Jump to ContentJump to Main Navigation
Measuring Poverty and Wellbeing in Developing Countries$

Channing Arndt and Finn Tarp

Print publication date: 2016

Print ISBN-13: 9780198744801

Published to Oxford Scholarship Online: January 2017

DOI: 10.1093/acprof:oso/9780198744801.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy).date: 16 November 2018

Estimating Utility-Consistent Poverty in Ethiopia, 2000–11

Estimating Utility-Consistent Poverty in Ethiopia, 2000–11

Chapter:
(p.55) 5 Estimating Utility-Consistent Poverty in Ethiopia, 2000–11
Source:
Measuring Poverty and Wellbeing in Developing Countries
Author(s):

David Stifel

Tassew Woldehanna

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780198744801.003.0005

Abstract and Keywords

This chapter describes how the standardized PLEASe computer code stream based on Arndt and Simler’s (2010) utility-consistent approach can be adapted to analyse poverty in Ethiopia in 2000, 2005, and 2011. Several data-related issues create challenges to estimating the spatial and temporal distribution of poverty in a manner that meets both consistency and specificity objectives. The chapter documents how the code stream is adapted to address changes in data collection periods and strata for the respective surveys over time. Changes in the duration and time of year for data collection can be especially problematic for consistency in the presence of annual inflation of over 30 per cent. In addition, the Ethiopia case provides an example of how to address convergence problems encountered when running the PLEASe code. Careful consolidation of spatial domains and limiting the number of iterations in the estimation of poverty lines are potential solutions.

Keywords:   poverty measurement, utility-consistent poverty lines, poverty dynamics, inequality, Ethiopia

5.1 Introduction

Since the turn of the century, the Ethiopian economy has experienced strong economic growth and structural improvements. Rapid infrastructure growth, increased agricultural production and commercialization, better-functioning food markets, and a strong social safety net programme are all part of the changing economic landscape (Dorosh and Schmidt 2010) that is likely to have paid dividends in terms of poverty reduction. Yet measuring these dividends in Ethiopia is complicated by conceptual and practical data-related issues. This is not surprising given the complexity of measuring poverty in a manner that is consistent over time and space, yet is also sensitive to local conditions.

There are two important measurement issues related to the consistency and specificity of poverty estimates over time and space. First, evidence that differing commodity lists (Pradhan 2000) and recall periods (Scott and Amenuvegbe 1990) affect the levels of reported consumption from household surveys highlights the importance of the comparability of the data used to construct nominal household consumption aggregates. Second, the appropriate estimation of poverty lines is also essential not only as a poverty threshold, but also as a cost-of-living index that allows interpersonal welfare comparisons when the costs of consuming basic needs vary over time and space (Ravallion 1998). The challenge is to estimate poverty lines that are consistent over time and space (i.e. the reference standard of living is fixed), and yet are also characterized by specificity in which the poverty lines reflect local consumption patterns and norms (Ravallion and Bidani 1994).

(p.56) The purpose of this chapter is to describe how the standardized PLEASe computer code stream based on Arndt and Simler’s (2010) utility-consistent approach to measuring consumption poverty can be adapted in order to analyse poverty in Ethiopia in 2000, 2005, and 2011. We document how the utility-consistent approach to spatial deflation differs from the approach undertaken by the national statistical office to produce the official poverty estimates (i.e. using consumer price indices), and how the trends in these estimates differ. Further, we highlight the importance of accounting for changes in the duration and time of year for data collection, and how this can be especially problematic for consistency in the presence of annual inflation of over 30 per cent. In addition, the Ethiopia case provides an example of the challenge of conducting revealed preference tests of the utility consistency of regionally estimated poverty lines (i.e. do the consumption patterns in other spatial domains cost no less than the own-domain consumption patterns when both are evaluated at own-domain prices) when spatial consumption patterns differ substantially.

The structure of this chapter is as follows. In section 5.2, we elaborate on the methodology and describe the primary data sources. Section 5.3 describes how the data were prepared for the exercise and how the PLEASe code was adapted for these data. In section 5.4, we present the estimates of poverty based on the utility-consistent approach to calculating poverty lines, and explore the differences between these estimates and the original estimates made by the Ethiopian Central Statistics Agency (CSA) (MoFED 2008 and MoFED 2012). Section 5.5 provides concluding remarks.

5.2 Methodology and Data

In this section, we briefly describe the methodology and primary data sources used to measure poverty and inequality in a manner that is consistent over time and space, and which is specific to local consumption patterns and norms.

5.2.1 Methodology

As with any analysis of poverty, choices need to be made regarding (i) the welfare indicator, (ii) the threshold between the poor and the non-poor, and (iii) the measure of poverty. First, in this particular analysis, we concentrate on a money measure of welfare—per capita household consumption. The household consumption aggregate that we use as our welfare indicator is constructed in a standard manner by aggregating food and non-food expenditures, the estimated value of own-produced food and non-food items and of (p.57) in-kind payments, gifts received, and the estimated use value of durable goods and housing (Deaton and Zaidi 2002).

