliminary estimates based

All prices Food prices Non-food prices

Food Prices Non-Food Prices All Prices

Mainland DSM Other Urban Rural

households – as rural households eat relatively

2007 2008 2009 b

All prices Food prices only

females by the standard definition. However,

men show slightly higher rates of

Self-employed earnings Private sector wages Public sector wages

table is restricted to non-agricultural wage 142

Turning to wage levels, the public sector

None Farm Wage-Pub W age-Pri Self

occupation, and earnings are

Table 13. Use of improved seeds

Table 14. Use of modern farming methods

Goal 6: Provision of reliable and

Table 18. % increase in number of

Table 20. Cluster 2 MKUKUTA indicators available in the NPS

“Moderate stunting” measures the proportion Moderate Severe

for-age, and weight-for-height.14 The

Figure 7 gives a more precise look at the

Table 24 shows the proportion of births

Table 24. Proportion of births attended by various service providers

Table 26. Toilet type (%), by stratum and year

Table 29. Net primary school enrolment

Among children in the age-range used to

This provides a quite small sub-sample o f

Proportion of orphaned Father alive, mother alive

Table 30 shows the primary school enrolment

Taking the results at face value, the enrolment

Table 30. Net primary school enrolment

Table 31. Cluster 3 MKUKUTA indicators available in the NPS

% who have been abused

% of those abused who

• Housing and amenities. The HBS and NPS

where is an error term with total variance

In practice (3) is estimated with household

Finally, because estimates of the headcount

CDF of Actual & Predicted Values

As with the main consumption model, the

variance model in equation (4) is estimated

variances for the 2009 NPS sample.

.5 Estimating the model

In many applications of Small Area Estimation,

Lacking firm theoretical grounds on which to

The center and bottom panels of Table 38

In brief, the results in the bottom parts of

Unit values for non-food consumption are not

Maize grain Irish potatoes

The second task in comparing prices over time

An alternative approach is to use a Paasche In specific economic models (where