Measuring Pro-Poor Growth in Non-Income Dimensions

Transcription

1 World Development Vol. 36, No. 6, pp. 7, 8 Ó 8 Elsevier Ltd. All rights reserved 35-75X/$ - see front matter doi:.6/j.worlddev.7..9 Measuring Pro-Poor Growth in Non-Income Dimensions MELANIE GROSSE, KENNETH HARTTGEN and STEPHAN KLASEN * University of Göttingen, Germany Summary. Current concepts and measures of pro-poor growth are entirely focused on the income dimension of well-being. This neglects non-income dimensions of poverty as well as the multidimensionality of poverty and well-being. In this paper, we extend the pro-poor growth toolbox to individual and composite measures of non-income achievements. In particular, we apply growth incidence curves, the Ravallion Chen pro-poor growth rate and the poverty-equivalent growth rate to a range of non-income indicators such as education, mortality, vaccinations, stunting, and a multidimensional well-being measure. We are thereby able to study improvements in these dimensions of well-being at various points of the distribution of those indicators as well as at various points of the income distribution. This way we can determine whether improvements in non-income indicators were pro-poor in an absolute or relative sense, and whether they benefited the income poor more than others. We illustrate this empirically for Bolivia during Ó 8 Elsevier Ltd. All rights reserved. Key words pro-poor growth, multidimensionality of poverty, growth incidence curve, povertyequivalent growth rate, Bolivia. INTRODUCTION Pro-poor growth has recently become a central issue for researchers and policy makers, especially in the context of reaching the millennium development goals (MDGs). The various proposals to measure pro-poor growth have also allowed a much more detailed assessment of progress on reducing poverty as they explicitly examine growth along the entire income distribution. However, a central shortcoming of current pro-poor growth concepts and measures is that to date they have been completely focused on the income dimension of poverty, thus ignoring critical non-income aspects of well-being which are also central to the MDGs. As is well known, however, a reduction in income poverty does not guarantee a reduction in non-income dimensions of poverty, such as education or health. In this context, Kakwani and Pernia () note that it would be futile if one operationalizes poverty reduction via pro-poor growth using just one single indicator because poverty is a multidimensional phenomenon, and thus pro-poor growth should also be understood as a multidimensional phenomenon. While non-income indicators of well-being and the multidimensionality of poverty have recently received more and more attention in poverty measurement no attempts have been made to date to measure pro-poor growth using non-income indicators. The aim of this paper is to extend the toolbox of pro-poor growth measurement to individual non-income dimensions and composite measures of well-being, including central capabilities in the areas of health, education, and nutrition which, following Sen (988), are central outcomes of well-being and also included in the MDGs. We do this by extending the growth incidence curve (GIC) by Ravallion and Chen (3) * The authors would like to thank Amartya Sen, Ravi Kanbur, Sonia Bhalotra, and Walter Zucchini and the participants at workshops in Göttingen and sessions at the Spring Meeting for Young Economists in Geneva, the Society for the Study of Economic Inequality in Palma, the Research Committee on Development Economics in Kiel, and the WIDER Project Meeting of A- ssessing and Forecasting Progress in MDGs in Helsinki for helpful comments and discussion. Final revision accepted: October 5, 7.

