Volume 45 Number 178,
July-September 2014
The Multidimensional Poverty Index and Poverty Traps in the Southern Cone
María Emma Santos *
Date received: August 2, 2013. Date accepted: January 6, 2014

This work proposes using the Multidimensional Poverty Index (MPI) as an indicator to quantify poverty traps in countries, as it allows us to measure the impact and intensity of poverty for those lacking in various basic areas simultaneously. The MPI is interesting for this purpose, both because it serves to complement measures of poverty by income over time, and because it may be a more viable tool for periodic use in the absence of panel data. This work examines the MPI for five Southern Cone countries.

Keywords: Southern Cone, poverty trap, multidimensional poverty index, measurement methodologies, social problems.

Poverty is frequently described as a vicious cycle or poverty trap because there is a self-reinforcing mechanism that causes poverty to persist (Azariadis et al., 2005: 326). From a formal perspective, a poverty trap refers to a situation in which there are at least two equilibria – one good and one bad – and the system works intrinsically to ensure that the bad equilibrium is perpetuated and the good one cannot be achieved given the current conditions (London and Rojas, 2013). This multiple equilibria situation is defined by one or more critical thresholds (usually wealth cutoffs) that are difficult for people to overcome from below (Barret and Peak, 2006).

The concept of poverty traps has long existed in development theory, with fundamental works by Young (1928), Rosestein-Rodan (1943), Nurske (1952), Myrdal (1957) and Leibenstein (1957). These ideas were revisited decades later and formalized in a variety of macroeconomic (economic growth models) and microeconomic models (focused on individual agents). Both types of models sought to explain different accumulation paths, some of which have inherent self-reinforcing poverty trap mechanisms, including: economies of scale, positive externalities, the presence of complements, imperfect competition, the failure of some markets (principally capital markets) and the recognition of the importance of an institutional framework to regulate economic transactions.1

From an empirical perspective, because poverty traps are important in designing policies, they are relevant when measuring poverty. However, the cross-sectional income poverty measurements that are typically used (such as the proportion of people living on less than 1.25 USD/day) do not quantify poverty traps, because they merely reflect insufficient income in a certain moment in time. They do not indicate if the problem is persistent, or whether the person is also deprived in other non-monetary ways.

To come up with a notion of a trap, we need to measure income poverty over time. The literature on measuring poverty over time is growing, and these works discriminate between chronic and temporary poverty. They estimate these measures using panel data to follow the same household or person over a time period and identify those that remain below the poverty line in multiple time periods.

However, panel data is scarce and, besides a few specific programs, tends to cover rather brief time periods, which makes it difficult to estimate how many people are in poverty traps. For this reason, this work proposes using the multidimensional poverty index introduced by Alkire et al. (2010, 2014) and published in the Human Development Report since 2010 (PNUD, 2010, 2011, 2013) as one alternative to measure poverty traps.

The MPI measures acute poverty, defined by two features. First, it refers to people living in conditions that do not meet the internationally defined standards for basic functions, such as good nutrition and access to drinking water.2 Second, it describes people that do not meet the minimum standards for various aspects simultaneously, that is, people that experience multiple deprivations.

According to Amartya Sen’s approach, these different sets of functions that people can achieve constitute capabilities. Many of these capabilities are connected in such a way that developing certain capabilities facilitates and promotes developing others, meaning they are mutually complementary. Moreover, in some cases, they are interrelated in such a way that certain thresholds must be achieved for some capabilities in order to obtain others. For example, many studies have found that adequate nutrition is a necessary precursor to intellectual development (Glewwe et al., 2001) and productive employment (Dasgupta et al., 1986). In addition, household conditions (such as access to drinking water and improved sanitation) also impact achieving minimum health (WHO and UNICEF, 2000), which in turn affects other areas.

Therefore, not being able to achieve certain functions can prevent people from achieving others, causing a poverty trap, whereby people are deprived of multiple basic needs at the same time. From this point of view, the MPI can be understood as an alternative way to measure poverty traps, because it is less demanding in terms of data than income poverty measurements over time by allowing for periodic estimates. Specifically, this work illustrates how to use the MPI for five countries in the Southern Cone: Argentina, Bolivia, Brazil, Paraguay and Uruguay.

The first section briefly describes the structure and approach to using the MPI, while the second section introduces the data sources used. The third section analyzes the results of the MPI for the five countries in the region and compares them to income poverty estimates. Finally, this study presents some conclusions.


