Poverty & Equity Global Practice Working Paper 110 MULTIDIMENSIONAL POVERTY IN THE PHILIPPINES, 2004–13 DO CHOICES FOR WEIGHTING, IDENTIFICATION AND AGGREGATION MATTER? Gaurav Datt June 2017 Poverty & Equity Global Practice Working Paper 110 ABSTRACT Multidimensional poverty comparisons can be sensitive to the choice of welfare indicators, the weights assigned to the indicators, as well as the aggregate poverty measure used. This paper examines the robustness of trends in multidimensional poverty in the Philippines to these choices by presenting estimates for three alternative weighting schemes and three measures of multidimensional poverty. The weighting schemes range from uniform weights similar to those used in the global multidimensional poverty indexes produced by the United Nations Development Programme, to weights based on inverse incidence of different deprivations and those derived from the estimated relationship of deprivations to a survey-based measure of subjective welfare. The multidimensional poverty measures similarly range from the “dual cut-off” indexes analogous to the United Nations Development Programme’s global Multidimensional Poverty Index, to “union-based” indexes that count all deprivations, to indexes that are also responsive to the distribution of deprivations. Using data for 2004–13, the paper finds evidence of a significant decline in multidimensional poverty that is robust to these alternatives, although the magnitude of the decline in, and especially the dimensional contributions to, aggregate multidimensional poverty are quite sensitive to the alternatives considered. This paper is a product of the Poverty and Equity Global Practice Group. It is part of a larger effort by the World Bank to provide open access to its research and contribute to development policy discussions around the world. The authors may be contacted at gaurav.datt@monash.edu. The Poverty & Equity Global Practice Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. ‒ Poverty & Equity Global Practice Knowledge Management & Learning Team This paper is co-published with the World Bank Policy Research Working Papers. Multidimensional Poverty in the Philippines, 2004-13: Do Choices for Weighting, Identification and Aggregation Matter? Gaurav Datt* Keywords: multidimensional poverty; weighting; identification and aggregation; Philippines JEL codes: I3, I32, N35 * Department of Economics and the Centre for Development Economics and Sustainability, Monash University (gaurav.datt@monash.edu). Background paper prepared for the World Bank’s Programmatic Task on Poverty and Shared Prosperity in the Philippines. The author would like to thank Regina Salve Baroma, Carolina Diaz-Bonilla, Nandini Krishnan, Xubei Luo, Paul Mariano, Sharon Faye Piza and Mara Warwick, for useful comments or other forms of help. The paper represents the views of the author which should not be attributed to the World Bank. Multidimensional Poverty in the Philippines, 2004-13: Do choices for weighting, identification and aggregation matter? 1. Introduction The multidimensional nature of poverty has been widely noted for a long time. Over the last decade or so, however, both the theory and practice of measurement of multidimensional poverty have made rapid advances. Indeed, one measure of the growing recognition of multidimensional poverty is the publication of Multidimensional Poverty Indices (MPI) for more than 100 countries by the United Nations Development Programme (UNDP) in its annual Human Development Report beginning with HDR 2010. Conceptually, the case for measuring multidimensional poverty over and above poverty measured in terms of consumption or income is straightforward. It is rooted in viewing poverty as “capability failure”, the most notable exponent of this view being Sen (1980, 1985, 1999).1 Viewed as such, multidimensional poverty then invites direct attention to a range of specific capabilities including those relating to health, education, shelter, and access to basic amenities. The key consideration then turns on the question whether consumption or income offers an adequate representation of an acceptable range of capabilities. If it did, there would really not be much of a case for a separate focus on multidimensional poverty.2 Conversely, the case for a distinctive focus on multidimensional poverty rests in the proposition that there are many aspects of poverty which, for a variety of reasons, may not be adequately captured by income or consumption. It is presently not necessary to dwell on these reasons which for the most part have to do with market failures of one sort or another.3 In addition, it is 1 Sen’s writings on this subject are many; referenced here are only a few examples (including one of the early ones). 2 Even from the welfarist perspective of consumption expenditure as a money-metric of utility, consumption may be inadequate as there are other arguments in individual utility functions for which markets (and hence prices) may not exist or if they exist may be distorted. 3 Even in the absence of market failures, total consumption or income as a linear combination of prices and quantities of goods and services implies perfect substitutability between these goods and services – an assumption that is questionable from a non-welfarist or human rights perspective that insists on the essentiality of minimum levels of achievement across a range of dimensions. 2 also worth noting that a focus on multidimensional poverty has the potential to draw attention to “government failures” too, especially in relation to provision of public goods and how markets function. While these conceptual arguments for an independent focus on measuring multidimensional poverty apply to all countries, there is also a significant framing issue that is especially relevant for the Philippines. During 2003-14, the Philippine economy grew at average rate of 5.4% per year. While there was fluctuation in growth over this period, per capita GDP grew by 48% and per capita consumption grew by 37% in real terms or at average annual growth rates of 3.5% and 2.9% respectively. However, the official (income) poverty estimates over this period largely indicate a flat line, changing little from 28% in 2003 to 26% in 2014 (see Figure 1).4 Why economic growth during 2003-14 did not translate into significant income poverty reduction has been puzzling. Of course, a decade is a relatively short period of time, and some of the investment activity may augment household incomes with a lag. Most recent estimates based on the 2015 Family Income and Expenditure Survey indicate a fall in poverty incidence to 21.6% (PSA, 2016). Nonetheless, this experience begs the question of how non-income indicators of poverty – including those related to health, education, quality of housing and access to services – have progressed over this period. This question has already initiated an investigation of trends in multidimensional poverty based on a range of non-monetary indicators of welfare. For instance, using data from three different sets of household surveys – the National Demographic and Health Survey (NDHS), Family Income and Expenditure Survey (FIES), and the Annual Poverty Indicators Survey (APIS), Balisacan (2015) constructs a series of multidimensional poverty indices. Defining poverty as deprivation in at least one-third of the chosen indicators, he finds that rates of multidimensional poverty reduction were higher than income poverty reduction during 1998-2011/12,5 leading him to conclude that “aggregate poverty, seen from the lens of multidimensional deprivation, actually declined as the economy expanded during the past decade” (p. 463). 4 These official estimates are based on per capita incomes from the first-semester Family Income and Expenditure Surveys (FIES) for 2003, 2006, 2009 and 2012, and the Annual Poverty Indicators Surveys (APIS) for 2013 and 2014. The APIS surveys are only conducted for the first semester of the survey year. Since 2013, the APIS has used the same income module as the FIES. 5 Balisacan (2015) in fact presents estimates for a longer period for some of his data sources. In particular, the study uses FIES data for the years 1998, 1991, 1994, 1997, 2000, 2003, 2006, 2009 and 2012; NDHS for 1993, 1998, 2003 and 2008; and APIS data for 1998, 2002, 2004, 2007, 2008, 2010 and 2011. 