Second, with regard to the poverty threshold, we estimate poverty lines1 for twenty spatial domains in Ethiopia (Addis Ababa, Harari, and urban and rural areas for the Afar, Amhara, Benishangul-Gumuz, Dire Dawa, Gambella, Oromiya, SNNP, Somali, and Tigray regions). Food poverty lines are estimated first, and are anchored to calorie requirements that are calculated for purposes of specificity separately for each domain based on the demographic structure and fertility patterns in the domain. This is a departure from the common practice for poverty analysis in Ethiopia of using a standard requirement of 2200 calories per person per day, with the poverty line calculated in 1995/6 and adjusted for inflation for analysis in later years. An iterative approach is used to find the least-cost consumption bundle that meets domain-specific calorie requirements and that reflects consumption patterns of the poor in the spatial domain. This provides specific initial estimates of the food poverty lines. Revealed preference tests are then conducted to test the utility consistency of these poverty lines (i.e. do the consumption patterns in other spatial domains cost no less than the own-domain consumption patterns when both are evaluated at own-domain prices). When the tests are violated, maximum-entropy methods are used to reconcile the differences so that domain specificity is maintained in the new poverty lines, while utility consistency is not violated.2 Once the region-specific food poverty lines are determined, they are scaled up by the share of non-food consumption representative of the households around the food poverty lines, to get the region-specific poverty lines.

With the welfare indicators and poverty lines in hand, we primarily employ the Foster–Greer–Thorbecke (1984) class of poverty indices to measure levels and changes in poverty. We also move beyond the use of poverty indices to analyse changes in poverty by employing standard tests of stochastic dominance. In order to do this, we note that poverty lines are more than poverty thresholds, they also serve as cost-of-living indexes that allow interpersonal welfare comparisons. As such, we use the poverty lines to map nominal household consumption to real household consumption using indexes constructed from these poverty lines (Blackorby and Donaldson 1987). Once mapped into comparable real values, the distributions of household consumption are then used to conduct dominance tests and to measure inequality.

(p.58) 5.2.2 Data

The primary data sources used in this analysis are the 1999/2000 (hereafter 2000), 2004/5 (hereafter 2005) and 2010/11 (hereafter 2011) Ethiopia Household Income, Consumption and Expenditure Surveys (HICES). The HICES, conducted by the Central Statistical Agency (CSA), are nationally representative stratified and clustered surveys that contain information on household characteristics, expenditure, activities, and infrastructure. The main objective of the HICES was to provide data on levels, distributions, and patterns of household income, consumption, and expenditures.

Given that the HICES are used to construct the household consumption aggregates for the analysis of monetary poverty, it is important to be aware of comparability issues related to them. Coverage of the three surveys is similar (major urban areas, rural regions, and other urban areas), and although the sample sizes grew from 17,332, to 21,274, to 27,830, for the 2000, 2005, and 2011 surveys, respectively, this is unlikely to affect the comparability of the welfare measures over time. There are, however, other differences in the data collection method that may be problematic. First, although the questionnaires are nearly identical, the item codes used for the expenditure/consumption recall differed for each of the three years. For example, the numbers of food codes used in the data collection process were 252, 872, and 653 in the 2000, 2005, and 2011 surveys respectively. Evidence that more detailed lists of commodity items are associated with higher levels of reported consumption from household surveys (Pradhan 2000) warrants care in interpreting changes in poverty given that the household consumption aggregates may not be entirely comparable.

Second, the change in the data collection period complicates comparability due to issues of seasonality and inflation. The 2000 and 2005 surveys were conducted in two relatively short and similarly timed rounds (July–August and January–February) during low inflation periods, whereas the 2011 survey was conducted over the course of a year (8 July 2010 to 7 July 2011) that was characterized by inflation of over 30 per cent.3 Further, it is difficult to gauge the consequences that seasonal variation in consumption patterns may have on the comparability of the 2011 consumption aggregate relative to the aggregates from the earlier surveys. As a form of sensitivity analysis, we estimated poverty lines on the subset of the sample of households in the 2011 survey who were interviewed in the same quarters as those in the 2000 and 2005 surveys. Although the poverty estimates from this subsample do not (p.59) differ substantively from those of the full sample, we remain cautious about interpreting changes in poverty between these surveys.

5.3 Application of PLEASe

5.3.1 Data Preparation

The bulk of the work in applying PLEASe to the Ethiopia household survey data was related to preparing the data themselves. The PLEASe manual (Arndt et al. 2013) provides guidance for creating standard data files with common variable names. We therefore do not elaborate on this here. But it is worth emphasizing that in following the manual it is important to pay close attention to the units (e.g. daily and metric) and to item codes when preparing the data as these have the potential to be an unnecessary source of error. In addition, certain country-specific decisions need to be made in the process of preparing the data.