2 WORLD DEVELOPMENT and related pro-poor growth measures to nonincome indicators and call our approach non-income growth incidence curves (NIGICs). We illustrate this approach using micro-survey data for Bolivia during We distinguish between relating growth in non-income achievements to the distribution of achievement in that non-income indicator and relating growth in non-income achievements to the distribution of income. While the former will allow us to study the distributional pattern of improvements in the non-income dimension, the latter provides insights about how far the income poor have benefited from improvements in non-income dimensions of well-being, thus complementing standard incidence analyses. In addition to investigating growth rates, we also investigate absolute changes of the non-income indicators. For our Bolivian illustration, we find that growth was relatively pro-poor both in the income and in the non-income dimension, but results are much less clear when absolute changes are considered and when changes in non-income achievement are related to the income distribution. The paper is organized as follows. In Section, we describe pro-poor growth measurement and in Section 3 our extension to non-income dimensions. The empirical illustration for Bolivia is contained in Section, and we conclude in Section 5.. METHODOLOGY (a) Defining pro-poor growth Categorizing the different definitions of propoor growth, we use three definitions of propoor growth in this paper: weak absolute pro-poor growth, relative pro-poor growth, and strong absolute pro-poor growth (see also Klasen, 8a). 3 Pro-poor growth in the weak absolute sense means that the growth rates are above for the poor. Pro-poor growth in the relative sense means that the growth rates of the poor are higher than the average growth rates; thus relative inequality falls (i.e., some indicator considering the relative gap between the rich and the poor falls). Pro-poor growth in the strong absolute sense requires that absolute increases of the poor are larger than the average, thus implying that absolute inequality falls (i.e., some measures considering the absolute gap between the rich and the poor falls; see, e.g., Klasen, 8a). Note that these definitions are sensitive to the definition of the poor, that is they depend on the poverty line. In the empirical illustration, we will use the poverty lines used in Bolivia. Also note that here we gloss over an aggregation issue as it may be the case that, for example, for only some of the poor growth is above, while for others it is not. Thus, there may be cases where we are unable to classify the episodes of growth using these definitions. The pro-poor growth measures described below, however, implicitly deal with these aggregation issues and come to a definitive judgment about the pro-poorness of a particular growth episode. They do so, however, in different ways, which has been the subject of some discussion. 5 The absolute definition is obviously the strictest definition of pro-poor growth and the hardest to be met as shown empirically by White and Anderson (). This is why most empirical and policy research has focused on the weak absolute and relative definition (OECD, 6). But this ignores that decreases in relative inequality might be and often are accompanied by increases in absolute inequality which is seen as undesirable by many and can be an important source of social tension (e.g., Atkinson & Brandolini, ; Duclos & Wodon, ; Klasen, ). Conversely, growth that is associated with falling absolute inequality would be particularly pro-poor and, therefore, it is useful to consider this strong absolute concept as well. As we argue in the following section, this is particularly important when examining pro-poor growth in the non-income dimension of poverty where even pro-poor growth in the relative definition might not be seen as sufficiently pro-poor. (b) The GIC A nice way to illustrate the definitions is to use growth incidence curves (GICs), which are also a useful visual tool for the assessment of the distributional pattern of growth. The GIC (Ravallion & Chen, 3) shows the mean growth rate g t in achievements (here incomes y) at each centile p of the distribution between two points in time, t and t. The GIC links the growth rates of different percentiles and is given by GIC ¼ g t ðpþ ¼ y tðpþ y t ðpþ : ðþ By comparing the two periods, the GIC plots the population centiles (from to ranked by income) against the annual per capita

3 MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS 3 growth rate in income of the respective centiles. If the GIC is above for all poor centiles (g t (p) > for all p), then it indicates weak absolute pro-poor growth. If the GIC is negatively sloped throughout it indicates relative pro-poor growth. 6 Starting from the GIC, Ravallion and Chen (3) define the pro-poor growth rate (PPGR) as the area under the GIC up to the headcount ratio H, which has a direct relation to poverty reduction if one uses the Watts index as the relevant poverty measure. The PPGR is formally expressed by PPGR ¼ g p t ¼ Z H t g H t ðpþdp; t ðþ which is equivalent to the mean of the growth rates of the poor up to the headcount. In order to determine whether growth was pro-poor according to the relative definition, one compares the PPGR with the growth rate in mean (GRIM). The GRIM is defined by GRIM ¼ c t ¼ l t ; ð3þ l t where l is mean income. If the PPGR exceeds the GRIM, growth is declared to be pro-poor in the relative sense. Examining pro-poor growth in the strong absolute sense one has to concentrate on the absolute changes in the income of the population centiles between the two periods. We define the absolute GIC by GIC absolute ¼ c t ðpþ ¼y t ðpþ y t ðpþ; ðþ which shows the absolute changes for each centile. By comparing the two periods, the absolute GIC plots the population centiles against the per capita change in income of the respective centile. If the absolute GIC is negatively sloped it indicates strong absolute pro-poor growth. Starting from the absolute GIC we define pro-poor change (PPCH) as the area under the absolute GIC up to the headcount H. The PPCH is formally expressed by PPCH ¼ c p t ¼ X H t c t ðpþ; ð5þ H t which is equivalent to the mean of the changes of the poor up to the headcount. We compare the PPCH with the change in mean (CHIM), which is defined by CHIM ¼ d t ¼ l t l t : ð6þ If the PPCH exceeds the CHIM, growth is declared to be pro-poor in the strong absolute sense. An alternative measure of pro-poor growth is the poverty-equivalent growth rate (PEGR) proposed by Kakwani and Son (6). The PEGR is the growth rate that would have delivered the same poverty reduction if income distribution had remained unchanged. It can be calculated for all the FGT poverty measures (headcount, depth, and severity). Applied to the poverty headcount it is calculated as PEGR ¼ PEH GEH GRIM; ð7þ where PEH is the total poverty elasticity of poverty headcount H with respect to a particular growth episode and GEH is the pure growth elasticity of poverty headcount H with respect to that growth episode (i.e., holding inequality constant), averaged over the two time paths (i.e., a forward headcount in the second year if all incomes had grown by the average income growth rate from the first to the second year and a backward headcount in the first year if all incomes had shrunk by the same rate from the second to the first year) 7 multiplied with the annualized growth rate (GRIM). If the PEGR is higher than the GRIM, growth was pro-poor in the relative definition PRO-POOR GROWTH IN NON-IN- COME DIMENSIONS (a) Concept We begin by extending the GIC to non-income dimensions. The calculation of the NI- GICs broadly follows the concept of the GIC. Instead of income (y) we apply formulas () (7) to selected non-income indicators to measure pro-poor growth directly via outcomebased welfare indicators. Thus, the NIGIC measures pro-poor growth in a non-income sense, for example, the improvement of the health status or the educational level between two periods for each centile of the distribution. We calculate the NIGIC in two different ways. We call the first way the unconditional NIGIC, in which we rank the individuals by each respective non-income variable and generate the population centiles based on this ranking. For example, using average years of