3 However, as is well-understood in the literature, multidimensional poverty comparisons are sensitive to assumptions in relation to the choice of indicators, the weights assigned to the indicators, the dimensional (deprivation) and the overall poverty cut-offs, as well as the choice of the aggregate poverty measure. For the Philippines, Balisacan (2015) already looks into some aspects of this robustness, for instance, variations in the choice of indicators across the three data sources, and sensitivity to alternative poverty cut-offs. However, the issues of alternative weighting schemes and poverty measures have remained underexplored. Using data from six rounds of the Annual Poverty Indicators Survey (APIS) conducted by the Philippines Statistics Authority (previously, the National Statistics Office) over the period 2004-13, this paper complements Balisacan (2015) in further investigating the robustness of trends in multidimensional poverty to methodological choices for the Philippines. In particular, the paper focuses on the choice of alternative weighting schemes and aggregate poverty measures, empirical investigation of which is relatively limited even for other countries. In addition, the paper also explores the implications of including an employment-based dimension among welfare indicators. The analysis thus helps reassess evidence on the larger question of the changes in poverty in the past decade by examining whether the absence of a trend in income poverty is also mirrored in the trends of multidimensional poverty over this period. Our main finding is that there is evidence of a significant decline in multidimensional poverty measures in the country during 2004-13 that appears robust across a range of alternatives considered. Dimensional contributions to aggregate multidimensional poverty however vary enormously across alternatives. The paper is organized as follows. Section 2 spells out the data and the methodology. In particular, it discusses the choice of dimensions and indicators, alternative approaches to setting weights for different indicators as well as alternatives for aggregate poverty measures, in particular, the use of “distribution-sensitive” measures. Section 3 presents the results of implementing these with the APIS data for 2004-13. It presents summary trends in individual indicators over this period, three sets of weighting schemes (two of which are estimated from APIS data for the initial year), and finally the estimates of multidimensional poverty measures as well as dimensional decompositions for the range of alternatives considered. Section 4 offers some concluding observations. 4 2. Data and methodology There now exists a large literature on the measurement of multidimensional poverty offering a variety of measurement methodologies; see Alkire et al. (2015) for a recent survey. Cutting across these different approaches, however, there are broadly three sets of issues that need to be addressed in any measurement framework for multidimensional poverty. These relate to: 1) Dimensions? Which indicators should be included in a multidimensional poverty measure? What sort of dimensional cut-offs should be used to define deprivation in different dimensions? 2) Weights? How should the different dimensions be weighted? 3) Identification and aggregation? How should dimensional deprivations of individuals be aggregated into an overall measure of multidimensional poverty? This includes the issue of how the multidimensionally-poor should be identified, in particular the use of a cross-dimensional (or poverty) cut-off as for instance in the counting approach of Alkire and Foster (2011), as well as the issue of how the multiple deprivations of those identified as poor should be aggregated over the population (or a sub-population). We discuss our approach to considering different options for each of these issues below, though, as mentioned in the Introduction, a key focus will be on alternative approaches to assigning weights to different dimensions and on the use of alternative poverty measures.6 2.1 Dimensions and indicators Our choice of dimensions and indicators to be included in the measurement of multidimensional poverty is guided both by common empirical practice as well as particular data considerations for the Philippines. Since multidimensional poverty is concerned with the joint distribution of deprivations in different dimensions, a key data consideration is the ability to observe deprivations across these dimensions for the same set of households or individuals. This limits the range of data sources that can be used for such analysis, and conversely, a given data source may limit the set of includable dimensions to those that can be readily measured from that source. In this paper, we use data from six rounds of the Annual Poverty Indicators Survey (APIS). With the exception of anthropometric information which it does not have, the APIS is quite well-suited to the task of multidimensional poverty measurement. As 6 Most of the literature has concentrated on the identification and aggregation issues with the choice of dimensions and weights having remained relatively under-explored. 5 noted before, data from two other sources – the National Demographic and Health Survey and the Family Income and Expenditure Survey – can also be used for constructing multidimensional poverty measures. However, their more limited frequency – once every five years for NDHS and once every three years for FIES – implies that only 2-3 rounds of these surveys are available every decade. By comparison, for the 2004-13 period covered by this study, we are able to use six rounds of APIS for the years 2004, 2007, 2008, 2010, 2011 and 2013. A notable benchmark in terms of empirical practice is the global Multidimensional Poverty Index (MPI) adopted by UNDP and now estimated for over 100 countries – including the Philippines. The UNDP’s MPI is based on the counting approach proposed by Alkire and Foster (2011), and uses ten indicators covering three broad dimensions related to education, health and what is dubbed as the “standard of living”. Education and health have two indicators each, and there are six indicators of the standard of living related to access to services (especially, electricity, water and sanitation) and asset ownership (see UNDP, 2014 for details). The choice of indicators used in this paper broadly follows the global MPI. While the same three main dimensions of health, education and standard of living are identified as in the global MPI, the specific indicators under each had to be adapted to what could be implemented with the data available from the APIS surveys. In particular, this paper presents results using the set of indicators as shown in Table 1. The two indicators for education deprivation – viz., if no one in the household has completed six years of schooling, and if any school-age child in the household is not attending school – are essentially the same as have been used in the global MPI, and do not require further explanation. In relation to health indicators, as noted above, APIS does not collect anthropometric information, nor information on child mortality. Instead, there are two other potential health indicators available from the APIS. First, as APIS has information on consumption, the adequacy of a household’s per capita food consumption can be assessed relative to the official food poverty thresholds. The indicator we use (“health1” in Table 1) deploys a stringent criterion of nutritional deprivation defined as per capita food consumption of the household falling below 80% of the food poverty threshold. The second potential health indicator (“health2” in Table 1) is defined in terms of past (self-reported) illness or injury of household members. There is however a data comparability issue related to this 6 indicator due to a change in the reference period. In the APIS surveys up to the round for 2011, the reference period for this question was the “past month”; however, in the 2013 round this was changed to the “past six months”, thus leading to much higher rates of reported morbidity. In light of this, our main results up to 2013 exclude “health2”, while we also report estimates including it, but the results for these are valid for the period up to 2011 only. The first five indicators related to the standard of living are also very similar to those in the global MPI, and do not require any special comment. The global MPI also includes a deprivation indicator based on the use of inferior sources of cooking fuel. The APIS data do not offer a direct measure for such an indicator. It is possible to proxy this with information on whether the household owns a gas range or stove with oven – one of the asset ownership items included in the questionnaire. However, this seems to be a rather noisy indicator with implausible fluctuations over certain rounds of the survey; for instance, the proportion of population in households not owning a gas range or stove with oven registered a sharp increase between APIS 2011 and 2013, from 78% to 87%. While the reasons for this are not entirely clear, the possibility that there was a sudden and marked deterioration in the use of improved sources of cooking fuel seems unlikely. Hence, we opted not to include this indicator in this assessment. Note that the final indicator in Table 1 relates to employment – a dimension that is not in included the UNDP MPI nor typically in most applications. Thus, a brief comment on the rationale for its inclusion is apposite. Employment generates income, so its pecuniary significance is obvious. But employment may also be considered an independent dimension in its own right as it represents more than just a means to earning a living. Sociologists have long recognized the importance of employment as a basis for social differentiation; see for instance, the centrality of “weak labor force attachment” as a defining characteristic of the “underclass” in the pioneering work of Wilson (1987). The linkage of employment with life satisfaction, happiness, self-esteem, social status, mental and physical health and wellbeing has also been widely documented.7 There is also a well-established tradition of using employment-based indicators in social exclusion measures in OECD countries; see for instance, indicators of quasi- joblessness included in EU-2020 measures/ targets for poverty and social exclusion. Employment-based 7 See for instance, Dooley, Catalano, and Wilson (1994), Dooley, Prause, and Ham-Rowbottom (2000), Helliwell and Putnam (2004), Stutzer and Lalive (2004), Lundin and Hemmingsson (2009), Stuckler et al. (2009a, 2009b), ILO (2010), Blanchflower and Oswald (2011). 