For Ethiopia, the choice of the spatial domains (‘spdomain’ in ‘hhdata.dta’) and the number of iterations used to calculate initial poverty lines were complicated by convergence problems encountered when running the PLEASe code on the 2011 data. Initially, the domains were defined over the urban and rural areas in the chartered city of Dire Dawa and the nine ethnically-based and politically autonomous regional states, as well as the chartered city of Addis Ababa (only urban). But when the PLEASe code was run on the 2011 data, the program encountered problems while iterating over the poverty lines that would then be used to prepare the data for the revealed preference tests. As noted in Arndt et al. (2013), the program estimates initial poverty lines by valuing the minimum cost of consuming domain-specific calorie requirements based on the consumption patterns of the poorest X per cent households in each domain, where X is defined by the user. This process is repeated over five iterations using the poverty lines from the previous iteration as the thresholds for determining the consumption patterns of the poor households. Five iterations generally result in poverty lines and consumption patterns that converge to steady values. In some spatial domains (e.g. rural and urban Benshangul, rural Gambella, and rural and urban Harari), however, poverty dropped so low after the second iteration that there were too few poor households to calculate poverty lines. In particular, when price observations for valuing the consumption patterns of the poor households are based on only a few observations, they are dropped. Consequently, the price files for these domains were empty and food poverty lines could not be calculated. It is not clear why the data led to this problem, but two adjustments proved sufficient to resolve it. First, the convergence process was limited to one iteration. We discuss the implications of this in section 5.3.2 (p.60) in the description of the PLEASe code preparation. Second, the rural and urban areas of Harari were merged into one spatial domain. Given the relatively small spatial area that makes up Harari, this is defensible. As a consequence of the latter adjustment, we ended up with twenty spatial domains (except for the 2005 data in which there were eighteen spatial domains because there was no data for urban and rural areas of Gambella).

5.3.2 PLEASe Code Preparation

Once the data were appropriately formatted and were sufficiently cleaned, the next step was to adjust the PLEASe code for the Ethiopia case. This involved adjusting two Stata do-files located in the PLEASe directory for each survey year entitled ‘new’. Each of these files is addressed in turn.

  1. 1. ‘000_boom.do’:

Aside from setting the path so that Stata recognized the locations of the various files on the analysts’ computers, the ‘year’ needed to be set for each of the three years of the analysis. For example, when PLEASe was run on the 2005 HICES, the appropriate line of code was

global year ‘2005’

It is worth noting here that intertemporal (between survey years) revealed preference tests cannot be conducted with these data since the number of food codes changed each year (see section 5.2.2). As such, the numerical value for the variable in the PLEASe code that indicates the previous year (‘prevyear’) was left blank:

global prevyear

  1. 2. ‘010_initial.do’:

This is an important file that defines the parameters and code options used in the remainder of the PLEASe code. The instructions in this file are self-explanatory, but it is worth noting that ‘spdom_n’ was set to 20 to reflect the number of spatial domains and to correspond to the numbers in the ‘spdomain’ variable.

As noted previously, one of the adjustments made in order to address the convergence problems in the 2011 data was to limit convergence process to one iteration. This is done in the ‘010_initial.do’ file by setting ‘it_n’ to 1. As a consequence of this, care must be taken in setting the initial quantile that defines the poor for purposes of estimating the minimum cost of consuming domain-specific calorie requirements for the food poverty line. Poverty line (p.61) estimates can be sensitive to this initial threshold. Thus for the 2011 data, we cautiously set this threshold equal to the fortieth percentile…

global bottom ‘40’

Estimating Utility-Consistent Poverty in Ethiopia, 2000–11

Figure 5.1. Cumulative distributions of household per capita consumption, Ethiopia 2000–11

Source: Authors’ calculations from HICES data

The rationale for using this particular threshold was the combination of a national poverty estimate of 46.0 per cent poor in 2005 using the PLEASe code combined with indications of considerable growth between 2005 and 2011 (see Figure 5.1). Using 46.0 per cent from 2005 appeared to be too high, while using the CSA estimate for 2011 of 29.6 was likely to result in low estimates of poverty that would be open to criticism. A conservative threshold of 40 per cent is a reasonable compromise.

5.4 Poverty Estimates

Table 5.1. Utility-consistent and original CSA poverty estimates, Ethiopia 2000–11

UC Estimates

CSA Estimates

Difference

2000

2005

2011

2000

2005

2011

2000

2005

2011

National

Headcount Ratio (P0)

46.8

46.0

23.8

44.2

38.7

29.6

−2.6

−7.3

5.8

Depth of Poverty (P1)

12.6

12.3

6.3

11.9

8.3

7.8

0.7

4.0

1.5

Severity of Poverty (P2)

4.8

4.5

2.4

4.5

2.7

3.1

0.3

1.8

0.7

Urban

Headcount Ratio (P0)

39.0

22.7

13.3

45.4

39.3

30.4

6.4

16.6

17.1

Depth of Poverty (P1)

10.8

4.7

3.2

12.2

8.5

8.0

1.4

3.8

4.8

Severity of Poverty (P2)

4.1

1.5

1.2

4.6

2.7

3.2

0.5

1.2

2.0

Rural

Headcount Ratio (P0)

48.0

50.0

25.9

36.9

35.1

25.7

−11.1

−14.9

−0.2

Depth of Poverty (P1)

12.9

13.5

6.9

10.1

7.7

6.9

2.8

5.8

0.0

Severity of Poverty (P2)

4.9

5.0

2.7

3.9

2.6

2.7

1.0

2.4

0.0

Notes: ‘UC’ indicates Arndt and Simler (2010) utility-consistent poverty lines estimated with PLEASe. ‘CSA’ indicates original poverty lines calculated by CSA. The rates are all multiplied by 100.