4 WORLD DEVELOPMENT schooling of adult household members, the poorest centile is now the education-poorest one with the lowest average household educational attainment. We call the second way conditional NIGIC, in which we rank the individuals by income and calculate the growth of non-income achievements for these income percentiles. Since the income-poorest group might not be the education-poorest group at the same time, these two ways of studying pro-poor growth will likely yield different results (see below). Conceptually, both ways of calculating the NIGIC are of relevance for policy making. The unconditional NIGIC shows the non-income distributional pattern of improvements of non-income dimensions of well-being. This is central for monitoring progress in non-income dimensions of well-being. It allows answers to questions such as: Have the education poor benefited disproportionately from improvements in education? Have those who were most lacking in nutrition been able to improve their nutritional status the most? In contrast, the conditional NIGIC allows an assessment of the income distributional pattern of improvements in non-income dimensions of well-being. This is a very useful supplement to standard benefit incidence analysis, commonly undertaken as part of public expenditure reviews and poverty assessments in developing countries. Standard benefit incidence studies typically analyze the impact of public spending by calculating shares of the total spending going to each centile and comparing the shares of the income poorest with the income-richest centile (see, e.g., Lanjouw & Ravallion, 998; Roberts, 3; Van de Walle, 998; Van de Walle & Nead, 995). With the conditional NI- GIC it is now possible to analyze how progress in a particular aspect of human welfare (e.g. health or education) is distributed across the income distribution. For example, it provides an instrument to assess if public social spending programs have reached the targeted incomepoorest population groups and if the public resources are effectively allocated and used. In this respect, the conditional NIGIC can be referred to as an outcome-oriented incidence analysis. As with the income measure, we generate GICs not only for relative improvements, but also for absolute improvements. In fact, we believe that tracking absolute improvements is particularly relevant in non-income dimensions for two reasons: First, it appears more usual and intuitive to consider absolute gaps in achievements (e.g. the literature on longevity and education gaps between the rich and poor usually reports absolute not relative gaps). Consequently, it seems harder to argue that a % growth in education for the education poor compared to 5% education growth for the education rich was pro-poor if the poor started with year of education (i.e. now have. years) and the rich with (i.e. now have.5). This is also related to the more ordinal character of some of the non-income indicators which make relative changes hard to interpret (see below). Second, the importance of absolute progress in non-income achievements would also be consistent with Sen s capability approach. As he emphasizes in Sen (98), poverty in the capability space should be considered using absolute achievements while often the means to achieve it (such as incomes) can be considered in a relative dimension. After constructing the GICs which track absolute and relative progress, we can then calculate the rates of pro-poor growth (PPGR) using the relative NIGICs, the PEGR, and PPCHs using the absolute NIGICs and compare each to the GRIM or CHIM. (b) Specification of the non-income indicators We calculate the unconditional and conditional NIGIC for education, health, nutrition, and for a composite welfare index (CWI) as described below. We are working with DHS data for Bolivia from the years 989 and 998 that do not contain information on income or consumption due to its focus on demographics, health, and fertility. However, in our DHS data set, we use simulated incomes based on a dynamic cross-survey micro-simulation methodology proposed by Grosse, Klasen, and Spatz (5). 9 The basic idea of this simulation methodology is the following. The authors use two kinds of surveys: first, the DHS (of 989 and 998) and, second, the Bolivian household surveys (the nd EIH of 989 and the ECH of 999). Then they estimate an income correlation in the household survey, apply the coefficients to the DHS, and predict, that is, simulate, incomes in the DHS. For each non-income indicator, we identify alternative variables to capture particular aspects of the non-income dimension in question. For education, we specify eight different vari-