7 indicators are also increasingly being included in National MPIs in many developing countries.8 There is thus a case for considering inclusion of the employment dimension in multidimensional welfare assessments. 2.2 Weights While a number of different approaches to weighting of indicators have been considered in the literature,9 in this paper we explore three of them, as described below. i) Nested Uniform Weights This set of weights follows the convention adopted by the UNPD for its global MPIs. Under this convention, the three main dimensions – corresponding to education, health and the standard of living – are first assigned equal weights of one-third each. Then, all indicators within a main dimension are also assigned equal weights. Thus, for instance, when using the 10 indicators in Table 1, the two education indicators have weights of 1/6th each, the same for the two health indicators, and the six indicators for the standard of living have weights of 1/18 each. When using 9 indicators (omitting health2), the weight for the sole health indicator (health1) becomes 1/3rd, while the weights for the other indicators remain unchanged. ii) Nested Inverse Incidence Weights Similar to the idea in the construction of an asset index where a more commonly-owned asset is deemed less valuable, this approach has been proposed in Desai and Shah (1988) and Cerioli and Zani (1990). The weights assigned to different deprivation indicators in such a weighting scheme vary inversely with the prevalence of those deprivations in the population; the higher (lower) the prevalence or incidence, the lower (higher) the weight. There is however a potential issue in that the weights will be sensitive to the number of indicator used to represent a certain type of deprivation. For instance, if one were to use four instead of, say, two indicators to represent education deprivation, then other 8 There are many examples. For instance, for Colombia’s MPI, two of the 15 indicators relate to the absence of long-term unemployment and formal employment, and these indicators are included under “Labour” as a separate dimension. For South Africa, “Economic Activity” is included as a separate dimension, with an indicator on unemployment among 15-64 year old adults. The Ecuador MPI has “Work and Social Security” as a separate dimension, with an indicator on unemployment or inadequate employment among those 18 years or older. Similarly, for Costa Rica, “Work” is included as a separate dimension, with indicators on long-duration unemployment, informal employment, and earnings less than the minimum wage. Employment dimension or indicators are also included in MPIs for El Salvador and Chile. 9 See Decanq and Lugo (2013) for a review. 8 things being equal, this will raise the combined weight of education deprivation indicators. Without any correction, the approach will ceteris paribus tend to favor dimensions represented by a larger number of indicators. To guard against this, we thus use a nested version of the inverse incidence approach, which in this respect is similar to the nested approach to uniform weights used in the global MPI. The nested inverse incidence weights are thus determined in two steps. First, we begin with the three main dimensions relating to education, health and the standard of living and livelihood, and establish their weights based on inverse incidence. In the second step, we establish inverse incidence weights for individual indicators within any main dimension such that their sum equals the weight established for that main dimension in the first step. These weights are determined for 2004 using APIS data for that year, and then kept fixed over time. iii) Subjective Welfare Weights This approach generates weights from an estimated relationship between a measure of subjective welfare (obtained through a survey) and the indicators that are sought to be included in a multidimensional welfare index.10 For constructing these weights, we exploit data from the 2004 APIS survey relating to the following question in the survey: “Imagine a ladder with ten steps. The first represents the poorest in society and the tenth represents the richest. On what step of the ladder would you be?” The percentage of the population at different steps of the ladder turned out to be as below: Step (poorest to richest) 1 2 3 4 5 6 7 8 9 10 % of population 6.3 15.8 21.9 17.9 25.6 7.7 3.3 1.2 0.2 0.2 Based on this, we define a binary subjective poverty variable which takes a unit value (and zero otherwise) if a household perceives themselves to be on the second rung of the ladder or lower.11 We then estimate a logit regression of subjective poverty on the 9 (or 10) multidimensional poverty indicators in Table 1. The marginal effects of indicators estimated from this regression are shown in 10 See Bellani (2013), Bellani et al. (2013), Decanq et al. (2013) and Mitra et al. (2013) for related applications of this approach. 11 This is arbitrary but yields a subjective poverty incidence of about 22% that is not very different to the official income poverty incidence of 25% (for 2003 based on FIES). 9 Table 2.12 The subjective welfare weights are then constructed as proportional to the estimated marginal effects. 2.3 Identification and aggregation Beyond the choice of indicators and their weights, the other important considerations in measuring multidimensional poverty have to do with the well-known issues of identification and aggregation. The first has to do with how, given the multiplicity of indicators, should the poor be identified, and the second has to do with how deprivations across individuals and dimensions be aggregated into a scalar measure of multidimensional poverty. Within the global MPI framework, following Alkire and Foster (2011), the poor are identified by reference to a cross-dimensional cut-off specified in terms of the minimum percentage of (weighted) dimensions a person must be deprived in for him/her to be considered poor. While the Alkire-Foster approach is more general (potentially allowing for the cross-dimensional cut-off to range from the minimum weight of any dimension to 100 percent), the global MPI sets this at one-third. There is thus the possibility of using more than one value for the cross-dimensional cut-off, or of not using the cross- dimensional cut-off at all. The former is in the spirit of investigating robustness of comparisons of multidimensional poverty to the choice of this cut-off. This is explored in Balisacan (2015) who finds a measure of robustness with respect to this choice. The latter takes us to the so-called “union approach” to identification, where a person is considered multidimensionally-poor if deprived in any dimension. The union approach asserts the essentiality of all deprivations. It also has the merit that it ensures regressive transfers from a more to a less deprived person are poverty-increasing – a property that is not always satisfied by measures that use cross-dimensional cut-offs.13 In this paper, we will present multidimensional poverty measures using both a cross-dimensional cut-off of one-third as well as union- based measures. Thus, from an identification perspective, the two measures we consider can be written as: 1 ( ) ( ) = (1) 12 The estimation of marginal effects takes into account the fact that the independent variables in the regression are also dichotomous. 13 See Datt (2017) for a formal statement of this property as well as further discussion of why the union approach may be worth considering. 10 where the multidimensional poverty measure ( ) is a function of the cross dimensional cut-off , where is a binary (0-1) variable denoting whether individual is deprived in dimension with weight ( ) among a total of dimensions and a population of size , and is a binary (0-1) variable denoting whether individual is deprived in at least fraction of the weighted dimensions. In particular, we will construct measures for = 1⁄3 , { }. In the union-case ( = { }), the multidimensional poverty measure can also be written more simply as: 1 ( ) = (2) Thus, the difference between the measures in (1) and (2) is that while ( ) counts all deprivations of all individuals, ( ) counts the deprivations of only those who are deprived in at least fraction of all weighted dimensions. There is one further aggregation issue that is worth considering, and it relates to the dispersion or inequality in the distribution of deprivations among the poor. This can be described by first noting that either of the above measures is equivalent to the ratio of the weighted sum of the deprived dimensions of the poor to the sum of all possible deprivations (when everyone is deprived in every dimension). Since the weights do not vary across persons and these measures aggregate deprivations by simply summing them up, they depend only on the total number of deprivations not their dispersion or concentration. Whether deprivations among the poor are evenly distributed or concentrated among a few, ( ) and ( ) return the same value of multidimensional poverty. In view of this, we also consider an alternative measure that is sensitive to the distribution of deprivations among the poor: 1 ( ; ) = ≥ 1 (3) where is interpretable as an “inequality-aversion” parameter such that for > 1, a higher concentration of a given set of deprivations yields a higher value for multidimensional poverty.14 Note that for = 1, ( ; ) coincides with ( ). In the following, we will thus present three multidimensional poverty measures: ( = 1⁄3), ( ; = 1) and ( ; = 2). 