Consumption Bundle Aggregation in Poverty Measurement : Implications for Poverty and its Dynamics in Uganda

TL;DR: In this article, the authors estimate a new set of poverty lines that accounts for the shortcomings of the official poverty figures in Uganda by the fact that the underlying poverty lines are based on a single national food basket constructed in the early 1990s.

Abstract: Official poverty figures in Uganda are flawed by the fact that the underlying poverty lines are based on a single national food basket that was constructed in the early 1990s. In this paper, we estimate a new set of poverty lines that accounts for the wid

Summary

1 Introduction

• During the past few decades, Uganda has experienced substantial economic growth.
• This growth has been attributed to the new government that has implemented a far-reaching economic reforms agenda, transforming Uganda into one of the most liberal economies in Africa south of the Sahara.
• Using the same methods the authors used to replicate o cial poverty gures in previous rounds of the UNHS, they estimate national headcount poverty to be 19.5 per cent in the UNHS 2012/13.
• To account for di erences in diets in di erent locations, the authors will construct di erent poverty thresholds for di erent spatial domains using the latest available nationally representative household survey (UNHS 2012/13).
• In other words, the authors will construct a detailed poverty pro le that takes dynamic aspects into account, de ning groups based on poverty transitions instead of a simple dichotomous poor/non-poor status (Boateng et al. 1992).

2 Poverty in Uganda: Trends and controversies

• During the 1990s, poverty in Uganda decreased substantially, falling by almost 40 per cent at the national level.
• This is because of the relatively high quality data available for this period.
• Even after correcting for income di erence, as regions that consume more expensive calories may do so simply because they have higher incomes, Appleton (2003) comes to the conclusion that the western region overtakes even the northern region as the poorest.
• Daniels and Minot (2014) use Demographic and Health Survey (DHS) data and methods related to poverty mapping and small area estimation to look at poverty trends across Uganda from 1995 to 2010.
• They nd that poverty indeed reduced over time, but much slower than o cial gures suggest.

3 Utility-consistent poverty lines using revealed

• From the above, the authors learn that one of the main weaknesses of the o cial poverty measures is that they are based on a poverty line that is constructed using a single food commodity bundle for the entire country.
• It is well known that in many instances - for example, if relative prices of basic commodities vary by region (or through time) and preferences permit substitution - the use of a single consumption bundle may yield inconsistent poverty comparisons (Tarp et al. 2002).
• The theory underlying absolute poverty lines is grounded in welfare economics and constrained utility maximization.
• Thus, any other bundle that yields the same level of utility (such as, for instance, the one chosen by the representative consumer in region r' ) should be equally expensive as or more expensive than the chosen bundle.
• One approach, which the authors will use in this paper, uses a minimum cross-entropy approach to adjust expenditure shares such that they meet revealed preference conditions (Arndt and Simler 2010).

4 A reassessment of poverty in Uganda

• The authors will mainly work with the four waves of the Uganda National Panel Survey (UNPS).
• This will then be compared to an analysis based on six separate spatial domains 7No UNPS survey has been done in 2012/13.
• The mapping from these basic caloric needs into basic needs consumption bundles is based on FAO (1986).
• The underlying poverty lines for the six spatial domains, in addition to a poverty line using only one spatial domain for comparison.

5 A pro le based on poverty dynamics

• Now that the authors developed a new set of poverty lines above, in this section, they will use the Uganda National Panel to construct pro les for di erent categories of households based on the evolution of their poverty status over time.
• These are households that are non-poor in all past waves but poor in all subsequent waves.
• If the authors weigh these households by population weights, the number of chronic poor individuals increases to 12.3 per cent.
• Next, 387 households have escaped poverty and 198 have fallen into poverty, corresponding to 19.0 and 8.2 per cent of the population, respectively.
• As such, the authors will also concentrate on characteristics that change only slowly over time, as opposed to those that may change signi cantly from year to year, such as crops cultivated.

5.1 Location

• Location and well-being are often found to be correlated.
• One prominent economic reason is that in sparsely populated areas with a thin road network that is often in bad shape, transaction costs are high, affecting economic activity (Stifel and Minten 2008).
• More than 44 per cent of the people that are always above the poverty threshold live there.
• People living in the western region seem to be moving in and out of poverty more than people living in other regions.
• The median for the non-poor is about 50 minutes, as opposed to about 60 minutes for the chronic poor.