5 MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS 5 ables. We calculate average years of schooling for all adult household members and for males and females separately. Age plays an important role when analyzing changes in nonincome indicators, especially for education. In particular, not much improvement in education can be expected among the adult population beyond usual school-going age (e.g. the education of 3 -year olds in 989 should not be very different from the education of 5-year olds in 999). To avoid misleading conclusion from potential low improvements, we therefore also restrict the sample to women aged between and 3 as only this age group is likely to have experienced a change in their educational achievement (the 3-year olds in 999 represent a new cohort of women who were educated later than the other cohorts). In addition, we calculate the maximal education per household (instead of the average) for all adults, males, females, and females aged between and 3. The idea here is that it might be sufficient that one household member is well educated to generate income for the whole household and to invest in education of other household members (i.e., intra-household externalities) (Basu & Foster, 998). For health, we specified three different variables. We calculate infant survival rates of children aged under 5 years and also for children aged under year. Furthermore, we take the average vaccinations of children aged between and 5 per household, with a maximum of eight possible vaccinations for each child. 3 The vaccination rate is a variable that represents access to health care and preventive medicines. A similar variable has, for example, been used in the monitoring of the health sector reform project in Bolivia in 999 (Montes, 3). For nutrition, we use stunting z-scores as the variable that measures chronic undernutrition for children aged between and 5 years. The stunting z-scores are defined as the difference of height at a certain age and the median of the reference population for height at that age divided by the standard deviation of the reference population. It takes values between approximately 6 and 6, where values below are considered as being moderately undernourished and below 3 as being severely undernourished (see, e.g., Klasen, 8b). Problematic might be that the z-score contains a lot of genetic noise in the sense that, for example, a low z-score interpreted as being undernourished might simply appear because the parents are genetically short but the child is small but well nourished and vice versa. Apart from studying individual non-income achievements, we also want to study progress in multidimensional non-income achievements using a CWI. Here, we follow the methodology of the human development index (HDI) to address the problem of difference scales of the variables (UN, 3). Each variable that enters the index is normalized to be between and in subtracting the individual value from the minimum value observed in the data set divided by the range X n CWI ¼ individual n min : ð8þ n max min i¼ The CWI is constructed by simply averaging the selected variable scores. It includes four of the above explained variables: average education of all adult household members, stunting z-scores, under 5 survival, and average vaccinations. As not all variables are available for all households (e.g., health and nutrition variables are only available for households who have children in the required age group), we calculate the CWI for two different samples. The first sample, called small sample, is the one for which all variables are available for all households. This reduces the sample size enormously (in 989, e.g., from 6,53 to,36 households) and, more importantly, in a non-random fashion. 5 The second sample, called big sample, includes all households, but the index is averaged over fewer variables for those households which do not have data for nutrition and/or health variables. The advantage of creating the CWI based on the big sample is the higher number of observations but the disadvantage is that the results for some centiles are driven by very few or only one variable. The smaller sample has fewer observations but contains for all households the same number of variables. For both the small and the big sample, we in addition augment the indices by also including simulated income as a fourth indicator. (c) Limitations of the indicators While we show below that these indicators yield important insights, we want to highlight a number of conceptual and empirical limitations of, and potential objections to, extending pro-poor growth measurement to non-income dimensions of well-being.

6 6 WORLD DEVELOPMENT The first limitation relates to the interpretation of the growth rates using the NIGIC, where we interpret an ordinal relation in a cardinal fashion. Examining an ordinally scaled variable such as years of schooling, one can say that 6 years of schooling is better than 3 years but one cannot be sure that the household is twice as well-educated. 6 This ordinal scaling leads to two different kinds of interpretation problems. First, averaging an ordinally scaled variable across household members leads to the problem that such averages can be misleading. 7 In order to partially address this issue, we additionally consider the individual educational attainment (instead of only the average of the household) and the gender gap in education within households. Second, concerning increases in the years of schooling, just comparing growth rates might be misleading. For example, Table shows for average education an increase of 8% for the nd decile compared to 6% of the 9th decile which might be overstating the improvement for the poor because the years of schooling of the poor increase from.3 to.37 years of schooling and those of the non-poor from.73 to.3. We try to address this problem in calculating absolute NIGIC and PPCHs. However, even when we use absolute changes which equal approximately, a further question remains open. An increase of.6 years of schooling of the nd decile might be less beneficial, because perhaps the persons are still more or less illiterate, compared to the increase of.7 years of schooling in the 9th decile, which might mean completing secondary schooling and getting a degree. Third, many of the non-income indicators are bounded above, that is, there are firm or likely upper limits on such achievements. One hundred percent survival in the first year is the upper limit for health, more than 8 or 9 years of education is very rare, more than eight vaccinations is not recommended, done, or measured, etc. This generates the problem (and indeed is the case in Bolivia) that some households have reached the upper limit and further growth is not possible or very hard to achieve. It generates the further problem that inequality in such indicators is typically low when a significant share of households has reached the upper limit. This may generate the misleading impression that there is no longer any inequality to be concerned about. One way to address this issue would be to transform our achievement variables in a way that reflects the greater difficulty to achieve further improvements as one approaches the upper bound, such as a percentage reduction of the absolute gap to the upper bound or an improvement index proposed by Kakwani (993). But as we are solely interested in the distributional pattern of welfare and welfare improvements, not in the difficulty of achieving them, we are refraining from making such adjustments here. If, due to the fact that the education- or income-rich already have close to the maximum possible education and thus any further growth in education is bound to reduce inequality in years of schooling, we believe that it is proper from a well-being assessment to report this decline in inequality in years of schooling. This does not, of course, mean that there are no further educational inequalities to consider. Our indicator purely focuses on the quantity of education and it may well be that educational quality continues to differ greatly, an issue that we cannot address here due to the lack of data on the quality of education. 8 The fourth type of problem in comparing relative changes relates to the stunting z-score. In our data sets, it ranges roughly from 6 to6. Relative changes in the stunting z-score cannot be calculated because of the coexistence of negative, positive and values in the variable range. For example, how to compare the relative improvement from to with an improvement from to from the year 989 to 998? We reduce this problem by transforming the z-score in such a way that all values are positive, that means by adding the minimum value of both data sets (in our case 5.89) to each z-score to get a range of only positive numbers. Another limitation is the problem of weighting which we illustrate with the example of child mortality. 9 For example, comparing two households A and B where A has one child and B has children the households should be weighted differently when in both households one child dies. Household A has a child mortality rate of % whereas B of only %, but one may argue that both deaths are equally lamentable and should be considered equally important. But any weighting would, however, be quite arbitrary and induce difficulties in justifying it with economic or welfare-theoretical judgments. Keeping this critical issue in mind we use unweighted child survival rates. Another limitation calculating the NIGIC in some contexts is that some non-income indicators do not vary much between households. This holds especially for under 5 and under