14 See Datt (2017) for the properties of this class of multidimensional poverty measures; also see Chakravarty and D’Ambrosio (2006) who proposed such a measure of social exclusion. 11 3. Results 3.1 Summary trends in indicators and alternative weighting schemes Table 3 shows the trends in the deprivation indicators over time. The data suggest significant declines in the deprivation indicators related to education and the standard of living and livelihood over the period 2004-13, though the improvement in water and sanitation appears limited to first sub-period 2004-07 with little or no change thereafter. The record of the two health indicators is mixed. While the illness/injury-related indicator shows a decline during 2004-11 (note that we do not have a comparable estimate for 2013 due to the change in the reference period for this variable, as discussed before), there is virtually no change in the food-consumption-based nutrition indicator over the entire period. Overall, since most indicators show an improvement over time, one may expect multidimensional poverty to register a decline over this period. However, to what extent that is the case or not will also depend upon changes in the joint distribution of indicators as well as on how these indicators are weighted. The set of weights corresponding to the three weighting schemes discussed above are shown in Table 4 for both the 9- and 10-indicator cases. Recall that the nested inverse incidence (NII) weights are based on the 2004 incidence rates for the deprivation indicators, and are kept fixed over time. The subjective welfare (SW) weights are also estimated for 2004 based on the socio-economic ladder question, and are also kept fixed over time. The nested uniform (NU) weights are fixed over time by construction. Table 4 shows some notable differences across the three sets of weights. (For comparison, the table also shows the weights in Balisacan, 2015.15) Looking first at the three main dimensions, relative to the nested uniform (NU) weights, the nested inverse incidence (NII) weights accord a higher value to education and health, while the subjective welfare (SW) weights accord a lower value to both. Indeed, the SW weights put the highest value on the standard of living and livelihood dimension, and lowest on the health dimension. By contrast, the NII weights put the highest value on education, and the lowest on the standard of living and livelihood. This pattern holds both in the 9- and 10-indicator cases. Looking at individual indicators, relative to the NU weights, the NII weights accord a premium to the two education indicators at the expense of the standard of living and livelihood indicators, while the SW 15 Note that Balisacan (2015) does not use exactly the same indicators as in this paper. The table reports his weights for the APIS application, lining them up as closely as possible against the indicators in this paper. 12 weights accord a premium to sanitation, electricity, roof-wall of the dwelling and assets at the expense of school attendance, water and employment, and the health indicators. Note also that under the 9- indicator scenario, there is only one health indicator related to food consumption, and the NU and NII weights accord primacy to this indicator with a weight of one-third or higher, which in turn has the implication of making deprivation in food consumption a sufficient condition for being deemed multidimensionally-poor when using the cross-dimensional cut-off of one-third of weighted dimensions. By contrast, the SW weighting scheme assigns a weight of less than one-fifth to this indicator. The employment indicator is also accorded a very low weight (of 2%) under the SW weighting scheme, while the NII (NU) weight for employment is about twice (nearly thrice) as high. Thus, there is considerable variation across these weighting schemes. The following results look at the implications of this for the assessment of multidimensional poverty for the Philippines. 3.2 Trends in multidimensional poverty The results on trends in multidimensional poverty based on 9 indicators are presented in Table 5 and those based on 10 indicators in Table 6. In both cases, results are presented for the three alternative weighting schemes and for the three alternative poverty measures. The first important finding is that between 2004 and 2013 (2011 when using 10 indicators) there is a statistically significant decline in multidimensional poverty for all three weighting schemes and all three poverty measures. Thus, the reduction in multidimensional poverty over this period seems fairly robust to alternative choices with respect to weighting, identification and aggregation. Nonetheless there are differences in the magnitude of decline across the weighting schemes. For measures based on 9 indicators, the fall in multidimensional poverty over 2004-13 is largest for SW weights, while the magnitude of decline is similar though smaller for NU and NII weights.16 Thus, for instance, the ( = 1⁄3) measure with SW weights fell by 42% between 2004 and 2013, while the decline in this measure with NU and NII weights was around 11-13%. Similarly, the ( ; = 1) and ( ; = 2) measures with SW weights fell by 26 and 37 percent respectively, while the corresponding declines were smaller at 14-18% with NU weights and 22-25% with NII weights. This largely reflects the fact that the SW weights are more positively correlated with the absolute reduction 16 Using the 10 indicators, however, the decline in multidimensional poverty over 2004-11 is of similar magnitude across the three weighting schemes. But this also omits the last year 2013 due to the comparability issue with the illness/injury-related indicator. 13 in deprivation indicators. That is, deprivation indicators that fell the most, e.g. those related to assets and electricity, have relatively higher SW weights, while the deprivation indicator related to food consumption which showed little change has a much lower SW weight. How do these results compare with Balisacan’s (2015) APIS-based estimates of multidimensional poverty (also reproduced in the last column of Table 6)? Among our estimates, the ones most comparable with Balisacan (2015) are those based on 10 indicators, using NU weights and ( = 1⁄3) as the multidimensional poverty measure for the period 2004-11. We find a fall in this measure by 26% between 2004 and 2013; by comparison, Balisacan reports a decline of 19% over the same period. Thus, the estimates in this paper corroborate the results in Balisacan, though we find evidence of a larger decline in multidimensional poverty. The decline is steeper if we use the subjective welfare weights, and as noted above, steeper still if the measures are based on 9 indicators. 3.3 Dimensional contributions to multidimensional poverty It is also instructive to look at how the contributions of different dimensions and indicators to aggregate multidimensional poverty vary by the weighting scheme and the aggregate measure used. Dimensional contributions can be evaluated using Shapley decomposition methods (Shorrocks, 2013); see Datt (2017) for how these methods can be applied to dimensional decomposition of multidimensional poverty measures. The decomposition evaluates the contribution of a particular dimension as the expected reduction in the multidimensional poverty measure (. . ) due to the elimination of deprivation when the expectation is taken over all possible elimination paths (ranging from being the first deprivation to be removed to its being the last, and for all possible combinations of the presence or absence of other deprivations). The results of this dimensional decomposition (for the 9 indicator case) are shown in Table 7 for alternative weighting schemes and aggregate multidimensional poverty measures. As seen in Table 7, dimensional contributions are highly sensitive to both the choice of weighting scheme as well as the aggregate poverty measure, especially the former. For instance, using nested uniform weights and ( = 1⁄3), the health dimension (based on the food consumption indicator) accounts for 82% of total multidimensional poverty in 2013. This is due, in part, to the relatively high incidence of food consumption deprivation (Table 3), but more importantly it reflects the large weight of 1/3 for this indicator (see Table 4). With the cross-dimensional cut-off also set at = 1⁄3, this large weight ensures that everyone deprived in food consumption is necessarily multidimensionally poor too, 14 while the same is not true of any other indicator. The nested inverse incidence scheme assigns a still higher weight of 0.37 to the food consumption indicator, and hence its contribution to ( = 1⁄3) is even higher at almost 87%. By comparison, the contribution of the food consumption indicator for both NU and NII weights is moderated to less than 70% when union-based measures, ( ; = 1) and ( ; = 2), are used. This is because union-based identification now allows other deprivations to be counted in more often. A major contrast in dimensional contributions is however observed with respect to subjective welfare weights. With SW weights, the standard of living and livelihood dimension accounts for the bulk of aggregate multidimensional poverty, its contribution in 2013 ranging between 53 and 62 percent. Education accounts for 9-12 percent, while health (food consumption) accounts for 26-38 percent. Alongside other considerations, these results on dimensional decomposition can help inform judgements about the choice of weighting schemes and aggregate measures. For instance, a combination of NU or NII weights with dual cut-off measures (with set at one-third) renders multidimensional poverty in the Philippines almost entirely a matter of food consumption deprivation. If this appears excessive, the dimensional decompositions remind us that it is largely a consequence of the particular weighting schemes and measures used, and it may be worthwhile to consider other reasonable alternatives. 