The MPI (Alkire et al., 2010, 2014; PNUD, 2010) is an internationally comparable poverty metric designed to measure acute poverty, understood, as described in the introduction, as not meeting two internationally established minimum standards of basic functions, simultaneously. The mathematical structure for the MPI is M00, a member of one of the families of multidimensional poverty metrics proposed by Alkire et al. (2011), which can be applied when at least one of the variables is on an ordinal scale.3

This study considered three dimensions: education, health and living standards, and ten indicators (see Table 1): years of schooling of household members, children attending school, nutrition, household child mortality, access to electricity, drinking water, adequate sanitation, clean cooking fuel, type of flooring and number of small assets (durable consumer goods).

These dimensions and indicators were selected based on the capabilities approach, and therefore, the analysis will focus on dimensions that are absolutely necessary for human development. The capabilities approach focuses not on the means, but rather the ends, of development. Income is considered as merely the means to development, and not an end in and of itself. In keeping with this approach, the indicators for the MPI were chosen to prioritize those that best reflect how people function. However, the data set had some limitations, and as such, not all indicators are functions in and of themselves. Specifically, the living standards dimension used indicators on access to services. However, even in these cases, it could be said that these resources are extremely closely connected to the needs they satisfy, and in that sense, are different from income, which is more varied.

Building the MPI required that indicator data came from the same source. It was necessary to determine whether or not each household met the thresholds of each indicator.4 Because there are more and more household surveys, however, this requirement was not too difficult.

The following steps are used to construct the MPI, also necessary for the M0 metric. The first evaluates each person by the achievements of the household in which he or she lives to determine if there is deprivation in any indicator. Let the achievement of each person in each indicator and let zj be the cutoff for indicator j, where a person is deprived of such indicator if he or she falls below the cutoff. Formally, deprivation is defined as when and in another case.5 It is notable that the MPI assumes positive and negative externalities within the household, in such a way that if one member is malnourished, all household members are considered to be malnourished in the nutrition indicator.6 Table 1 details the types of deprivation, many of which are taken from millennium development goals (MDG) (UN, 2000).

Secondly, each person’s privation is defined by the weight of the indicator, given by such that . Each dimension receives the same relative weight (1/3), evenly distributed among the indicators of each dimension. Table 1 details how each indicator is weighted and, based on that, how a deprivation score is calculated for each person, defined as the weighted sum of the deprivations: .

Later on, this score can be used to identify poor people with a second deprivation cutoff, called which specifies the proportion of minimum deprivations that a person must have to be considered poor, that is, someone is poor when Specifically, in the MPI, people are identified as poor wen the sum of the weighted deprivations is k=33.33%. Deprivations of those not identified as poor are not taken into account in the MPI, so technically, they are censored. Formally, the censored deprivations are defined as when and in the other case. Analogously, the censored deprivation score is built as .

Once the multidimensionally poor have been identified, the index combines two fundamental pieces of information: the “proportion” of multidimensionally poor people (known as poverty incidence) and the "intensity" of their deprivations, given by the average proportion of deprivations (weighted) of the poor. More formally, the proportion of poor people is given by , where q is the number of people identified as poor. The intensity is given by . The MPI, like the M0, is the product of these two metrics:

Because the MPI contains the additive structure of M0, it admits two types of decompositions, which are extremely useful in analyzing poverty traps. First, this index can be decomposed by population groups, because, when there are two or more mutually exclusive groups that sum up to the total population, the weighted sum of poverty in each sub-group is equal to the aggregate poverty metric, where the weights correspond to the population share of each group. In turn, once the poor have been identified (such that the deprivations of the non-poor have been censored), the MPI can be broken down into so-called recount rates censored by indicator, which indicate the proportion of people that are poor and experience deprivation in each indicator j. These decompositions allow us to determine the contributions of different population groups and the contribution of the deprivation of each indicator to total aggregate poverty.

As the introduction argued, in this work, the MPI can be understood as a way to measure poverty traps, for two reasons. First, because it estimates deprivation in basic human functions: nutrition, mortality, basic education and basic living services. A poverty index that includes fewer basic deprivations would be less reliable as a poverty trap metric. Second, the MPI takes into account deprived people for at least 33.33% of the weighted indicators. The simultaneous nature of deprivations in these basic functions means it is reasonable to assume that the people defined as poor by the MPI are likely in a poverty trap and, without outside help, it is unlikely that they will overcome this situation.