Source: Authors’ elaboration based on data from CSA and authors’ calculations based on data from HICES

Based on utility-consistent poverty lines derived from application of the PLEASe code to the HICES data, we find that poverty rates in Ethiopia at the turn of the century were high, but that they fell substantially by 2011 (Table 5.1). In 2000, 46.8 per cent of the population was poor, compared to 23.8 per cent in 2011. Most of the decline, however, occurred between 2005 and 2011 as the poverty rate only fell by just under one percentage point (p.62) between 2000 and 2005. The more distribution-sensitive poverty measures (i.e. the depth (P1) and severity (P2) of poverty) indicate similar patterns of decline over time. That is, marginal declines in the depth and severity of poverty between 2000 and 2005 were followed by substantial improvements between 2005 and 2011. Figure 5.1 illustrates this more completely as the nearly overlapping distributions of per capita consumption for 2000 and 2005 (spatially and regionally deflated by the utility-consistent poverty lines) are first-order dominated by the 2011 distribution.

Poverty is largely a rural phenomenon, with 48.0 per cent of the rural population below the poverty line in 2000, compared to 39.0 per cent in urban areas. Although the rural headcount ratio fell by a remarkable 22.1 percentage points, urban areas as a whole saw even greater declines in poverty, as the urban poverty rate fell to under 14 per cent by 2011. Most of the decline in urban poverty took place in the first half of the decade, falling by just over sixteen percentage points. Conversely, rural poverty rose marginally during this period, with all of the gains occurring after 2005.

These utility-consistent poverty estimates differ considerably from CSA’s original estimates (MoFED 2008 and MoFED 2012). As illustrated in Table 5.1, the original national headcount ratio estimates are lower than the utility-consistent estimates by 2.6 percentage points for 2000 and by 7.3 percentage points in 2005, and they are higher by 5.8 percentage points for 2011. The urban utility-consistent poverty estimates are all lower than the CSA estimates, while the rural utility-consistent estimates are higher for 2000 (p.63) and 2005 and are nearly identical for 2011. Although the patterns are the same for the depth and severity of poverty, the differences are less stark.

Both approaches indicate that poverty fell substantially in Ethiopia over the course of the 2000s. But the utility-consistent poverty estimates suggest that poverty fell by even more than the original CSA estimates did despite using a higher initial cutoff of 40 per cent for 2011 (see section 5.3). It is worth noting, however, that the differences in the estimated declines are greater for the headcount ratios than for the distribution-sensitive poverty measures, suggesting that the two approaches estimate spatially price-adjusted real household consumption aggregates that are more similar at the lower end of the distribution than around the poverty line.

Table 5.2. Original CSA and utility-consistent poverty lines, Ethiopia 2000–11

2000

2005

2011

Orig

UC

% Diff

Orig

UC

% Diff

Orig

UC

% Diff

Addis Ababa

4.58

3.22

−29.8

5.13

2.27

−55.8

16.10

8.86

−45.0

Afar—rural

3.05

3.07

0.5

3.59

3.09

−13.9

10.58

8.89

−16.0

Afar—urban

3.05

3.27

7.3

3.59

2.68

−25.3

10.58

8.00

−24.3

Amhara—rural

2.68

2.52

−5.8

3.47

3.84

10.5

9.83

7.77

−21.0

Amhara—urban

2.68

2.78

3.8

3.47

3.31

−4.5

9.83

8.52

−13.3

Benshangul—rural

2.65

2.66

0.3

3.71

4.54

22.3

9.92

6.77

−31.7

Benshangul—urban

2.65

2.83

6.7

3.71

3.99

7.7

9.92

7.41

−25.3

Dire Dawa—rural

3.45

3.58

3.9

3.90

4.07

4.5

12.90

8.68

−32.7

Dire Dawa—urban

3.45

3.42

−0.9

3.90

2.69

−31.1

12.90

9.19

−28.8

Gambela—rural

3.01

2.79

−7.3

11.03

7.76

−29.7

Gambela—urban

3.01

2.80

−6.8

11.03

7.22

−34.6

Harari

3.76

3.48

−7.3

4.54

2.87

−36.7

12.71

9.10

−28.4

Oromiya—rural

2.66

2.26

−15.0

3.52

3.94

11.9

10.16

7.52

−26.0

Oromiya—urban

2.66

2.43

−8.7

3.52

3.20

−9.2

10.16

8.00

−21.3

SNNP—rural

2.52

2.36

−6.3

2.93

3.73

27.3

9.39

5.57

−40.7

SNNP—urban

2.52

2.62

4.0

2.93

3.31

12.9

9.39

6.93

−26.2

Somali—rural

3.25

2.90

−10.8

3.82

3.05

−20.1

11.73

8.31

−29.1

Somali—urban

3.25

3.43

5.5

3.82

2.83

−26.0

11.73

8.69

−25.9

Tigray—rural

3.82

2.84

−25.7

4.67

3.44

−26.3

10.71

9.17

−14.4

Tigray—urban

3.82

3.10

−18.7

4.67

2.94

−37.0

10.71

8.86

−17.3

Notes: ‘Orig’ indicates original poverty lines calculated by CSA. ‘UC’ indicates Arndt and Simler (2010) utility-consistent poverty lines estimated with PLEASe. ‘% Diff’ indicates the percentage difference.