4. Conclusion Official estimates of income poverty for the Philippines over the decade or so since 2003 have indicated little decline despite respectable rates of economic growth over this period. Though the most recent estimates show a decline in poverty incidence in 2015, the earlier trend led to conjectures on whether this had been a lost decade for poverty reduction for the country. This concern also prompted inquiry into trends in non-income indicators of welfare to assess whether progress in the reduction of multidimensional poverty has fared any better. Recent work by Balisacan (2015) has suggested better performance on the multidimensional front. This paper has sought to contribute further to this line of investigation by constructing estimates of multidimensional poverty over the period 2004-13 using data from six rounds of the Annual Poverty Indicators Survey. A key emphasis of the paper has been on 15 examining the robustness of trends in poverty to alternative weighting, identification and aggregation choices in the construction of multidimensional poverty measures. In particular, the paper presents estimates for three alternative weighting schemes for different dimensions, and using three alternative measures of multidimensional poverty. The weighting schemes range from uniform weights broadly following the approach of global multidimensional poverty indices (MPI) produced by the UNDP, to weights based on inverse incidence of different deprivations and those derived from the estimated relationship of deprivations to a survey-based measure of subjective welfare. The multidimensional poverty measures similarly range from the “dual cut-off” indices analogous to UNDP’s global MPI, to “union-based” indices that count all deprivations, to indices that are also responsive to the distribution of deprivations. The main finding of the paper is that there is evidence of a significant decline in multidimensional poverty that is robust to these alternatives, although the magnitude of decline differs across the alternatives. The reduction in multidimensional poverty tends to be largest for subjective welfare weights and for union-based distribution-sensitive measures. The results thus corroborate and supplement the findings in Balisacan (2015). Notwithstanding the absence of a significant trend in income poverty, the period 2004-13 has witnessed improvement in several non-income dimensions of poverty in the Philippines. An examination of the factors underlying different trends in income and multidimensional poverty however remains a topic for further research. The paper also finds that the contributions of different dimensions to aggregate multidimensional poverty can vary enormously with the choice of the weighting scheme and the aggregate poverty measure. Dimensional decompositions of multidimensional poverty such as those presented in this paper alert us to particular aspects of the judgments implicit in these choices, and thus offer further input into making these choices. 16 References Alkire, S. and J.E. Foster. 2011. Counting and multidimensional poverty measurement. Journal of Public Economics, 95(7-8): 476-487. Alkire S., Foster, J.E., Seth, S., Santos, M.E., Roche, J.M., and P. Ballon. 2015. Multidimensional Poverty Measurement and Analysis. Oxford University Press, Oxford. Balisacan, Arsenio M. 2015. The Growth-Poverty Nexus: Multidimensional Poverty in the Philippines. Chapter 24 in Balisacan, Arsenio M., Chakravorty, Ujjayant and Majah-Leah V. Ravago (eds.). Sustainable Economic Development: Resources, Environment, and Institutions. Academic Press. Oxford. Bellani, L. 2013. Multidimensional indices of deprivation: the introduction of reference groups weights. Journal of Economic Inequality 11, 495-515. Bellani, L., Hunter, G., and P. Anand. 2013. Multidimensional Welfare: Do Groups Vary in Their Priorities and Behaviours?. Fiscal Studies 34, 333-354. Blanchflower, David G. and Andrew J. Oswald. 2011. International Happiness: A New View on the Measure of Performance. Academy of Management Perspectives, 25: 6-22. Cerioli, A. and S. Zani. 1990. A fuzzy approach to measurement of poverty. In: Dagum, C. and M. Zenga (eds.) Income and Wealth Distribution, Inequality and Poverty. Springer, Berlin, 272-284. Chakravarty, Satya R. and C. D'Ambrosio. 2006. The measurement of social exclusion. Review of Income and Wealth 523, 377–398. Datt, Gaurav. 2017. Distribution-sensitive multidimensional poverty measures. World Bank Economic Review (forthcoming). Decanq, Koen and Maria Ana Lugo. 2013. Weights in Multidimensional Indices of Wellbeing: An Overview. Economic Review, 32: 721-746. Decancq, K., Van Ootegem, L., and E. Verhofstadt. 2013. What If We Voted on the Weights of a Multidimensional Well-Being Index? An Illustration with Flemish Data. Fiscal Studies 34, 315- 332. Desai, M. and A. Shah, A. 1988. An Econometric Approach to the Measurement of Poverty. Oxford Economic Papers 40, 505-522. Dooley, David, Catalano, Ralph and Georjeanna Wilson. 1994. Depression and Unemployment: Panel Findings from the Epidemiologic Catchment Area Study. American Journal of Community Psychology 61 (3): 745-65. 17 Dooley, David, Prause, JoAnn and Kathleen Ham-Rowbottom. 2000. Inadequate Employment and High Depressive Symptoms: Panel Analyses. International Journal of Psychology 35: 294. Helliwell, John, and Robert Putnam. 2004. The Social Context of Well-Being. Philosophical Transactions of the Royal Society B 359: 1435-46. ILO (International Labour Organization). 2010. Emerging Risks and New Patterns of Prevention in a Changing World of Work. Geneva: ILO. Lundin, Andreas, and Tomas Hemmingsson. 2009. Unemployment and Suicide. The Lancet 374: 270–1. Mitra, S., Jones, K., Vick, B., Brown, D., McGinn, E., and M.J. Alexander. 2013. Implementing a Multidimensional Poverty Measure Using Mixed Methods and a Participatory Framework. Social Indicators Research 110, 1061-1081. Philippines Statistics Authority (PSA). 2016. Poverty Incidence among Filipinos registered at 21.6% in 2015. Press Release, 27 October. PSA-SSO-SSSS-PHDSD-1610-04. Sen, Amartya. 1980. Equality of What? (Lecture delivered at Stanford University, 22 May 1979), in MacMurrin, Sterling M., The Tanner lectures on human values 1, Salt Lake City, Utah: University of Utah Press. _____. 1985. Commodities and Capabilities. New York, NY: North-Holland. _____. 1999. Development as Freedom. New York: Oxford University Press. Shorrocks, A.F. 2013. Decomposition procedures for distributional analysis: a unified framework based on the Shapley value, Journal of Economic Inequality, 11(1): 99-126. Stuckler, David, Sanjay Basu, Marc Suhrcke, and Adam Coutts. 2009a. The Public Health Effect of Economic Crises and Alternative Policy Responses in Europe: An Empirical Analysis. The Lancet 374: 315-23. _____. 2009b. The Health Implications of Financial Crisis: A Review of the Evidence. Ulster Medical Journal 78 (3): 142-5. Stutzer, Alois, and Rafael Lalive. 2004. The Role of Social Work Norms in Job Searching and Subjective Well-Being. Journal of the European Economic Association 2 (4): 696-719. UNDP. 2014. Human Development Report 2014: Sustaining Human Progress: Reducing Vulnerability and Building Resilience. Macmillan, New York. Wilson, W.J. 1987. The Truly Disadvantaged: the Inner City, the Underclass, and Public Policy. University of Chicago Press, Chicago. 18 Figure 1: TRENDS IN REAL PER CAPITA INCOME AND CONSUMPTION (FROM THE NATIONAL ACCOUNTS) AND OFFICIAL POVERTY RATES (BASED ON SURVEYS) 40 150 Per capita GDP 145 35 Per capita GNP 140 Index of real per capita GNP/ GDP/ HFCE Official incidence of %of population below the poverty poverty (1st 135 30 semester) Per capita HFCE 130 25 125 120 20 115 110 15 105 10 100 2003 2006 2009 2012 2013 2014 Note: HFCE refers to household final consumption expenditure from the national accounts. Source: Based on official poverty estimates from the National Statistics Coordination Board, and national accounts statistics from the Philippines Statistics Authority. 