5.2 Household demographics

• The size and composition of the household are also variables that often feature in poverty regressions.
• This last feature may be captured better when using relating the di erent types of household in terms of poverty dynamics to dependency ratios.
• The underlying reasons should be sought in di erences between male-headed and female-headed households in terms of access to secure land tenure, labor, credit, technology, and extension services (e.g. Quisumbing and Pandolfelli 2010).
• Those that are never poor have small households and low dependency ratios.
• At the same time, households where the head is never married are clearly more likely to be non-poor, as are households where the head has divorced.

5.3 Activity

• While in general 35.8 per cent of Ugandans fall in the non-poor category, only 25.7 per cent of the Ugandan subsistence farmers are in the non-poor subgroup.
• It seems the group of vulnerable households is disproportionately represented within the group of subsistence farmers.
• Wage employment also seems to be an activity that is prevalent among the nonpoor.
• The most clear results are for those who mention their main source of income is property - virtually all are non-poor.
• People that depend on transfers are also non-poor.

5.4 Education

• In traditional poverty pro les, the education level of the household head is often signi cant.
• Indeed, skills are important for the self-employed, and schooled labor is likely to be better rewarded.
• Education is also among the initial characteristics associated with chronic poverty in rural communities in Ethiopia (Dercon, Hoddinott, and Woldehanna 2012).
• Table 7 looks at the highest education level reported by the household head.
• Within the group of individuals in households that have always been poor, the share of households that are headed by someone without formal education is 37 per cent.

5.5 Health

• Illness and health shocks have been reported to a ect poverty dynamics.
• At the other extreme, the authors nd that households that live in chronic poverty reported highest median distance to health facilities.
• Figure 5 looks at average days that household heads reported being inactive due to illness in the last six months in the 2005/06 UNPS wave conditional on subsequent poverty transitions.
• For most of the categories, the number of days lost is on average about 8.5 days.
• On the other hand, the households that report the highest number of days lost by the household head due to illness are those that are in the subgroup of households that eventually fall into poverty or are living in chronic poverty.

5.6 Shocks and Coping

• The poor are known to be more vulnerable to shocks, due to their lower ability to insure (Dercon 2004).
• The authors also look at how the households deal with shocks ex post conditional on their wealth dynamics category.
• A substantial share of the non-poor report to have been exposed to drought shocks, but this share is only about 5 percentage points lower than the overall share that reports drought-related shocks.
• Bad seed quality is reported more among the non-poor than average.
• This category also shows up relatively more in the category of households that slide into poverty.

6 Conclusion

• The authors reassess the evolution of poverty over the past ten years in Uganda.
• In addition, this poverty line relies on a single food consumption basket for Uganda, despite the fact that Uganda consists of a diverse set of regions, each with their own diets.
• These poverty lines are then tested to check if they obey revealed preference conditions.
• The fact that in this region, relatively few households are escaping poverty and relatively more households are falling into poverty needs attention.
• It now seems that the households that slide below the poverty threshold have a surprisingly high dependency ratio.

Figures

Table 5: Marital status

Table A.1: Utility-consistent poverty lines

Figure 2: Time to fetch water

Table 4: Poverty headcounts 2002-2012 using six spatial domains

Figure 6: Coping with adverse shocks

Table A.3: Poverty severity 2002-2012 using six spatial domains

Table 6: Most important source of earnings

Figure 5: Days inactive

Figure 4: Distance to health infrastructure

Table A.2: Poverty gap 2002-2012 using six spatial domains

Figure 3: Household size and child dependency ratios

Table 3: Poverty headcounts 2002-2012 using one spatial domain

Figure 1: Regional poverty dynamics

Table 1: O cial poverty headcounts 2002-12

Table 2: Utility-consistent poverty lines based on UNHS 2012/13

Full text

World Institute for Development Economics Research wider.unu.edu
WIDER Working Paper 2014/150
Consumption bundle aggregation in poverty
measurement
Implications for poverty and its dynamics in Uganda
Bjorn Van Campenhout,
1
Haruna Sekabira,
2
and Dede H.
Aduayom
1
November 2014