7 Table. Non-income indicators and related variables (unconditional, Bolivia, 989 and 998) Mean : Gini Mean of deciles (unconditional), 989 Income Education Average education Average education of respondents n.d..5 Average education of respondents ( 3) Average education of partners n.d.. Maximum education Maximum education of respondents n.d..5 Maximum education of respondents ( 3) Maximum education of partners n.d.. Individual education n.d..7 Health Under 5 child survival rate (%) Under child survival rate (%) Average vaccination per child (age P ) Nutrition Stunting z-score n.d..9 (continued on next page) MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS 7

8 Table (continued) Mean : Gini Mean of deciles (unconditional), 998 Income Education Average education Average education of respondents n.d.. Average education of respondents ( 3) Average education of partners Maximum education Maximum education of respondents n.d.. Maximum education of respondents ( 3) Maximum education of partners Individual education Health Under 5 child survival rate (%) Under child survival rate (%) Average vaccination per child (age P ) Nutrition Stunting z-score n.d..6 Notes: The explanation for the variables is the following. Income: Real household income per capita in Bolivianos per month (CPI of 995 = ). Education: All variables for education are measured in average single years per adult married household members. Respondents and partners are only couples, for which the respondent knows the education of her partner. Health: Under 5 () survival are estimated with life table estimations and we take the sample of children born (5) years prior to the sample. Survival rates are averaged over the household. Vaccinations: Average vaccinations of the children in the household older than, where the possible vaccinations are three against polio, three against DPT, one against measles, and one against BCG. Nutrition: Stunting z-score of the last born child of each respondent (averaged over the household) with a child defined to be stunted if her z-score is below. Source: Own calculations. 8 WORLD DEVELOPMENT

9 Table. Non-income indicators and related variables (conditional, Bolivia, 989 and 998) Mean : Mean of the deciles (conditional), 989 Income Education Average education Average education of respondents Average education of respondents ( 3) Average education of partners Maximum education Maximum education of respondents Maximum education of respondents ( 3) Maximum education of partners Individual education Health Under 5 child survival rate (%) Under child survival rate (%) Average vaccination per child (age P ) Nutrition Stunting z-score n.d. (continued on next page) MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS 9

10 Table (continued) Mean : Mean of the deciles (conditional), 998 Income Education Average education Average education of respondents Average education of respondents ( 3) Average education of partners Maximum education Maximum education of respondents Maximum education of respondents ( 3) Maximum education of partners Individual education Health Under 5 child survival rate (%) Under child survival rate (%) Average vaccination per child (age P ) Nutrition Stunting z-score n.d. Notes: The explanation for the variables is the following. Income: Real household income per capita in Bolivianos per month (CPI of 995 = ). Education: All variables for education are measured in average single years per adult married household members. Respondents and partners are only couples, for which the respondent knows the education of her partner. Health: Under 5 () survival are estimated with life table estimations and we take the sample of children born (5) years prior to the sample. Survival rates are averaged over the household. Vaccinations: Average vaccinations of the children in the household older than, where the possible vaccinations are three against polio, three against DPT, one against measles, and one against BCG. Nutrition: Stunting z-score of the last born child of each respondent (averaged over the household) with a child defined to be stunted if her z-score is below. Source: Own calculations. 3 WORLD DEVELOPMENT

