Carlos Felipe Balcázar, Eduardo A. Malásquez, Sergio Olivieri and Julieth Pico GLOBAL PRACTICES PROFILE Knowledge Brief The multidimensional poverty index (MPI) is a useful literature on how to update the indicators in the MPI measure to assess deprivations in a society beyond once these are no longer capturing significant traditional money-metric approaches. During the deprivations. last decade, the MPI has been used to complement traditional monetary poverty measures in several Latin American countries such as Chile, Costa Rica, Colombia, Dominican Republic, Ecuador, El Salvador, Honduras, Mexico and Panama (World In Colombia, the MPI was adopted as an instrument Bank, 2017). for monitoring public policy following the Plan Moreover, nowadays several countries are Nacional de Desarrollo 2010-2013 “Prosperidad para implementing the MPI not only for monitoring todos.” The selection of indicators included in the multidimensional poverty over time but also for official MPI relied on data gathered by the Life public policy design, as well as for program Quality Survey (Encuesta de Calidad de Vida, ECV), monitoring and evaluation; targeting the poorest while the implementation of the MPI follows the groups and other beneficiary groups or geographic methodology proposed by Alkire and Foster (2007). zones. For instance, Colombia uses the MPI to Figure 1: Dimensions of the MPI in Colombia define regions for the allocation of the conditional cash transfer program Mas Familias en Acción, and the program Unidos (a household is graduated from the program if and only if it is not extreme poor by income nor multidimensional poor). This note discusses the evolution of the MPI in Colombia since 2010 and describes some of the challenges associated with the spatial heterogeneity of multidimensional poverty across urban and rural areas, and the relationship between life cycle and the evolution of the MPI over time. Also, this note opens a discussion that has not been yet addressed by the Colombia’s index is comprised of five dimensions, expansion of the subsidized regime, particularly after each with multiple indicators (Figure 1). Under the 2004. Although we observe progress on informal current approach, the index is a weighted sum of employment, the rate of people working informally is dichotomized attainments (or indicators). Then, each still high (as high as 73.6% in 2016). dimension is given the same weight and, following a The only indicator that regressed was the long-term nested weighting scheme, within a dimension each unemployment, linked to the weak performance in indicator is given the same weight. economic activity faced by the country in recent From 2010 to 2016, Colombia achieved a significant years. There was an increase on the percentage of reduction in its official MPI (see Figure 2). The official people experiencing long-term unemployment. MPI not only fell from 30.4 percent in 2011 to 17.8 BOX: The MPI in Latin America percent in 2016, but Colombia experienced an 1. In Mexico, the MPI considers two dimensions: improvement in 14 of the 15 indicators that social rights and economic welfare. A person is composed the MPI. considered multidimensionally poor if her income is below the poverty line and she is Figure 2: Changes in Multidimensional Poverty deprived in any of the 6 indicators used to Rate in Colombia by indicator, 2010-2012-2016 construct the social right dimension. 2. In Panama, the MPI is composed of 5 2 Progress 2010 - 2016 dimensions with equal weight. A person will be Progress 2012-2016 considered as multidimensional poor if she 0 lives in a household deprived in more than 30% Percentage points of the 17 weighted indicators. -2 3. The MPI in Honduras uses 15 indicators, grouped in 4 dimensions. The dimensions on -4 health, education, and labor have 3 indicators, while the dimension on housing has 6 -6 indicators. A person is multidimensionally poor if she lives in a household deprived in more -8 than 25% of the 15 weighted indicators. 4. The MPI in Ecuador uses 12 indicators grouped -10 in 4 dimensions (education; labor and social security; health, water and nutrition; and, -12 housing and environment). A person is No Access to Healthcare Services… considered as multidimensional poor if she is Child Labour Illiteracy Long-Term Unemployment Low Educational Achievement No Adequate External Walls Low School Attendance Critical Overcrowding No Adequate Floors No Access to Drinking Water Informal Employment Educational Gap No Health Insurance Poor Access to Childcare Services No Acess to Sanitation deprived in at least 33.33% of the 12 weighted indicators. 5. The MPI in Chile is composed of 5 dimensions each one with 3 indicators. A person is considered as multidimensional poor if she is deprived in at least 25% of the 12 weighted indicators. 6. Costa Rica defined 5 dimensions for its MPI. A person will be considered as multidimensional poor if she is deprived in more than 20% of the weighted indicators. The main drivers of the reduction in the MPI were: 7. The MPI in El Salvador uses 20 indicators increased access to the health system, reduction of grouped in 5 dimensions (each dimension has the educational gap and increase on educational a total of 4 indicators). A household will be considered as multidimensional poor if she is attainment. Regarding the health dimension, an deprived in more than 7 of the 20 indicators additional 11.4% of the population got access to the (35% of the weighted indicators). health system in the last 7 years mainly due to the Source: Own elaboration. more than thrice higher (16.3 percent) in rural areas. This reveals a significant spatial disparity between areas. Similarly, even though child labor and school attendance are below 3 percent at the national level, these