برآورد فقر جهانی: تجزیه و تحلیل حساسیت

Current estimates of global poverty vary substantially across studies. We undertake a sensitivity analysis to highlight the importance of methodological choices by measuring global poverty using different data sources, parametric and nonparametric estimation methods, and multiple poverty lines. Our results indicate that estimates of global poverty vary significantly when they are based alternately on data from household surveys versus national accounts but are relatively consistent across estimation methods. The decline in poverty over the past decade is found to be robust across methodological choices.

Global poverty monitoring has been brought to the forefront of the international policy arena with the adoption of the Millennium Development Goals (MDGs) by the United Nations. The first of the eight MDGs proposes reducing global poverty by the year 2015 and is stated as “halving the proportion of people with an income level below $1/day during 1990–2015” (United Nations, 2000). Progress toward attaining this MDG is monitored using global poverty estimates published by the World Bank and a number of independent scholars. The process is not only expensive (Moss, 2010) but also mired with conceptual, methodological, and data-related problems (Klasen, 2009). Estimates of global poverty differ significantly in magnitude as well as in the rate of change in poverty. Consider, for instance, Chen and Ravallion (2010) and Pinkovskiy and Sala-i-Martin (2009)—two studies that estimate global poverty using the international poverty line of $1/day (see Figure 1). Chen and Ravallion (2010) find that during 1981–2005 the global poverty rate fell from 51.8% to 25.2%. Pinkovskiy and Sala-i-Martin (2009) show that the global poverty rate over the same period declined from 4.4% to 2.4%. Although there is general agreement in the literature that global poverty has decreased over time, the estimated level of poverty and the rate of poverty decline vary substantially across studies. Full-size image (18 K) Figure 1. Estimates of global poverty during 1981–2005. Notes: The poverty rates are based on 2005 PPPs but are not strictly comparable across studies because of differences in methodological approach (see Section 2(c)). Figure options This paper aims to contribute to the debate on global poverty not by providing a new set of estimates but by addressing two important questions. First, we ask why estimates from different studies differ so much. As we unravel the various assumptions made by researchers, we realize that global poverty estimates are simply not comparable across studies. For instance, they differ in terms of underlying data sources, number of countries included, welfare metric, adjustments to mean incomes, and statistical methods employed to estimate the income distribution. Given this variety of methodological choices, we arrive at our second question: Can we assess the impact of different approaches on the resulting poverty estimates? Since global poverty estimation requires making multiple assumptions simultaneously, we isolate and measure separately the relative importance of each such assumption by undertaking a sensitivity analysis. The paper documents how sensitive global poverty estimates are to underlying assumptions. A key methodological choice made when estimating global poverty is whether to use data from household surveys (HS) or national account statistics (NAS) or whether to combine data from the two sources. Data on income are typically collected through HS of nationally representative samples. In fact HS data are the only source of information on the relative distribution of incomes, that is, on the share of income or consumption possessed by population quintiles or deciles in a country. Distributional data can be scaled by mean income or consumption from HS or from NAS. HS means are the most natural choice; the World Bank estimates global poverty largely based on data from HS (Chen and Ravallion, 2001, Chen and Ravallion, 2004 and Chen and Ravallion, 2010). But HS-based estimates of mean income or consumption are not available for all countries and years, primarily because they are not published by statistical agencies. To overcome this limitation, researchers often replace HS means with estimates from NAS (Ahluwalia et al., 1979, Bhalla, 2002, Pinkovskiy and Sala-i-Martin, 2009 and Sala-i-Martin, 2006). NAS data are a more readily-accessible and consistently-recorded source of information on average incomes, and are available for most countries on a yearly basis. However, rescaling distributional data from HS with means from NAS has been criticized because it implicitly assumes that the discrepancy between the two data sources is distribution-neutral (Bourguignon, 2005, Chen and Ravallion, 2010 and Deaton, 2005). The choice between HS and NAS data generates significantly different poverty estimates, as shown in Figure 1 and documented in the paper. However, these differences cannot be entirely attributed to this particular choice, since studies also differ in other aspects, such as the number of countries covered, the choice of welfare metric (income or consumption), and the statistical techniques used to estimate the income distribution. Given the variety of methodological approaches proposed in the literature on global poverty, our main goal is to assess the relative significance of these choices. We do so by means of a sensitivity analysis. We start with a benchmark poverty estimate based entirely on HS data and quantify the extent to which poverty estimates vary when the relative distribution is anchored alternately to mean income or consumption from NAS. This is our first sensitivity exercise. The second sensitivity exercise concerns the choice of statistical method used to estimate income distributions from grouped data representing income shares of population deciles. We measure global poverty by estimating each country’s distribution using different methods employed in the literature. These include the General Quadratic (GQ) Lorenz curve, the Beta Lorenz curve, the lognormal density function, and the Singh–Maddala density function.1 In addition to these parametric specifications, we consider the nonparametric kernel density method whose performance we assess in conjunction with four different bandwidths—a parameter that controls the smoothness of the income distribution. As a benchmark, we follow the World Bank methodology to the extent possible and estimate global poverty in 1995 and 2005—the latest year for which data are available for many countries. We collect distributional data for 65 countries from the World Bank’s poverty monitoring website PovcalNet. Our sample covers almost 90% of the developing world population and includes all countries for which HS and NAS data are available in both years. Global poverty is estimated using poverty lines ranging from $1/day to $2.5/day. By using a range of poverty lines, we are able to determine how methodological choices impact poverty rates at different income cutoffs. Furthermore, we take into account the fact that the $1/day international poverty line is relatively low compared to the average standard of living in developing countries. Our results are twofold. First, a large share of the variation in estimated poverty levels and trends can be attributed to the choice between HS and NAS means. Global poverty estimates vary not only in terms of the proportion of the poor, and correspondingly the number of poor, but also in terms of the rates of poverty reduction. Poverty estimates based on HS and NAS do not tend to converge in higher income countries. Second, the choice of statistical method used to estimate the income distribution affects poverty levels to a lesser extent. A comparison of poverty estimates across parametric and nonparametric techniques reveals that the commonly-used lognormal specification consistently underestimates poverty levels. While there is little doubt that the proportion of poor declined during 1995–2005, our results underscore the fact that global poverty counts are highly sensitive to the methodological approach. The remainder of the paper is structured as follows. Section 2 consists of a review of the literature on global poverty. We describe the sensitivity analysis and the data in Section 3, and the results in Section 4. Conclusions are presented in Section 5. The statistical techniques used in the exercise are described in the Appendix.