19 TABLE 1: DIMENSIONS AND INDICATORS FOR MULTIDIMENSIONAL POVERTY ASSESSMENT FOR THE PHILIPPINES Main Variable Indicator Description dimension name Education educ1 Primary No household member has completed at least six years of schooling schooling educ2 School A school-age child (up to grade 10, i.e. between age 6-16) is not attendance attending school Health health1 Food If per capita food consumption < 4/5 of food poverty line consumption health2 Illness If more than 50% of household members report illness or injury over the past month (past 6 months for 2013) Standard of sliv1 Sanitation If household does not use own flush or closed pit toilet living and sliv2 Water If household's main source of water is not piped water or livelihood protected well sliv3 Electricity If there is no electricity in the house sliv4 Roof-wall If roof and outer wall is not made of predominantly strong materials sliv5 Assets If household does not have at least one communication asset (among phone, tv, radio, or PC) AND at least one mobility asset (among car, motorcycle or motorboat) OR at least one livelihood asset (among agricultural land, livestock, refrigerator/freezer) sliv6 Employment If less than 50% of working age members who are not students worked over the past 6 months Source: APIS survey instruments and data sets for rounds 2004 through 2013. 20 TABLE 2: MARGINAL EFFECTS OF DEPRIVATION INDICATORS ON SUBJECTIVE POVERTY (2004) 9 indicators 10 indicators Marginal Marginal Indicator Std. Err. Std. Err. effect effect educ1 Primary schooling 0.088 0.008 0.084 0.008 educ2 School attendance 0.027 0.007 0.028 0.007 health1 Food consumption 0.095 0.006 0.096 0.006 health2 Illness 0.036 0.006 sliv1 Sanitation 0.055 0.005 0.054 0.005 sliv2 Water 0.009 0.005 0.009 0.005 sliv3 Electricity 0.069 0.007 0.068 0.007 sliv4 Roof-wall 0.055 0.006 0.055 0.006 sliv5 Assets 0.089 0.007 0.088 0.007 sliv6 Employment 0.009 0.006 0.010 0.006 Note: The dependent variable is a measure of subjective poverty defined as a binary variable taking the value 1 if the household reports being on the lowest two steps of a 10-step economic ladder, zero otherwise. Source: Estimated using data from APIS 2004. 21 TABLE 3: INCIDENCE OF DEPRIVATION, 2004-2013 Fraction of population deprived Dimension/ indicator 2004 2007 2008 2010 2011 2013 Education educ1 Primary schooling 0.064 0.051 0.050 0.048 0.045 0.040 educ2 School attendance 0.137 0.139 0.126 0.118 0.104 0.082 Total education 0.180 Health health1 Food consumption 0.253 0.299 0.233 0.274 0.281 0.261 health2 Illness 0.100 0.116 0.075 0.091 0.080 0.257 Total health 0.327 Standard of sliv1 Sanitation 0.237 0.203 0.199 0.182 0.196 0.200 living and sliv2 Water 0.205 0.177 0.167 0.162 0.162 0.172 livelihood sliv3 Electricity 0.207 0.161 0.161 0.131 0.129 0.108 sliv4 Roof-wall 0.178 0.171 0.139 0.124 0.126 0.112 sliv5 Assets 0.152 0.142 0.126 0.104 0.103 0.078 sliv6 Employment 0.127 0.132 0.118 0.119 0.118 0.096 Total std. of living and livelihood 0.554 Note: The incidence for total education refers to the incidence of deprivation in educ1 or educ2 (i.e. any deprivation in terms of education indicators), and similarly for total health and total standard of living and livelihood. The value of health2 for 2013 is not comparable with previous years due to a change in the reference period for this indicator (see text for details). Source: Calculated from APIS survey data sets for rounds 2004 through 2013. 22 TABLE 4: ALTERNATIVE SETS OF WEIGHTS FOR MULTIDIMENSIONAL POVERTY ASSESSMENT FOR THE PHILIPPINES 9 indicators 10 indicators Main Variable Indicator Nested Nested Subjective Nested Nested Subjective Balisacan dimension name Uniform Inverse Welfare Uniform Inverse Welfare (2015) Weights Incidence Weights Weights Incidence Weights weights Weights Weights Education educ1 Primary schooling 0.167 0.212 0.172 0.167 0.220 0.155 0.167 educ2 School attendance 0.167 0.195 0.057 0.167 0.203 0.056 0.167 Total education 0.333 0.407 0.230 0.333 0.423 0.211 Health health1 Food consumption 0.333 0.371 0.189 0.167 0.157 0.178 0.167 health2 Illness 0 0 0 0.167 0.190 0.071 Total health 0.333 0.371 0.189 0.333 0.347 0.249 Standard of sliv1 Sanitation 0.056 0.035 0.114 0.056 0.036 0.105 0.083 living and sliv2 Water 0.056 0.036 0.019 0.056 0.037 0.019 0.083 livelihood sliv3 Electricity 0.056 0.036 0.139 0.056 0.037 0.129 0.083 sliv4 Roof-wall 0.056 0.037 0.112 0.056 0.039 0.105 0.083 sliv5 Assets 0.056 0.038 0.177 0.056 0.040 0.163 0.083 sliv6 Employment 0.056 0.039 0.020 0.056 0.041 0.019 Other sources of income 0.083 Total std. of living and 0.333 0.221 0.581 0.333 0.230 0.541 livelihood All indicators Total sum of weights 1.000 1.000 1.000 1.000 1.000 1.000 1.000 Note: Balisacan (2015) weights are those corresponding to the APIS-based indicators Source: Calculated from APIS 2004, and Balisacan (2015). 23 TABLE 5: MULTIDIMENSIONAL POVERTY ESTIMATES BASED ON 9 INDICATORS Nested Uniform Nested Inverse Subjective Welfare Weights Incidence Weights Weights Year Multidim. Std. Multidim. Std. Multidim. Std. Poverty Err. Poverty Err. Poverty Err. Estimate Estimate Estimate M(k=1/3) 2004 0.140 0.002 0.139 0.002 0.114 0.002 2007 0.152 0.002 0.155 0.002 0.104 0.002 2008 0.122 0.002 0.122 0.002 0.089 0.002 2010 0.135 0.003 0.138 0.003 0.085 0.003 2011 0.135 0.002 0.138 0.002 0.080 0.002 2013 0.121 0.004 0.125 0.004 0.066 0.004 M(union; β=1) 2004 0.179 0.002 0.174 0.002 0.176 0.002 2007 0.186 0.002 0.185 0.002 0.169 0.002 2008 0.158 0.002 0.155 0.002 0.149 0.002 2010 0.164 0.003 0.165 0.003 0.143 0.003 2011 0.165 0.002 0.165 0.002 0.145 0.002 2013 0.150 0.004 0.149 0.004 0.130 0.004 M(union; β=2) 2004 0.081 0.001 0.083 0.001 0.081 0.001 2007 0.084 0.001 0.088 0.001 0.075 0.001 2008 0.068 0.001 0.070 0.001 0.064 0.001 2010 0.073 0.002 0.077 0.002 0.062 0.002 2011 0.071 0.001 0.075 0.001 0.060 0.001 2013 0.061 0.002 0.065 0.002 0.051 0.002 H(k=1/3) : fraction poor in at least one-third of weighted dimensions 2004 0.281 0.004 0.270 0.004 0.207 0.004 2007 0.318 0.005 0.310 0.004 0.190 0.004 2008 0.258 0.004 0.248 0.004 0.165 0.003 2010 0.288 0.006 0.282 0.006 0.155 0.005 2011 0.295 0.004 0.289 0.004 0.149 0.003 2013 0.274 0.008 0.269 0.008 0.126 0.006 H(union) : fraction poor in any dimensions 2004 0.635 0.004 0.635 0.004 0.635 0.004 2007 0.630 0.004 0.630 0.004 0.630 0.004 2008 0.587 0.004 0.587 0.004 0.587 0.004 2010 0.564 0.006 0.564 0.006 0.564 0.006 2011 0.583 0.004 0.583 0.004 0.583 0.004 2013 0.560 0.008 0.560 0.008 0.560 0.008 Note: These estimates omit the health2 indicator. The standard errors allow for design effects due to stratification and clustering. Source: APIS survey data sets for rounds 2004 through 2013. 24 TABLE 6: MULTIDIMENSIONAL POVERTY ESTIMATES BASED ON 10 INDICATORS Nested Uniform Nested Inverse Subjective Welfare Balisacan Weights Incidence Weights Weights (2015) weights Year Multidim. Std. Multidim. Std. Multidim. Std. Multidim. Poverty Err. Poverty Err. Poverty Err. Poverty Estimate Estimate Estimate Estimate M( k=1/3) 2004 0.084 0.002 0.067 0.001 0.107 0.002 0.154 2007 0.082 0.002 0.068 0.002 0.099 0.002 0.140 2008 0.064 0.001 0.051 0.001 0.082 0.002 0.130 2010 0.069 0.002 0.056 0.002 0.080 0.003 0.128 2011 0.062 0.001 0.050 0.001 0.075 0.002 0.124 2013 0.072 0.003 0.063 0.003 0.068 0.004 M(union; β=1) 2004 0.154 0.002 0.143 0.002 0.171 0.002 2007 0.156 0.002 0.146 0.002 0.166 0.002 2008 0.131 0.002 0.122 0.001 0.144 0.002 2010 0.134 0.002 0.126 0.002 0.140 0.003 2011 0.131 0.002 0.122 0.001 0.141 0.002 2013 0.149 0.003 0.144 0.003 0.140 0.004 M(union; β=2) 2004 0.053 0.001 0.049 0.001 0.073 0.001 2007 0.053 0.001 0.049 0.001 0.069 0.001 2008 0.042 0.001 0.039 0.001 0.058 0.001 2010 0.044 0.001 0.041 0.001 0.056 0.002 2011 0.041 0.001 0.038 0.001 0.055 0.001 2013 0.047 0.002 0.046 0.002 0.050 0.002 H(k=1/3) : fraction poor in at least one-third of weighted dimensions 2004 0.186 0.003 0.140 0.003 0.204 0.004 0.319 2007 0.185 0.004 0.146 0.003 0.190 0.004 0.301 2008 0.145 0.003 0.110 0.003 0.160 0.003 0.281 2010 0.157 0.005 0.123 0.004 0.153 0.005 0.277 2011 0.144 0.003 0.110 0.003 0.147 0.003 0.275 2013 0.167 0.007 0.140 0.006 0.138 0.006 H(union) : fraction poor in any dimension 2004 0.668 0.004 0.668 0.004 0.668 0.004 2007 0.668 0.004 0.668 0.004 0.668 0.004 2008 0.615 0.004 0.615 0.004 0.615 0.004 2010 0.599 0.006 0.599 0.006 0.599 0.006 2011 0.612 0.004 0.612 0.004 0.612 0.004 2013 0.663 0.008 0.663 0.008 0.663 0.008 Note: The estimates for 2013 are not comparable due to a change in the reference period for the health2 indicator (see text for details). The standard errors allow for design effects due to stratification and clustering. Source: APIS survey data sets for rounds 2004 through 2013; and Balisacan (2015). 