Source: CSA and authors’ calculations from HICES

What accounts for these differences? Both approaches use similar methods to construct the nominal household consumption aggregate (Deaton and Zaidi 2002), and indeed the nominal household consumption aggregates are themselves similar. The source of the differences thus follows from the handling of the poverty lines and deflation. As shown in Table 5.2, the CSA and utility-consistent poverty lines differ for each of the spatial domains, and those differences are larger in 2005 and 2011 than in 2000. While the utility-consistent poverty lines on average are 5.6 per cent lower on average in 2000, they are 10.5 per cent lower in 2005 and 26.6 per cent lower in 2011. However, the utility-consistent poverty lines are only uniformly lower across all spatial domains in 2011. In both 2000 and 2006, they are lower than the CSA poverty lines in roughly 60 per cent of the cases. Even in 2011, the differences were not uniformly even. Indeed, they ranged from 13 per cent in urban Amhara to 45 per cent in Addis Ababa.

To understand why the poverty lines differ for the two approaches, we must understand how the CSA poverty lines were derived. The original CSA approach to maintaining consistency was to use the 1995 poverty line as the benchmark. More specifically, the national poverty line was calculated for 1995/6 in Addis Ababa values. In subsequent years this poverty line was scaled up to 2000, 2005, and 2011 prices using the CPI. The inflated 1995/6 poverty line was then applied to the 2000, 2005, and 2011 regionally deflated household consumption aggregates to calculate poverty. The consumption aggregates were regionally deflated using price indices calculated in each stratum relative to the consumption basket for the capital (Addis Ababa) using the maximum number of common items (i.e. items consumed in all of the strata). This differs from the utility-consistent approach in that the latter estimates poverty lines for each region for each year and relies on revealed preference tests and maximum-entropy methods to maintain consistency.

Table 5.3. Region- and time-specific minimum calorie requirements

Difference from

CSA standard (2200)

2000

2005

2011

2000

2005

2011

Addis Ababa

2289

2314

2305

89

114

105

Afar—rural

2172

2177

2226

28

23

26

Afar—urban

2276

2253

2232

76

53

32

Amhara—rural

2157

2164

2186

43

36

14

Amhara—urban

2191

2224

2259

9

24

59

Benishangul—rural

2141

2179

2146

59

21

54

Benishangul—urban

2179

2210

2217

21

10

17

Dire Dawa—rural

2168

2138

2146

32

62

54

Dire Dawa—urban

2212

2285

2249

12

85

49

Gambella—rural

2201

2172

1

28

Gambella—urban

2193

2,205

7

5

Harari

2202

2190

2175

2

10

25

Oromiya—rural

2132

2127

2142

68

73

58

Oromiya—urban

2192

2213

2246

8

13

46

SNNP—rural

2151

2134

2141

49

66

59

SNNP—urban

2219

2196

2263

19

4

63

Somali—rural

2171

2151

2131

29

49

69

Somali—urban

2186

2170

2142

14

30

58

Tigray—rural

2118

2151

2173

82

49

27

Tigray—urban

2144

2176

2192

56

24

8

Source: Authors’ calculations from HICES data

Further, the original 1995/6 national food poverty line, which forms the basis of the national poverty line, was estimated as the cost of consuming 2200 calories per adult per day based on the consumption patterns of poor (p.64) households ranked by the consumption aggregate. This also differs from the utility-consistent approach, which does not fix the calorie requirements to be the same across all regions. Rather it allows the demographic characteristics of the particular region to dictate the differing calorie requirements. In particular, it calculates the average calorie requirements in a spatial domain for people of all ages, not just adults. As illustrated in Table 5.3, the utility-consistent minimum calorie requirements differ across regions and range from 114 calories higher than the CSA-standard 2200, to 82 calories lower. One would thus expect, ceteris paribus, that the utility-consistent poverty lines would be higher than the original when the minimum calorie requirement of the former is greater than 2200, given that the former is based on the estimated cost of acquiring more calories than the latter. Conversely, one would expect the utility-consistent poverty lines to be lower when the utility-consistent minimum calorie requirement is less than 2200. This, however, is only the case for half of the comparisons.

Table 5.4. Household food consumption baskets by spatial domain, Ethiopia HICES 2011