11 MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS 3 survival which is very low in Bolivia at the household level. For both years, Table shows that from the 3rd or th decile upward, the maximum value % survival is already reached in both years, so that no improvement is possible any more. This translates into growth rates of, so that the unconditional NI- GIC becomes flat and takes the value of from the 3rd or th decile onward. The problem of flat curves always arises when the variable values are bounded above (as, e.g., a maximum of 9 years of schooling or eight vaccinations). In these cases, it is more interesting to examine the conditional NIGIC, in which we link the survival rates and vaccination to income. Here, low or variation is less problematic than for the unconditional NIGIC because the variables are ranked by income. As Table and all figures show there is no flat part any more. Now we generate interesting information regarding the changes on the non-income indicators when ranked according to their income situation and how improvements are distributed.. EMPIRICAL ILLUSTRATION (a) Inequality in non-income dimensions of wellbeing Bolivia is one of the countries with a very unequal income distribution in Latin America. We find high and persisting income inequality as measured with the Gini coefficient that falls from.56 in 989 to.5 in 998 (Table ). This high inequality is also reflected in the high and only slightly falling : ratio, that is mean incomes in the th, divided by mean incomes in the st decile. Turning from inequality to growth we find that all deciles increased their incomes. Especially in the 99s, Bolivia experienced relatively high growth rates (which also were pro-poor in urban and rural areas). However, Bolivia was and is one of the poorest countries of the region, and the positive economic trend has reversed since 999 combined with some episodes of social and political turmoil. For social indicators such as life expectancy or literacy, Bolivia used to show much worse outcomes compared to other countries in the region. However, there have been notable and sustained improvements in many social indicators since the late 98s which continued to improve during the recent economic slowdown (Klasen et al., 7). Tables and show achievements in income and non-income dimensions of well-being in Bolivia for 989 and 998 sorting households by the respective non-income achievement in Table, and by incomes in Table. We also calculate Gini coefficients and : decile ratios. Table 3 does the same for the composite indicator and Table reports the respective pro-poor growth measures. Table 3. Deciles of the composite welfare index (Bolivia, 989 and 998) Mean : Gini Mean of the deciles (unconditional), 989 Composite welfare index Small sample Big sample Mean of the deciles (unconditional), 998 Composite welfare index Small sample Big sample Mean of the deciles (conditional), 989 Composite welfare index Small sample Big sample Mean of the deciles (conditional), 998 Composite welfare index Small sample Big sample Notes: The composite welfare index includes average education of married household members, under 5 survival rate, average vaccination per child, and stunting. Source: Own calculations.

12 Table. Mean growth rates, mean absolute changes, pro-poor growth rates, poverty equivalent growth rates, and absolute pro-poor changes (Bolivia, 989 and 998) NIGIC Poverty equivalent NIGIC Unconditional Conditional Growth rate Unconditional Condtional GRIM PPGR moderate PPGR extreme PPGR moderate PPGR extreme PEGR moderate PEGR extreme CHIM PPCH moderate PPCH extreme PPCH moderate PPCH extreme Income Education Average education Average education of all respondents Average education of respondents ( 3) Average education of partners Maximal education Maximum education of all respondents Maximum education of respondents ( 3) Maximum education of partners Individual education Health Under 5 child survival rate (%) Under child survival rate (%) Average vaccination per child (age P ) Nutrition Stunting z-score Composite welfare index (CWI) Small sample Big sample Notes: For the explanation of the variables, see Table. We are using two poverty lines. The moderate poverty line leads to an income headcount of 77 and the extreme poverty line to an income headcount of 56, which we also use for the non-income indicators. GRIM: Growth rate in mean. PPGR: Pro-poor growth rate. PEGR: Poverty equivalent growth rate. CHIM: Change in mean. PPCH: Pro-poor change. Changes are for the entire period and not annualized. Source: Own calculations. 3 WORLD DEVELOPMENT

13 MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS 33 The Ginis for education variables are all in the range of. and fall between 989 and 998. As shown in Table, this is related to the fact that the education rich already are quite close to the upper bound so that inequality in the years of schooling is bound to be smaller and falling over time, as the poor are catching up. As shown by Thomas, Wang, and Fan (), educational Ginis are much larger in African countries (in fact larger than typical income Ginis) where the majority of people have close to no education, and only the better off are benefitting from educational expansion (see also below). Interesting to note is that the highest Ginis exist for the group of all (female) respondents, both for average and maximal education indicating larger educational gaps among females than males. These findings are also reflected in the : ratio. The conditional deciles in Table also show that the level of schooling increases with increasing income for all educational variables, but the : ratio is much lower than in the unconditional case. This suggests an imperfect correlation between education and income poverty. We also find that there has been an improvement for all educational variables in all deciles for both the unconditional and the conditional cases during , a rather encouraging finding (Tables and ). The extremely low Ginis for the under and under 5 survival rates can be explained by the low overall incidence of child mortality in Bolivia at the household level. For both age groups, child mortality is about %. The conditional deciles in Table indicate that the risk of child mortality is higher for the income-poor compared to the income-rich. For vaccination, we find only little improvements over time for the lower deciles and also for the higher deciles, which is also due to the fact that the best-vaccinated deciles had only limited room for improvements. The inequality of the stunting z-score is relatively low and falls slightly. Malnutrition decreases with an increasing position in the income distribution, but the differences for the income deciles are lower but clearly existing. The CWI reflects the findings from above where the Gini coefficients decrease for the selected variables (Table 3). For the CWI, the Gini coefficient is higher for the big sample than for the small sample. (b) GICs and pro-poor growth Figure shows the relative and absolute GIC for income. The relative GIC plots the annual Percentile GIC (lhs) 95% CI upper bound (lhs) 95% CI lower bound (lhs) Absolute GIC 95% CI upper bound abs. 95% CI lower bound abs. Figure. GIC (relative and absolute). Source: Own calculations.