25 TABLE 7: CONTRIBUTION OF DIMENSIONS TO MULTIDIMENSIONAL POVERTY FOR DIFFERENT WEIGHTING SCHEMES AND MEASURES (%AGE) Poverty Dimension/ Indicator measure/ Nested Uniform Nested Inverse Subjective welfare var. name weights Incidence weights weights 2004 2013 2004 2013 2004 2013 M( k=1/3) educ1 Primary schooling 4.7 2.8 4.7 2.7 8.5 8.6 educ2 School attendance 7.0 3.8 7.1 3.8 4.5 3.7 Education 11.7 6.6 11.8 6.5 13.0 12.3 health1 Food consumption 71.9 82.0 78.4 86.6 23.8 25.8 Health 71.9 82.0 78.4 86.6 23.8 25.8 sliv1 Sanitation 3.4 2.7 2.0 1.5 12.0 12.9 sliv2 Water 2.9 2.4 1.7 1.5 2.4 2.9 sliv3 Electricity 3.8 2.2 2.2 1.3 18.2 16.7 sliv4 Roof-wall 2.8 1.9 1.7 1.2 10.6 10.7 sliv5 Assets 2.7 1.6 1.7 1.0 19.6 18.2 sliv6 Employment 0.7 0.6 0.5 0.4 0.4 0.4 Std. of living & livelihood 16.4 11.5 9.9 6.9 63.2 61.9 100.0 100.0 100.0 100.0 100.0 100.0 M(union; β=1) educ1 Primary schooling 5.9 4.4 7.7 5.7 6.2 5.3 educ2 School attendance 12.7 9.1 15.3 10.7 4.5 3.6 Education 18.6 13.5 23.0 16.4 10.7 8.9 health1 Food consumption 47.1 58.1 53.8 64.9 27.2 37.9 Health 47.1 58.1 53.8 64.9 27.2 37.9 sliv1 Sanitation 7.4 7.4 4.7 4.6 15.4 17.6 sliv2 Water 6.4 6.4 4.2 4.1 2.2 2.5 sliv3 Electricity 6.4 4.0 4.3 2.6 16.4 11.5 sliv4 Roof-wall 5.5 4.1 3.8 2.8 11.4 9.6 sliv5 Assets 4.7 2.9 3.3 2.0 15.3 10.5 sliv6 Employment 3.9 3.5 2.9 2.5 1.5 1.5 Std. of living & livelihood 34.3 28.4 23.2 18.7 62.1 53.2 100.0 100.0 100.0 100.0 100.0 100.0 M(union; β=2) educ1 Primary schooling 7.0 5.3 9.1 6.7 8.0 7.2 educ2 School attendance 13.1 9.0 15.6 10.6 3.7 2.6 Education 20.1 14.2 24.7 17.2 11.7 9.9 health1 Food consumption 52.6 63.2 58.8 69.5 25.8 33.7 Health 52.6 63.2 58.8 69.5 25.8 33.7 sliv1 Sanitation 5.7 5.4 3.2 3.0 13.7 15.0 sliv2 Water 4.9 4.9 2.9 2.9 1.6 1.8 sliv3 Electricity 6.1 4.1 3.6 2.4 18.0 14.5 sliv4 Roof-wall 4.7 3.7 2.9 2.2 11.1 10.2 sliv5 Assets 4.3 3.0 2.8 1.9 17.7 14.6 sliv6 Employment 1.7 1.5 1.1 1.0 0.4 0.4 Std. of living & livelihood 27.3 22.6 16.5 13.2 62.5 56.5 100.0 100.0 100.0 100.0 100.0 100.0 Note: See text for details. Source: APIS survey data sets for rounds 2004 and 2013. 26 Poverty & Equity Global Practice Working Papers (Since July 2014) The Poverty & Equity Global Practice Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. This series is co-published with the World Bank Policy Research Working Papers (DECOS). It is part of a larger effort by the World Bank to provide open access to its research and contribute to development policy discussions around the world. For the latest paper, visit our GP’s intranet at http://POVERTY. 1 Estimating poverty in the absence of consumption data: the case of Liberia Dabalen, A. L., Graham, E., Himelein, K., Mungai, R., September 2014 2 Female labor participation in the Arab world: some evidence from panel data in Morocco Barry, A. G., Guennouni, J., Verme, P., September 2015 3 Should income inequality be reduced and who should benefit? redistributive preferences in Europe and Central Asia Cojocaru, A., Diagne, M. F., November 2014 4 Rent imputation for welfare measurement: a review of methodologies and empirical findings Balcazar Salazar, C. F., Ceriani, L., Olivieri, S., Ranzani, M., November 2014 5 Durable goods and poverty measurement Amendola, N., Vecchi, G., November 2014 6 Inequality stagnation in Latin America in the aftermath of the global financial crisis Cord, L., Barriga Cabanillas, O., Lucchetti, L., Rodriguez-Castelan, C., Sousa, L. D., Valderrama, D. December 2014 7 Born with a silver spoon: inequality in educational achievement across the world Balcazar Salazar, C. F., Narayan, A., Tiwari, S., January 2015 8 Long-run effects of democracy on income inequality: evidence from repeated cross-sections Balcazar Salazar,C. F., January 2015 Updated on September 12, 2017 by POV GP KL Team | 1 9 Living on the edge: vulnerability to poverty and public transfers in Mexico Ortiz-Juarez, E., Rodriguez-Castelan, C., De La Fuente, A., January 2015 10 Moldova: a story of upward economic mobility Davalos, M. E., Meyer, M., January 2015 11 Broken gears: the value added of higher education on teachers' academic achievement Balcazar Salazar, C. F., Nopo, H., January 2015 12 Can we measure resilience? a proposed method and evidence from countries in the Sahel Alfani, F., Dabalen, A. L., Fisker, P., Molini, V., January 2015 13 Vulnerability to malnutrition in the West African Sahel Alfani, F., Dabalen, A. L., Fisker, P., Molini, V., January 2015 14 Economic mobility in Europe and Central Asia: exploring patterns and uncovering puzzles Cancho, C., Davalos, M. E., Demarchi, G., Meyer, M., Sanchez Paramo, C., January 2015 15 Managing risk with insurance and savings: experimental evidence for male and female farm managers in the Sahel Delavallade, C., Dizon, F., Hill, R., Petraud, J. P., el., January 2015 16 Gone with the storm: rainfall shocks and household well-being in Guatemala Baez, J. E., Lucchetti, L., Genoni, M. E., Salazar, M., January 2015 17 Competition in Kenyan markets and its impact on income and poverty: a case study on sugar and maize Argent, J., Begazo Gomez,T. P., January 2015 18 Handling the weather: insurance, savings, and credit in West Africa De Nicola, F., February 2015 19 The distributional impact of fiscal policy in South Africa Inchauste Comboni, M. G., Lustig, N., Maboshe, M., Purfield, C., Woolard, I., March 2015 20 Interviewer effects in subjective survey questions: evidence from Timor-Leste Himelein, K., March 2015 21 No condition is permanent: middle class in Nigeria in the last decade Corral Rodas, P. A., Molini, V., Oseni, G. O., March 2015 22 An evaluation of the 2014 subsidy reforms in Morocco and a simulation of further reforms Verme, P., El Massnaoui, K., March 2015 23 The quest for subsidy reforms in Libya Araar, A., Choueiri, N., Verme, P., March 2015 Updated on September 12, 2017 by POV GP KL Team | 2 24 The (non-) effect of violence on education: evidence from the "war on drugs" in Mexico Márquez-Padilla, F., Pérez-Arce, F., Rodriguez Castelan, C., April 2015 25 “Missing girls” in the south Caucasus countries: trends, possible causes, and policy options Das Gupta, M., April 2015 26 Measuring inequality from top to bottom Diaz Bazan, T. V., April 2015 27 Are we confusing poverty with preferences? Van Den Boom, B., Halsema, A., Molini, V., April 2015 28 Socioeconomic impact of the crisis in north Mali on displaced people (Available in French) Etang Ndip, A., Hoogeveen, J. G., Lendorfer, J., June 2015 29 Inequality of outcomes and inequality of opportunity in Tanzania Belghith, N. B. H., Zeufack, A. G., May 2015 30 How unfair is the inequality of wage earnings in Russia? estimates from panel data Tiwari, S., Lara Ibarra, G., Narayan, A., June 2015 31 Fertility transition in Turkey—who is most at risk of deciding against child arrival? Greulich, A., Dasre, A., Inan, C., June 2015 32 The socioeconomic impacts of energy reform in Tunisia: a simulation approach Cuesta Leiva, J. A., El Lahga, A., Lara Ibarra, G., June 2015 33 Energy subsidies reform in Jordan: welfare implications of different scenarios Atamanov, A., Jellema, J. R., Serajuddin, U., June 2015 34 How costly are labor gender gaps? estimates for the Balkans and Turkey Cuberes, D., Teignier, M., June 2015 35 Subjective well-being across the lifespan in Europe and Central Asia Bauer, J. M., Munoz Boudet, A. M., Levin, V., Nie, P., Sousa-Poza, A., July 2015 36 Lower bounds on inequality of opportunity and measurement error Balcazar Salazar, C. F., July 2015 37 A decade of declining earnings inequality in the Russian Federation Posadas, J., Calvo, P. A., Lopez-Calva, L.-F., August 2015 38 Gender gap in pay in the Russian Federation: twenty years later, still a concern Atencio, A., Posadas, J., August 2015 39 Job opportunities along the rural-urban gradation and female labor force participation in India Chatterjee, U., Rama, M. G., Murgai, R., September 2015 40 Multidimensional poverty in Ethiopia: changes in overlapping deprivations Yigezu, Biratu., Ambel, A. A., Mehta, P. A., September 2015 Updated on September 12, 2017 by POV GP KL Team | 3 41 Are public libraries improving quality of education? when the provision of public goods is not enough Rodriguez Lesmes, P. A., Valderrama Gonzalez, D., Trujillo, J. D., September 2015 42 Understanding poverty reduction in Sri Lanka: evidence from 2002 to 2012/13 Inchauste Comboni, M. G., Ceriani, L., Olivieri, S. D., October 2015 43 Exploring the sources of downward bias in measuring inequality of opportunity Lara Ibarra, G., Martinez Cruz, A. L., October 2015 44 Women’s police stations and domestic violence: evidence from Brazil Perova, E., Reynolds, S., November 2015 45 From demographic dividend to demographic burden? regional trends of population aging in Russia Matytsin, M., Moorty, L. M., Richter, K., November 2015 46 Hub-periphery development pattern and inclusive growth: case study of Guangdong province Luo, X., Zhu, N., December 2015 47 Unpacking the MPI: a decomposition approach of changes in multidimensional poverty headcounts Rodriguez Castelan, C., Trujillo, J. D., Pérez Pérez, J. E., Valderrama, D., December 2015 48 The poverty effects of market concentration Rodriguez Castelan, C., December 2015 49 Can a small social pension promote labor force participation? evidence from the Colombia Mayor program Pfutze, T., Rodriguez Castelan, C., December 2015 50 Why so gloomy? perceptions of economic mobility in Europe and Central Asia Davalos, M. E., Cancho, C. A., Sanchez, C., December 2015 51 Tenure security premium in informal housing markets: a spatial hedonic analysis Nakamura, S., December 2015 52 Earnings premiums and penalties for self-employment and informal employees around the world Newhouse, D. L., Mossaad, N., Gindling, T. H., January 2016 53 How equitable is access to finance in turkey? evidence from the latest global FINDEX Yang, Judy., Azevedo, J. P. W. D., Inan, O. K., January 2016 54 What are the impacts of Syrian refugees on host community welfare in Turkey? a subnational poverty analysis Yang, J., Azevedo, J. P. W. D., Inan, O. K., January 2016 55 Declining wages for college-educated workers in Mexico: are younger or older cohorts hurt the most? Lustig, N., Campos-Vazquez, R. M., Lopez-Calva, L.