Addis

Afar

Amhara

Benishangul

Dire Dawa

Ababa

Rural

Urban

Rural

Urban

Rural

Urban

Rural

Urban

Harari

Teff—unmilled

0.035

0.004

0.004

Teff—milled

0.198

0.025

0.108

0.080

0.168

0.024

0.034

0.072

0.061

Wheat—unmilled

0.004

0.005

0.003

0.003

0.009

0.004

0.004

Wheat—milled

0.026

0.042

0.047

0.060

0.040

0.008

0.014

0.190

0.061

0.094

Barley—unmilled

0.008

0.005

Barley—milled

0.034

0.005

0.017

Maize—unmilled

0.020

0.003

0.030

0.027

0.015

Maize—milled

0.009

0.284

0.062

0.054

0.034

0.064

0.032

0.060

0.016

0.092

Sorghum—unmilled

0.004

0.008

0.006

Sorghum—milled

0.026

0.089

0.136

0.125

0.162

0.126

0.293

0.093

0.150

Millet—milled

0.024

0.005

0.057

0.058

Rice

0.006

0.025

0.010

Mixed cereals—milled

0.004

Other cereals—unmilled

0.025

0.011

0.011

0.015

Other cereals—milled

0.003

0.003

0.044

0.006

0.003

0.020

0.004

Horse beans—unmilled

0.006

Horse beans—milled

0.004

0.004

0.009

0.053

0.056

0.014

0.020

0.006

Chick peas—unmilled

0.003

0.004

0.006

Chick peas—milled

0.003

0.008

0.005

Peas—unmilled

Peas—milled

0.048

0.006

0.022

0.027

0.029

0.013

0.025

0.012

0.028

0.008

Lentils—unmilled

0.003

0.011

0.002

Lentils—milled

0.021

0.004

0.008

0.009

0.009

Haricot beans—unmilled

0.002

0.012

0.015

0.010

Haricot beans—milled

0.002

0.058

0.033

Vetch—milled

0.013

0.004

0.055

0.043

0.036

0.010

0.018

Fenugreek—unmilled

0.004

Fenugreek—milled

0.031

0.014

0.039

Soya beans—unmilled

0.004

0.042

0.022

0.003

Mixed pulses—milled

0.042

0.013

0.005

0.009

0.026

0.004

0.013

0.012

(p.67) Other pulses—unmilled

0.003

Other pulses—milled

0.002

Linseed—oilseed

0.002

0.006

0.004

Other oilseeds

0.003

0.003

Spaghetti

0.003

0.009

0.007

0.010

0.013

Macaroni

0.005

0.008

0.012

0.006

0.005

0.015

0.024

0.009

Injera

0.077

0.031

0.005

0.019

0.010

0.006

0.075

0.065

Wheat bread

0.092

0.010

0.043

0.005

0.039

0.006

0.011

0.016

0.110

0.073

Biscuit

0.002

0.004

0.007

0.005

0.004

0.003

Other prepared foods

0.019

0.002

0.004

0.003

0.011

0.007

Beef

0.015

0.003

0.009

0.009

0.015

0.021

0.022

Mutton–Goat

0.003

0.003

0.003

0.020

Chicken

0.006

0.005

Other meat

0.003

Fresh fish

Dried fish

0.012

0.008

0.006

Milk

0.004

0.322

0.014

0.005

0.002

0.086

0.011

0.049

Cottage cheese

0.003

0.004

0.004

Yogurt

0.003

0.005

Butter (milk)

0.006

Other dairy

0.005

Butter (oil)

0.002

0.006

0.013

Edible oils

0.096

0.037

0.082

0.029

0.061

0.073

0.085

0.073

0.072

0.066

Ethiopian kale

0.005

0.005

0.004

0.007

0.017

Cabbage/Lettuce/Spinach

0.004

0.003

Tomato

0.007

0.006

0.024

0.005

0.005

0.007

0.012

0.041

0.032

Onion

0.041

0.018

0.052

0.014

0.032

0.048

0.047

0.016

0.041

0.041

Garlic

0.003

0.003

0.002

0.011

0.011

0.003

Green pepper

0.004

0.003

0.006

0.011

0.007

Pumpkin

0.004

0.005

Canned tomato

Other vegetables

0.038

0.005

Banana

0.003

0.006

0.003

Mango

0.003

0.005

Other fruit

0.003

0.003

Potato

0.013

0.004

0.030

0.020

0.008

0.029

0.010

0.018

0.017

(p.68) Sweet potato

0.005

0.026

Kocho (from enset)

Amicho (from enset)