14 3 WORLD DEVELOPMENT growth rates in monthly household per capita income for each household of the distribution. The absolute GIC plots absolute increases in real Bolivianos for the whole period for each percentile. Included are (also in all other figures) the bootstraped 95% confidence intervals and the moderate and extreme poverty headcounts for 989 of 77% and 56%, respectively. As can be seen from the position and slope of the GIC, income growth was pro-poor in the weak absolute and relative sense since the curve is above for all and negatively sloped for nearly all percentiles. However, as expected, we do not find strong absolute pro-poor growth since the absolute GIC is positively sloped meaning that absolute increases in income were much higher for the non-poor than for the poor. The confidence intervals are extremely tight so that we can be quite confident about the statistical validity of our findings. Figure a shows the relative and absolute unconditional NIGIC for average education per household. Figure b shows for this variable the relative and absolute conditional (smoothed ) NIGIC. Whereas for the unconditional NIGIC the growth rates and absolute changes are shown in percentiles ( ), for the conditional NI- GIC the growth rates and absolute changes are shown in vintiles (, each comprising 5%). The reason for using vintiles instead for percentiles for the conditional NIGIC is to get a higher number of observations for each group when households are ranked by income. Even with income vintiles, the confidence intervals are much larger, due to the high heterogeneity of non-income achievements and growth rates thereof within each income vintile. For the unconditional NIGIC, we find pronounced weak absolute as well as relative propoor growth. 3 The relative pro-poorness of average education is also reflected by comparing the PPGR with the GRIM where the PPGR for moderate poverty is 3.89% and the PPGR for extreme poverty is.88, both much higher than the GRIM of.8% (Table ). The PEGR by Kakwani and Son (6) applied to education is also larger than the GRIM for all education indicators, when the extreme poverty line is used, supporting a relatively pro-poor pattern of expansion of education. For some of the education indicators, this is not the case when the moderate poverty line is used, suggesting that differences in the aggregation procedure inherent in these two pro-poor growth measures have some effects on the finding of pro-poor growth. The conditional NIGIC is more volatile than the unconditional NIGIC and also shows weak Percentile NIGIC unconditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC uncon. abs. 95% CI upper b. abs. 95% CI lower b. abs. Figure a. Unconditional NIGIC for average education (relative and absolute). Source: Own calculations.

15 MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS Vintile NIGIC conditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC con. abs. 95% CI upper b. abs. 95% CI lower b. abs. Figure b. Conditional NIGIC for average education (relative and absolute). Source: Own calculations. absolute and relative pro-poor growth but to a lower extent. Thus, the conditional NIGIC shows that the income-poor have experienced slightly higher educational growth than the average. This is also reflected in the higher PPGR (.% for moderate and.% for extreme poverty) compared to the GRIM (.8%). However, the wide confidence interval raises some doubts about the statistical significance of these findings. We do not find strong absolute pro-poor growth for the absolute unconditional NIGIC for education as the slope of the absolute curves in Figures and 3 are not negative, but even positive for the poorest deciles. This is quite interesting, because it puts the findings of the unconditional NIGIC in Figure a in perspective where we have found high relative pro-poor growth for the first three deciles. This seemingly contradictory finding is due to the fact that the high percentage growth rates for the lower deciles are due to the very low base in 989. The absolute conditional NIGIC is virtually flat, meaning that the income-poor have not been able to improve their educational attainment by more than the average. These findings are also reflected in comparing the PPCH with the CHIM. As Table shows, the unconditional PPCH is still larger than the CHIM, however, only slightly: the average years of schooling only increased by.8 years in the mean, by.3 years for the moderately poor, and.3 for the extremely poor. For the absolute conditional changes and for both poverty lines, the CHIM is higher than the PPCH of.. To take into account also possible intrahousehold inequalities Figures 3a and 3b show the unconditional and conditional NIGIC for individual education in single years. Both figures reflect the picture that was found for average education per household showing relative pro-poor growth both for the unconditional and for the conditional case. Another way to look at intra-household inequalities is to look at the gender gaps in education within households. For this purpose, we calculate the female minus male education in the households (in years of education). Ranking the households by the gap, we plot in Figure the unconditional and conditional absolute change in the gender gap. We find that the intra-household gender gaps were reduced for nearly all household except those between the th and the th percentile. Again, especially households in the middle of the distribution showed the strongest reductions. When looking at the conditional NIGIC, we find no clear trend meaning that the reduction in