-F., January 2016 Updated on September 12, 2017 by POV GP KL Team | 4 56 Sifting through the Data: labor markets in Haiti through a turbulent decade (2001-2012) Rodella, A.-S., Scot, T., February 2016 57 Drought and retribution: evidence from a large-scale rainfall-indexed insurance program in Mexico Fuchs Tarlovsky, Alan., Wolff, H., February 2016 58 Prices and welfare Verme, P., Araar, A., February 2016 59 Losing the gains of the past: the welfare and distributional impacts of the twin crises in Iraq 2014 Olivieri, S. D., Krishnan, N., February 2016 60 Growth, urbanization, and poverty reduction in India Ravallion, M., Murgai, R., Datt, G., February 2016 61 Why did poverty decline in India? a nonparametric decomposition exercise Murgai, R., Balcazar Salazar, C. F., Narayan, A., Desai, S., March 2016 62 Robustness of shared prosperity estimates: how different methodological choices matter Uematsu, H., Atamanov, A., Dewina, R., Nguyen, M. C., Azevedo, J. P. W. D., Wieser, C., Yoshida, N., March 2016 63 Is random forest a superior methodology for predicting poverty? an empirical assessment Stender, N., Pave Sohnesen, T., March 2016 64 When do gender wage differences emerge? a study of Azerbaijan's labor market Tiongson, E. H. R., Pastore, F., Sattar, S., March 2016 65 Second-stage sampling for conflict areas: methods and implications Eckman, S., Murray, S., Himelein, K., Bauer, J., March 2016 66 Measuring poverty in Latin America and the Caribbean: methodological considerations when estimating an empirical regional poverty line Gasparini, L. C., April 2016 67 Looking back on two decades of poverty and well-being in India Murgai, R., Narayan, A., April 2016 68 Is living in African cities expensive? Yamanaka, M., Dikhanov, Y. M., Rissanen, M. O., Harati, R., Nakamura, S., Lall, S. V., Hamadeh, N., Vigil Oliver, W., April 2016 69 Ageing and family solidarity in Europe: patterns and driving factors of intergenerational support Albertini, M., Sinha, N., May 2016 70 Crime and persistent punishment: a long-run perspective on the links between violence and chronic poverty in Mexico Rodriguez Castelan, C., Martinez-Cruz, A. L., Lucchetti, L. R., Valderrama Gonzalez, D., Castaneda Aguilar, R. A., Garriga, S., June 2016 Updated on September 12, 2017 by POV GP KL Team | 5 71 Should I stay or should I go? internal migration and household welfare in Ghana Molini, V., Pavelesku, D., Ranzani, M., July 2016 72 Subsidy reforms in the Middle East and North Africa Region: a review Verme, P., July 2016 73 A comparative analysis of subsidy reforms in the Middle East and North Africa Region Verme, P., Araar, A., July 2016 74 All that glitters is not gold: polarization amid poverty reduction in Ghana Clementi, F., Molini, V., Schettino, F., July 2016 75 Vulnerability to Poverty in rural Malawi Mccarthy, N., Brubaker, J., De La Fuente, A., July 2016 76 The distributional impact of taxes and transfers in Poland Goraus Tanska, K. M., Inchauste Comboni, M. G., August 2016 77 Estimating poverty rates in target populations: an assessment of the simple poverty scorecard and alternative approaches Vinha, K., Rebolledo Dellepiane, M. A., Skoufias, E., Diamond, A., Gill, M., Xu, Y., August 2016 78 Synergies in child nutrition: interactions of food security, health and environment, and child care Skoufias, E., August 2016 79 Understanding the dynamics of labor income inequality in Latin America Rodriguez Castelan, C., Lustig, N., Valderrama, D., Lopez-Calva, L.-F., August 2016 80 Mobility and pathways to the middle class in Nepal Tiwari, S., Balcazar Salazar, C. F., Shidiq, A. R., September 2016 81 Constructing robust poverty trends in the Islamic Republic of Iran: 2008-14 Salehi Isfahani, D., Atamanov, A., Mostafavi, M.-H., Vishwanath, T., September 2016 82 Who are the poor in the developing world? Newhouse, D. L., Uematsu, H., Doan, D. T. T., Nguyen, M. C., Azevedo, J. P. W. D., Castaneda Aguilar, R. A., October 2016 83 New estimates of extreme poverty for children Newhouse, D. L., Suarez Becerra, P., Evans, M. C., October 2016 84 Shedding light: understanding energy efficiency and electricity reliability Carranza, E., Meeks, R., November 2016 85 Heterogeneous returns to income diversification: evidence from Nigeria Siwatu, G. O., Corral Rodas, P. A., Bertoni, E., Molini, V., November 2016 86 How liberal is Nepal's liberal grade promotion policy? Sharma, D., November 2016 Updated on September 12, 2017 by POV GP KL Team | 6 87 CPI bias and its implications for poverty reduction in Africa Dabalen, A. L., Gaddis, I., Nguyen, N. T. V., December 2016 88 Building an ex ante simulation model for estimating the capacity impact, benefit incidence, and cost effectiveness of child care subsidies: an application using provider-level data from Turkey Aran, M. A., Munoz Boudet, A., Aktakke, N., December 2016 89 Vulnerability to drought and food price shocks: evidence from Ethiopia Porter, C., Hill, R., December 2016 90 Job quality and poverty in Latin America Rodriguez Castelan, C., Mann, C. R., Brummund, P., December 2016 91 With a little help: shocks, agricultural income, and welfare in Uganda Mejia-Mantilla, C., Hill, R., January 2017 92 The impact of fiscal policy on inequality and poverty in Chile Martinez Aguilar, S. N., Fuchs Tarlovsky, A., Ortiz-Juarez, E., Del Carmen Hasbun, G. E., January 2017 93 Conditionality as targeting? participation and distributional effects of conditional cash transfers Rodriguez Castelan, C., January 2017 94 How is the slowdown affecting households in Latin America and the Caribbean? Reyes, G. J., Calvo-Gonzalez, O., Sousa, L. D. C., Castaneda Aguilar, R. A., Farfan Bertran, M. G., January 2017 95 The impact of mining on spatial inequality recent evidence from Africa Boly, A., Mveyange, A., Addison, T., February 2017 96 Are tobacco taxes really regressive? evidence from Chile Fuchs Tarlovsky, A., Meneses, F. J., March 2017 97 Design of a multi-stage stratified sample for poverty and welfare monitoring with multiple objectives: a Bangladesh case study Yanez Pagans, M., Roy, D., Yoshida, N., Ahmed, F., March 2017 98 For India's rural poor, growing towns matter more than growing cities Murgai, R., Ravallion, M., Datt, G., Gibson, J., March 2017 99 Leaving, staying, or coming back? migration decisions during the northern Mali conflict Hoogeveen, J. G., Sansone, D., Rossi, M., March 2017 100 Arithmetics and Politics of Domestic Resource Mobilization Bolch, K. B., Ceriani, L., Lopez-Calva, L.-F., April 2017 101 Can Public Works Programs Reduce Youth Crime? Evidence from Papua New Guinea’s Urban Youth Employment Project Oleksiy I., Darian N., David N., Sonya S., April 2017 102 Is Poverty in Africa Mostly Chronic or Transient? Evidence from Synthetic Panel Data Dang, H.-A. H., Dabalen, A. L., April 2017 Updated on September 12, 2017 by POV GP KL Team | 7 103 To Sew or Not to Sew? Assessing the Welfare Effects of the Garment Industry in Cambodia Mejía-Mantilla, C., Woldemichael, M. T., May 2017 104 Perceptions of distributive justice in Latin America during a period of falling inequality Reyes, G. J., Gasparini, L. C., May 2017 105 How do women fare in rural non-farm economy? Fuje, H. N., May 2017 106 Rural Non-Farm Employment and Household Welfare: Evidence from Malawi Adjognon, G. S., Liverpool-Tasie, S. L., De La Fuente, A., Benfica, R. M., May 2017 107 Multidimensional Poverty in the Philippines, 2004-13: Do Choices for Weighting, Identification and Aggregation Matter? Datt, G., June 2017 108 But … what is the poverty rate today? testing poverty nowcasting methods in Latin America and the Caribbean Caruso, G. D., Lucchetti, L. R., Malasquez, E., Scot, T., Castaneda, R. A., June 2017 109 Estimating the Welfare Costs of Reforming the Iraq Public Distribution System: A Mixed Demand Approach Krishnan, N., Olivieri, S., Ramadan, R., June 2017 110 Beyond Income Poverty: Nonmonetary Dimensions of Poverty in Uganda Etang Ndip, A., Tsimpo, C., June 2017 111 Education and Health Services in Uganda: Quality of Inputs, User Satisfaction, and Community Welfare Levels Tsimpo Nkengne, C., Etang Ndip, A., Wodon, Q. T., June 2017 112 Rental Regulation and Its Consequences on Measures of Well-Being in the Arab Republic of Egypt Lara Ibarra, G., Mendiratta, V., Vishwanath, T., July 2017 113 The Poverty Implications of Alternative Tax Reforms: Results from a Numerical Application to Pakistan Feltenstein, A., Mejia-Mantilla, C., Newhouse, D. L., Sedrakyan, G., August 2017 114 Tracing Back the Weather Origins of Human Welfare: Evidence from Mozambique? Baez Ramirez, J. E., Caruso, G. D., Niu, C., August 2017 115 Many Faces of Deprivation: A multidimensional approach to poverty in Armenia Martirosova, D., Inan, O. K., Meyer, M., Sinha, N., August 2017 116 Natural Disaster Damage Indices Based on Remotely Sensed Data: An Application to Indonesia Skoufias, E., Strobl, E., Tveit, T. B., September 2017 For the latest and sortable directory, available on the Poverty & Equity GP intranet site. http://POVERTY WWW.WORLDBANK.ORG/POVERTY Updated on September 12, 2017 by POV GP KL Team | 8