Godere

Other tubers

Salt

0.004

0.015

0.012

0.007

0.006

0.021

0.017

0.009

0.004

0.007

Sugar

0.045

0.057

0.065

0.003

0.014

0.017

0.022

0.023

0.056

0.039

Sugar cane

0.002

0.003

Candy

0.003

Other refined food

0.020

0.019

0.004

Outside meals

0.055

0.032

0.024

0.071

0.060

0.045

0.027

0.046

0.092

0.025

Spices

0.080

0.026

0.084

0.144

0.109

0.088

0.110

0.015

0.028

0.030

Number of food items

34

26

34

42

35

46

45

23

31

27

(p.69) Teff—unmilled

0.002

0.010

Teff—milled

0.006

0.018

0.045

0.108

0.016

0.079

0.045

0.139

Wheat—unmilled

0.009

0.009

0.008

0.008

0.007

0.018

0.007

Wheat—milled

0.009

0.056

0.051

0.057

0.013

0.018

0.133

0.070

0.147

0.127

Barley—unmilled

0.004

0.006

0.005

0.015

0.005

Barley—milled

0.021

0.012

0.006

0.005

0.067

0.027

Maize—unmilled

0.016

0.020

0.045

0.021

0.059

0.035

0.052

0.022

0.005

0.006

Maize—milled

0.295

0.221

0.135

0.078

0.127

0.124

0.062

0.033

0.061

0.019

Sorghum—unmilled

0.005

0.046

0.019

Sorghum—milled

0.041

0.006

0.068

0.056

0.017

0.035

0.029

0.175

0.103

Millet—milled

0.004

0.013

0.011

0.003

0.002

Rice

0.005

0.044

0.073

Mixed cereals—milled

0.003

0.004

0.009

0.011

Other cereals—unmilled

0.013

0.033

0.008

0.028

0.009

0.023

0.008

Other cereals—milled

0.028

0.006

0.004

0.021

0.087

0.004

Horse beans—unmilled

0.006

0.003

0.013

0.009

Horse beans—milled

0.022

0.021

0.026

0.010

0.004

0.055

0.038

Chick peas—unmilled

0.002

0.002

Chick peas—milled

0.008

0.006

Peas—unmilled

0.004

Peas—milled

0.030

0.025

0.012

0.033

0.005

0.018

0.005

0.040

0.020

Lentils—unmilled

Lentils—milled

0.006

0.003

0.008

0.007

0.003

0.002

Haricot beans—unmilled

0.005

0.007

0.008

0.006

0.004

0.019

Haricot beans—milled

0.015

0.008

Vetch—milled

0.009

0.008

0.006

0.012

0.018

0.057

Fenugreek—unmilled

0.002

Fenugreek—milled

0.005

0.005

0.006

0.004

Soya beans—unmilled

0.013

0.002

0.016

0.008

0.003

(p.70) Mixed pulses—milled

0.006

0.004

0.008

0.006

Other pulses—unmilled

Other pulses—milled

Linseed—oilseed

Other oilseeds

Spaghetti

0.011

0.020

Macaroni

0.006

0.003

0.010

0.010

Injera

0.005

0.009

0.039

0.008

0.056

0.021

0.010

Wheat bread

0.019

0.005

0.043

0.006

0.063

0.021

0.004

0.017

Biscuit

0.006

0.002

0.005

0.003

Other prepared foods

0.007

0.003

0.005

0.009

0.003

Beef

0.004

0.029

0.007

0.014

0.025

0.015

0.012

0.034

Mutton–Goat

0.003

0.006

Chicken

0.010

0.002

0.007

Other meat

0.007

0.005

0.008

Fresh fish

0.079

0.059

0.003

Dried fish

0.004

0.016

0.012

0.006

Milk

0.079

0.131

0.060

0.021

0.019

0.005

0.114

0.060

0.002

Cottage cheese

0.008

0.004

0.009

Yogurt

0.005

0.003

Butter (milk)

0.007

0.011

0.003

0.010

Other dairy

0.004

0.009

Butter (oil)

0.019

0.006

0.019

0.009

Edible oils

0.043

0.058

0.056

0.085

0.033

0.059

0.111

0.090

0.047

0.067

Ethiopian kale

0.007

0.019

0.022

0.014

0.064

0.058

0.004

Cabbage/Lettuce/Spinach

0.003

0.002

0.010

Tomato

0.012

0.004

0.010

0.003

0.010

0.020

0.010

0.020

Onion

0.024

0.022

0.025

0.036

0.015

0.025

0.010

0.018

0.017

0.030

Garlic

0.009

0.006

0.006

0.016

0.013

0.003

Green pepper

0.007

0.008

0.009

0.009

0.002

Pumpkin

0.005

0.005

0.002

0.002

0.003

(p.71) Canned tomato

0.003

0.004

Other vegetables

0.100

0.106

0.014

0.010

0.016

0.013

Banana

0.003

0.002

0.005

Mango

0.007

Other fruit

0.002

0.004

0.007

0.002

Potato

0.008

0.009

0.019

0.006

0.016

0.008

0.003

0.011

Sweet potato

0.009

0.011

0.002

0.034

0.035

Kocho (from enset)

0.009

0.045

0.003

0.162

0.052

Amicho (from enset)

0.030

0.007

Godere

0.006

0.036

0.022

Other tubers

0.004

0.002

0.015

0.007

Salt

0.015

0.017

0.018

0.009

0.014

0.010

0.011

0.007

0.008

0.009

Sugar

0.025

0.021

0.012

0.031

0.014

0.220

0.216

0.012

0.026

Sugar cane

0.002

0.003

Candy

Other refined food

0.007

0.009

Outside meals

0.018

0.030

0.040

0.037

0.021

0.059

0.007

0.062

0.066

0.058

Spices

0.033

0.007

0.061

0.070

0.044

0.052

0.006

0.008

0.083

0.081

Number of food items

36

29

49

48

43

46

22

30

35

33

Source: Authors’ calculations from HICES data

The source of the differences in the utility-consistent and CSA poverty lines thus must also follow from the composition of the basket used to value the region-specific calorie requirements. Unfortunately, the original code used to construct the 1995/6 poverty line and regional deflators is not available. (p.65) Thus we cannot compare the consumption baskets used to create the utility-consistent poverty lines with the original from 1995/6. But the food consumption baskets derived from the utility-consistent approach shown in Table 5.4 give an indication of how the baskets differ substantially over the spatial domains in 2011, including urban and rural areas within regions. Given that the CSA poverty lines are defined over the regions (urban and rural combined), not over these more disaggregated spatial domains, differences in food consumption baskets are likely to be an important contributor to the different poverty line estimates.

5.5 Concluding Remarks

This chapter describes the application to Ethopia of the standardized PLEASe computer code stream based on Arndt and Simler’s (2010) utility-consistent approach to measuring consumption poverty. In doing so, we highlight the importance of adapting the code stream to address changes in data collection periods and strata for the respective surveys over time. Indeed, changes in the (p.66) (p.72) duration and time of year for data collection can be especially problematic for consistency in the presence of annual inflation of over 30 per cent. In addition, the Ethiopia case provides an example of how to address convergence problems encountered when running the PLEASe code. Careful consolidation of spatial domains and limiting the number of iterations in the estimation of poverty lines are potential solutions.