Over the past decades, global poverty monitoring has gained significance in international policy-making, and increasingly so with the adoption of the Millennium Development Goals. However, measuring global poverty has proven to be a difficult exercise, both conceptually and empirically. Estimates of global poverty vary substantially due to different methodological choices made by researchers and are rarely comparable across studies. In this paper we present a sensitivity analysis of global poverty estimates by proposing a step-by-step approach to assessing the relative importance of different assumptions. If countries had complete information on every individual’s income then with an agreed-upon global poverty line, identifying the poor would be a straightforward exercise. However, the task is rendered complex by data-driven limitations. The debate over the use of survey data and national accounts data continues to dominate the literature on global poverty measurement. For instance, the World Bank estimates global poverty relying largely on survey data. However, countries conduct surveys at irregular intervals, which means that poverty estimates based on survey data still have to use data from the national accounts to interpolate missing survey values. Household surveys are also known to suffer from the problems of nonresponse and under-reporting. Another approach to estimating global poverty is to use national accounts estimates of mean income (or consumption) to scale survey-based relative distributions. This method rests on the assumptions that national accounts data are a reliable estimate of mean income and that the discrepancy between the two data sources is distribution-neutral. We believe these are strong assumptions and that conceptually it is difficult to defend replacing the survey mean with the national accounts mean to anchor relative distributions from surveys. Nonetheless, given the paucity of consistent survey data across countries and over time, some combination of survey and national accounts data is often needed to estimate poverty at the global level. The purpose of this paper is first to assess the sensitivity of global poverty estimates to the choice between household survey and national accounts estimates of mean income/consumption. Although nationally representative surveys are, in our view, a better and more direct source of information on private consumption, we believe that neither of these estimates is unbiased, but both are plausible. Our sensitivity analysis reveals that global poverty estimates vary not only in terms of the proportion of the poor, and correspondingly the number of poor, but also in terms of the rates of decline in poverty. Poverty estimates based on surveys are higher than those based on national accounts and do not tend to converge in countries with higher income. Our second sensitivity exercise concerns the choice of statistical technique used to estimate income distributions. The results indicate that global poverty rates vary little across statistical methods. Among the techniques discussed in the paper, we prefer parametric methods since these are more reliable than nonparametric methods when applied to a small number of observations, such as income shares by population deciles. Although the decline in the global poverty rate during 1995–2005 appears robust across methodological approaches, the estimated number of poor and the rate of poverty reduction differ significantly. The purpose of the paper is not to provide evidence supporting the choice of one particular methodological approach over others. Instead, our sensitivity analysis underscores the consequences of this choice by quantifying their impact on poverty estimates, thus alerting the reader to the importance of this issue. We believe that examining the robustness of global poverty estimates to alternate methodological choices helps put these estimates in perspective and boosts public confidence in global poverty studies.