16 36 WORLD DEVELOPMENT Percentile NIGIC unconditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC uncon. abs. 95% CI upper b. abs. 95% CI lower b. abs. Figure 3a. Unconditional NIGIC for individual education (relative and absolute). Source: Own calculations Vintile NIGIC conditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC con. abs. 95% CI upper b. 95% CI lower b. Figure 3b. Conditional NIGIC for individual education (relative and absolute). Source: Own calculations. gender gaps is equally distributed across all income groups (Figure b). Figures 5a and 5b show the results for average vaccination. The unconditional NIGIC shows pro-poor growth in the weak absolute and is also slightly negatively sloped. Table confirms the pro-poorness in the relative sense. Here, both PPGR exceed the GRIM and also

17 MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS 37 Change in Absolute Gap Percentile NIGIC uncon. abs. 95% CI upper b., uncon. abs. 95% CI lower b., uncon. abs. Figure a. Unconditional absolute gender gap in education. Source: Own calculations. Change in Absolute Gap Vintile NIGIC conditional 95% CI upper b., con. 95% CI lower b., con. Figure b. Conditional absolute gender gap in education. Source: Own calculations. the PEGR is larger when the extreme poverty line is used. However, improvements are relatively low which was also shown in Table. The conditional NIGIC shows no clear propoor growth trend, also visible in the wide confidence intervals. In addition, the PPGR are lower than the GRIM and for some deciles we even find deterioration. The same findings also hold for the absolute curves which show very small improvements over the time period. When examining the high relative growth in the unconditional NIGIC for education and

18 38 WORLD DEVELOPMENT Percentile NIGIC unconditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC uncon. abs. 95% CI upper b. abs. 95% CI lower b. abs. Figure 5a. Unconditional NIGIC for average vaccinations (relative and absolute). Source: Own calculations Vintile NIGIC conditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC con. abs. 95% CI upper b. abs. 95% CI lower b. abs. Figure 5b. Conditional NIGIC for average vaccinations (relative and absolute). Source: Own calculations. vaccinations, Figures 3a and 5a do not report growth rates for the very poor percentiles. This is due to two reasons. First, the very poor began and ended with no education and no vaccinations (see discussion below). Second, a sizable group started with no education or no vaccination and ended up having positive levels of education and vaccinations in the second

19 MEASURING PRO-POOR GROWTH IN NON-INCOME DIMENSIONS 39 period. But in this case the growth rate is not defined and thus not reported. Remember that the very high growth rates that appear on the graphs at the left are thus based on percentiles that have had some small amount of education and vaccinations and even a minute absolute expansion translates into a very high growth rate. Examining the absolute unconditional NI- GICs for education and vaccinations also reveal an important finding regarding the very low tail of the distribution. As Figures a, 3a, and 5a show, the very education-poor and vaccination-poor had no education (vaccinations) in the first period and this continued to be the case in the second period. This is true for the first few percentiles in the education indicator (e.g. the first 8% in Figure 3a) and the entire first decile in the vaccination indicator. Thus, whatever expansion has taken place in non-income improvements, it bypassed a core group of very poor. For all the other educational indicators shown in Tables and, the findings are qualitatively the same. 5 For both survival variables the unconditional NIGIC and the absolute NIGIC are only interpretable for the first few deciles where they show clear improvements in the sense of weak absolute and relative pro-poor growth, but they become flat from the th decile onward in the case of under 5 survival since % survival is already reached as shown in Figures 6a and 6b. Also the conditional NIGIC, which oscillate closely to but always above, reflects the moderate and more or less equally distributed mortality by the income groups. Also, the deciles of Table show only a small income gradient of the mortality risk. Figures 7a and 7b show the NIGIC for stunting. The unconditional NIGIC indicates weak absolute and relative pro-poor growth. For the conditional NIGIC we only find weak absolute but no relative pro-poor growth. 6 These results are also found when looking at the PPGR and the GRIM for the stunting z-score. Both absolute NIGICs show that the absolute changes are distributed nearly equally over the sample. Aggregating the several variables in the CWI (using the small sample), Figures 8a and 8b summarize the development of the social indicators in one single NIGIC. As expected, we find pro-poor growth in the weak absolute and relative sense for the unconditional NIGIC. Looking at Table, we find very high relative pro-poor growth Percentile NIGIC unconditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC uncon. abs. 95% CI upper b. abs. 95% CI lower b. abs. Figure 6a. Unconditional NIGIC for under 5 survival (relative and absolute). Source: Own calculations.

20 WORLD DEVELOPMENT Vintile NIGIC conditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC con. abs. 95% CI upper b. abs. 95% CI lower b. abs. Figure 6b. Conditional NIGIC for under 5 survival (relative and absolute). Source: Own calculations Percentile NIGIC unconditional (lhs) 95% CI upper b. (lhs) 95% CI lower b. (lhs) NIGIC uncon. abs. 95% CI upper b. abs. 95% CI lower b. abs. Figure 7a. Unconditional NIGIC for stunting (relative and absolute). Source: Own calculations. as both PPGR clearly exceed the GRIM and also the PEGR is larger than the GRIM, at least for the extreme poverty line. Despite being somewhat more volatile, the conditional NIGIC shows also pro-poor growth in the weak absolute but not in the relative sense.