1
International Food Policy Research Institute (IFPRI) – Kampala;
2
Goettingen University, Department for Agricultural Economics
and Rural Development; corresponding author: b.vancampenhout@cgiar.org
This study has been prepared within the UNU-WIDER ‘Reconciling Africa's Growth, Poverty and Inequality Trends: Growth and
Poverty Project’ (GAPP), directed by Finn Tarp.
Copyright © UNU-WIDER 2014
ISSN 1798-7237 ISBN 978-92-9230-871-1
Typescript prepared by the authors.
UNU-WIDER gratefully acknowledges the financial contributions to the research programme from the governments of Denmark,
Finland, Sweden, and the United Kingdom.
The World Institute for Development Economics Research (WIDER) was established by the United Nations University (UNU)
as its first research and training centre and started work in Helsinki, Finland in 1985. The Institute undertakes applied research and
policy analysis on structural changes affecting the developing and transitional economies, provides a forum for the advocacy of
policies leading to robust, equitable and environmentally sustainable growth, and promotes capacity strengthening and training in
the field of economic and social policy-making. Work is carried out by staff researchers and visiting scholars in Helsinki and through
networks of collaborating scholars and institutions around the world.
UNU-WIDER, Katajanokanlaituri 6 B, 00160 Helsinki, Finland, wider.unu.edu
The views expressed in this publication are those of the author(s). Publication does not imply endorsement by the Institute or the
United Nations University, nor by the programme/project sponsors, of any of the views expressed.
Abstract: Official poverty figures in Uganda are flawed by the fact that the underlying poverty
lines are based on a single national food basket that was constructed in the early 1990s. In this
paper, we estimate a new set of poverty lines that accounts for the widely divergent diets
throughout the country using the latest available household survey. Using these updated poverty
lines, we then look at poverty dynamics using four waves of the Uganda National Panel Survey.
We classify households into categories depending on their change in poverty status over time and
relate this to characteristics that are likely to change only slowly. This enables us to explore the
characteristics of households that, for instance, grow out of poverty and how they differ from
households that appear to be trapped in poverty. Our approach generates poverty measures that
are more credible from a theoretical point of view and are more in line with what other researchers
find.
Keywords: poverty, cost of basic needs, revealed preferences, Uganda
JEL classification: D31, O12, O55
Acknowledgements: This project is under revision control. All source code to replicate the
analysis can be found at: https://bitbucket.org/bjvca/wider/

1 Introduction
During the past few decades, Uganda has experienced substantial economic
growth. Since 1986, when the National Resistance Movement took over gov-
ernment, real gross domestic product (GDP) has grown at an annual rate of
6.8 per cent, making its economy one of the fastest growing in Africa. This
growth has been attributed to the new government that has implemented
a far-reaching economic reforms agenda, transforming Uganda into one of
the most liberal economies in Africa south of the Sahara. Indeed, as argued
in World Bank (1993: 22), the government 'liberalized the trade regime by
abolishing export and import licensing; dismantled all price controls, which
were few to begin with; repealed the Industrial Licensing Act, promulgated a
new investment code, returned properties expropriated by the Amin regime
and commenced privatizing public industrial enterprises; made important
strides in abolishing export and distribution monopolies; embarked upon a
major overhaul of the civil service; restructured the tax system and improved
tax administration; and has made an impressive start in restructuring pub-
lic expenditures towards critical economic and social services'. Such policy
changes were seen as essential preconditions for sustainable economic growth.
This growth has been accompanied by equally impressive declines in the
levels of poverty as reported by the government. While aggregate headcount
poverty stood at about 57 per cent in the early 1990s, the most recent ocial
estimate puts 19.5 per cent of the population below the ocial poverty line.
1
But despite these successes at the aggregate level, researchers warn that
1
For the most recent ocial estimate, we take the poverty estimate based on the
2012/13 Uganda National Household Survey (UNHS). The data on which these estimates
are based were obtained from the Uganda Bureau of Statistics (UBOS) in August 2014.
As is mostly the case with UNHS data obtained from UBOS, the dataset came with a
compiled welfare aggregate based on consumption expenditure and a set of ocial poverty
lines. Using the same methods we used to replicate ocial poverty gures in previous
rounds of the UNHS, we estimate national headcount poverty to be 19.5 per cent in the
UNHS 2012/13. This is slightly lower than ocial poverty estimates at the time of the
UNHS 2012/13 dissemination and reported in the press (22.1 per cent).
1