indicators are around 6 percent in rural areas. Five out of 15 indicators in the Colombia´s MPI no longer measure significant levels of deprivations at The spatial heterogeneity also occurs among other the national level. These indicators are: no adequate indicators. An important indicator such as lack of floors (4.5%), no access to health services when access to drinking water is below 10% at the national needed (4%), school non-attendance (2.9%), child level, but it is four times higher (37%) among the labor (2.6%) and lack of adequate walls (1.9%). Thus, rural population. Likewise, in the case of lack of the Colombian government is revisiting and access to sanitation, the indicator at the national level updating the list of indicators, cut-off points and is 9.6 percent, while among rural areas is more than weights of its official MPI. 21 percent. These cases suggest the importance for policymakers to consider spatial heterogeneity when Figure 3: Urban, Rural and National MPI in using the MPI to evaluate the evolution of welfare, as Colombia (2016) well as to update the MPI. 100 Urban area Rural area 80 Nation Percentage points 60 MPI indicators typically apply to different units of analysis in the household such as household 40 members of certain age, under the assumption that in absence of such unit of analysis, the household is non-deprived on that indicator. For instance, the 20 child nutrition indicator in a household with no children will reflect non-deprivation. 0 No Access to Healthacare Services… Figure 4 suggests that indicators follow a life cycle Poor Access to Childcare Services No Heatlh Insurance Educational Gap Child Labour Informal Employment No Acess to Sanitation Illiterarcy Low School Attendance No Adequate Floors Critical Overcrowding Long-Term Unemployment No Adequate External Walls No Access to Drinking Water Low Educational Achievement model. Levels of deprivation in child-work are higher when household heads are older (Panel a), and levels of deprivation in early childhood indicators are higher for those household heads in their most fertile ages (Panel b). In contrast, educational attainment indicators reflect intergenerational changes: younger individuals are more educated than older ones (Panel c). They also reflect that household heads accumulate assets during their life cycle (panels e and f). The MPI However, an indicator might not seem informative at does not allow the researcher and policymaker to the national level, but spatial disparities could be observe and disentangle these dynamics and address hidden (Figure 3). For instance, while the indicator the problems that censoring indicators to population for lack of adequate floors at the national level is 4.5 groups entail. percent (and just 1.3 percent in urban areas), it is This challenge could be illustrated by comparing households with and without children and youth. For including lack of education, is to adjust the criteria of example, in Colombia, the second dimension of the deprivation by the age of the individual and consider Colombian MPI is Youth and Children, whose weight all household members, instead of constraining the is 20 percent. This implies that the maximum population to which the requirements will be applied, percentage of weighted deprivation that could that is the case of the lack of education indicator. experience a household without children and youth (around 23 percent of the Colombian households) is 80 percent instead of 100 percent. Figure 4: Smoothed deprivation rates This note discusses the evolution of the MPI in Colombia since 2010 and highlights two challenges that arise when analyzing its evolution over time: the spatial heterogeneity of multidimensional poverty (e.g., across urban and rural areas), and the relationship between life cycle and evolution of the MPI over time. Although, over time an indicator might turn less relevant at the national level, spatial disparities might be hidden in the national aggregate. Thus, public policy design and monitoring need be informed by spatial disaggregation rather than driven only by national averages. In addition, a change in the MPI over time could be explained by demographic changes. For instance, households deprived today in the indicator “gap of education”, since they have a 17-year child with only 9 years of education, will not be deprived anymore in When a criterium as age is used for defining who is this dimension next year (i.e., when child will be 18 deprived or not in each indicator, this threshold years old) even when such child hasn’t accumulated becomes a trigger. For example, in Colombia, the more years of education. child and youth dimensions are applied to people These two features need to be considered before between 0 and 17 years. Then, a reduction in the MPI creating or modifying the MPI as a monitoring could be associated with a demographic change indicator, and before creating or modifying the even if the reduction is not related to social progress. country’s public policies. For instance, in case a demographic transition would have occurred in Colombia between 2017 and 2018, ABOUT THE AUTHORS Felipe Balcazar is a PhD student in the Wilf Family we would have observed a fall in the MPI attributable Department of Politics, New York University. to a change in the age profile. Therefore, we could Eduardo A. Malásquez is an Economist in the Poverty conclude that demographic changes could partially and Equity Global Practice at the World Bank. explain some of the reduction on the MPI. Sergio Olivieri is a Senior Economist in the Poverty and Equity Global Practice at the World Bank. A potential solution, useful for several indicators Julieth Pico is a Consultant for the Poverty and Equity Global Practice at the World Bank.