According to our estimates using utility-consistent poverty lines from the application of the PLEASe code stream, national poverty fell from 46.8 per cent in 2000, to 46.0 per cent in 2005, and finally to 23.8 per cent in 2011. Poverty is considerably higher in rural areas (48.0 per cent) where more than 80 per cent of the population lives, compared to urban areas (39.0 per cent). Although the rural headcount ratio fell by 11.2 percentage points, urban areas as a whole saw even greater declines in poverty, as the urban poverty rate fell to 13.3 per cent by 2011.

Although the patterns of decline in poverty as estimated using utility-consistent poverty lines are similar to those from the original CSA estimates, the utility-consistent poverty estimates fell by even more than the CSA estimates did. These differences stem from the handling of the poverty lines and deflation. Unlike the CSA approach that maintains consistency over time by using the 1995 poverty line as a benchmark and scales it up to 2000, 2005, and 2011 prices using the CPI, the utility-consistent approach estimates poverty lines for each region for each year and relies on revealed preference tests and maximum-entropy methods to maintain consistency. Although differing region-specific calorie requirements contribute partly to the disparity among the poverty lines of the two approaches, the differing compositions of the baskets used to value these calorie requirements likely played a more important role. The specificity of these utility-consistent weights, based on consumption patterns of the poor in the spatial domains, is a strength of this approach compared to the previous approach taken by the CSA.

References

Bibliography references:

Arndt, Channing, Ulrik Richardt Beck, M. Azhar Hussain, Kenneth Simler, and Finn Tarp (2013). ‘User Guide to Poverty Line Construction Toolkit: Version 2.0’, Development Economics Research Group, University of Copenhagen, Denmark.

Arndt, Channing and Kenneth Simler (2010). ‘Estimating Utility‐Consistent Poverty Lines with Applications to Egypt and Mozambique’, Economic Development and Cultural Change, 58(3): 449–74.

Blackorby, Charles and David Donaldson (1987). ‘Welfare Ratios and Distributionally Sensitive Cost-Benefit Analysis’, Journal of Public Economics, 34: 265–90.

(p.73) Deaton, Angus and Salman Zaidi (2002). ‘Guidelines for Constructing Consumption Aggregates for Welfare Analysis’, Living Standards Measurement Study Working Paper 135. Washington, DC: World Bank.

Dorosh, Paul and Emily Schmidt (2010). ‘The Urban–Rural Transformation in Ethiopia’, ESSP II Working Paper 13, International Food Policy Research Institute/Ethiopia Strategy Support Program II, Addis Ababa, Ethiopia.

Foster, James, Joel Greer, and Erik Thorbecke (1984). ‘A Class of Decomposable Poverty Measures’, Econometrica, 52: 761–6.

Headey, Derek, Fantu Nisrane, Ibrahim Worku, Mekdim Dereje, and Alemayehu Seyoum Taffesse (2012). ‘Urban Wage and Food Price Inflation: The Case of Ethiopia’, ESSP II Working Paper 41, International Food Policy Research Institute/Ethiopia Strategy Support Program II, Addis Ababa, Ethiopia.

MoFED (2008). ‘Dynamics of Growth and Poverty in Ethiopia (1995/96–2004/05)’, Development Planning and Research Department, Ministry of Finance and Economic Development, Addis Ababa, Ethiopia.

MoFED (2012). ‘Ethiopia’s Progress towards Eradicating Poverty: An Interim Report on Poverty Analysis Study (20010/11)’, Development Planning and Research Department, Ministry of Finance and Economic Development, Addis Ababa, Ethiopia.

Pradhan, Menno (2000). ‘How Many Questions Should Be in a Consumption Questionnaire? Evidence from a Repeated Experiment in Indonesia’, Working Paper 112, Cornell Food and Nutrition Policy Program, Cornell University, Ithaca, NY.

Ravallion, Martin (1998). ‘Poverty Lines in Theory and Practice’, Living Standards Measurement Study Working Paper No. 133. Washington, DC: World Bank.

Ravallion, Martin and Benu Bidani (1994). ‘How Robust Is a Poverty Profile?’, World Bank Economic Review, 8: 75–102.

Scott, Christopher and Ben Amenuvegbe (1990). ‘Effect of Recall Duration on Reporting of Household Expenditures: An Experimental Study in Ghana’, Social Dimensions of Adjustment in Sub-Saharan Africa Working Paper No. 6. Washington, DC: World Bank.

Notes:

(1) See Chapter 2 of this book for more details about the general procedure. The household consumption aggregates and poverty lines were calculated using the PLEASe software.

(2) We note that revealed preference conditions should also hold over time (i.e. do the consumption patterns in the same spatial domain but in different time periods cost no less than the own-domain consumption patterns at a specific time when both are evaluated at own-domain prices for that specific time). When these conditions are violated over time, similar maximum-entropy methods can be used to reconcile the differences (Arndt and Simler 2010).

(3) Headey et al. (2012) document a rapid rise in urban food prices for the poor during the 2011 survey period that outpaced the growth of urban nominal wages.