this growth has not been shared equally by the population at large. For
instance, marked spatial heterogeneity in baseline poverty and subsequent
poverty reductions mean that dierences in the standard of living between
locations are often much higher now than what they used to be.
Apart from the observed heterogeneity in terms of poverty and poverty
reduction, the gures itself have been called into question as well. Some
argue that the lack of progress on assets accumulation and non-monetary
well-being proxies suggest much more modest poverty reductions, raising
suspicion about the poverty lines and the welfare indicator used by the gov-
ernment of Uganda (Daniels and Minot 2014; Kakande 2010). Some scholars
have also been questioning the spatial pattern of poverty as reported in of-
cial documents, arguing that a single national food poverty line is likely
to overstate poverty in some areas while underestimating poverty in others
(Appleton 2003; Jamal 1998).
In this paper, we want to update existing knowledge about the state of
poverty and its dynamics in Uganda, while at the same time address some
of the problems with the ocial gures that have been identied in recent
studies. To account for dierences in diets in dierent locations, we will
construct dierent poverty thresholds for dierent spatial domains using the
latest available nationally representative household survey (UNHS 2012/13).
For each spatial domain, we construct a food basket that produces a certain
minimum of calories that reects the diets of the poorest households in that
region. These baskets are then multiplied by prices prevailing in that region
to arrive at food poverty lines. An allowance for basic non-food necessities is
then added to get a set of Cost of Basic Needs (CBN) poverty lines, one in
each spatial domain. We then test these poverty lines to see if they are utility-
consistent. The idea is that a basic needs bundle in a certain spatial domain
A should always be cheaper than a bundle from any other region valued
at prices of region A. If a bundle does not satisfy these revealed preference
conditions, we use an information theoretic approach to adjust the bundles
2

until they do, as outlined in Arndt and Simler (2010).
While comparing poverty estimates using these new poverty lines with
the ocial estimates is interesting in its own right, we will use the utility-
consistent poverty lines to look at poverty dynamics using the recently re-
leased Uganda National Panel Survey (UNPS). The UNPS is a yearly panel
survey collected by the Uganda Bureau of Statistics (UBOS) supported by
the World Bank's Living Standard Measurement Study (LSMS) project that
tracks about 3,000 households. We can use the panel nature of this survey to
study how many households are chronically poor (dened as always falling
below the poverty threshold) and investigate how their characteristics dier
from other groups, such as households that successfully escaped poverty. In
other words, we will construct a detailed poverty prole that takes dynamic
aspects into account, dening groups based on poverty transitions instead of
a simple dichotomous poor/non-poor status (Boateng et al. 1992). As for the
characteristics we contrast within each group, we conne ourselves to those
that change only slowly over time, and we look at the initial conditions
at the start of the panel. We hope that this can enlighten us on the pre-
conditions that need to be in place to be in a particular poverty dynamics
group.
2
The remainder of this paper is structured as follows: We rst give an
overview of poverty in the past few decades and look at the present ocial
poverty estimates in Uganda. We also present some studies that point to
shortcomings in ocial poverty measurement. In Section 3, we briey ex-
plain the reasoning behind the use of spatially disaggregated poverty lines
and the role of revealed preferences to test for utility consistency. Section
4 presents the construction of the new set of poverty lines and our poverty
estimates based on the UNHS 2012/13. Section 5 then looks at poverty dy-
namics and relates households with diering poverty dynamics to a selection
2
For example, we could check if households that are always below the poverty line are
dierent in terms of their reported ability to cope with adverse shocks at the start of the
panel from those that grow out of poverty over the course of the panel.
3

Frequently asked questions

Q1. What are the contributions in "Implications for poverty and its dynamics in uganda" ?

In this paper, the authors used the most recent available nationally representative household survey ( UNHS 2012/13 ) to construct a food basket that produces a certain minimum of calories that re ects the diets of the poorest households in that region.