This work introduces and analyzes the MPI for five Southern Cone countries: Argentina, Bolivia, Brazil, Paraguay and Uruguay. Unfortunately, Chile was not included, because it does not have a database containing health information indicators required for the MPI. The databases used were as follows. For Argentina, the National Nutrition and Health Survey, conducted by the Health Ministry between 2004 and 2005, was used. This survey was not designed to obtain nationally representative estimates, both because it only encompasses urban areas and because the target population included households with children and women in reproductive age. For Brazil, Paraguay and Uruguay, the World Health Survey, WHS) conducted in 2003 was used. This is a standardized survey designed by the World Health Organization. Finally, for Bolivia, the Demographic and Health Survey, DHS) was used from 2003, designed and supervised by USAID. Table 2 presents the sample sizes for each survey, and how much of the sample could be used (because it had complete information) in calculating the multidimensional poverty index.

It is important to clarify that these results should be seen as merely indicative, in terms of between-country comparisons, given the limitations of available data sources. First, in Argentina, the sample is from 2004-2005, not 2003, and is not nationally representative. Second, MPI estimates for Brazil, Paraguay and Uruguay do not consider child school attendance, because the WHS does not collect data on this variable, meaning that the years of schooling indicator took the entire weight of the education dimension (16.67%). In addition, for Brazil, there was no mortality indicator, and as such, nutrition carried the entire weight of the health dimension.7

Despite the limitations of the data, the exercise is interesting as a first approximation to quantifying poverty traps in Southern Cone countries, and as an example of the multidimensional poverty index approach.


Incidence and Intensity of Poverty Traps on the Aggregate Level

This section analyzes the outcome of MPI estimates for five Southern Cone countries and interprets these results in light of the concept of a poverty trap. Figure 1 introduces the MPI and its components and the incidence and intensity of poverty for the five countries studied. The countries were ordered from highest to lowest MPI, starting with Bolivia, at 0.175, followed by Brazil, at 0.083, Paraguay, 0.064, Argentina, 0.011 and Uruguay, 0.006. Incidence estimates indicate that in Bolivia, 36.3% of the population is poor, according to the MPI, while this figure is 21.6% in Brazil and 13.3% in Paraguay. Argentina and Uruguay were considerably lower, at 2.9 and 1.7%, respectively.

Given that the MPI identifies people as poor who live in households with deprivations in at least 33.33% of the indicators (weighted), these rates of incidence are often a mere approximation of the proportion of people really trapped in poverty. These are people experiencing at least one health deprivation and one education deprivation, or both deprivations in one of these dimensions, or one deprivation in one of these dimensions and three in the living standards dimension. Any of these minimum combinations suggests that poverty will be reinforced over time. For example, a household with one malnourished member and children not attending school (or no member with five years of schooling) will find restrictions in trying to overcome the thresholds of any of these categories on its own.

When incidence rates translate into numbers of people, the estimates take on new importance. According to the population values given in 2007 (UN, 2010), there were approximately 41 million people in poverty traps in Brazil, 3.4 million in Bolivia, 1.1 million in Argentina, 0.8 million in Paraguay and 0.05 million Uruguay. Logically, in keeping with country size, Brazil has the greatest number of people in poverty traps in the Southern Cone.

The advantage of the MPI, versus the rate of incidence, is that it also includes the intensity of poverty among the poor, which provides additional and valuable information, especially in terms of quantifying poverty traps. It is noteworthy that Paraguay has an intensity comparable with that of Bolivia; on average, the poor were deprived in 48% of the weighted indicators, whereas the intensity in Argentina was comparable to that of Brazil, where, on average, the poor were deprived in 38% of the weighted indicators. As a result, even when incidence rates are substantially different, the poverty intensity may be similar among countries, in such a way that poverty traps are deeper in countries with greater poverty intensity.8 Uruguay had the lowest poverty intensity: the average poor person was deprived in 35% of the indicators (barely above the cutoff line of 33.33%).

Figure 1. The MPI and its Components in Five Southern Cone Countries

Source: Estimates from Alkire et al. (2010).

This study also analyzed the distribution of intensity among the poor, to gain an idea of the magnitude and sectors of the population with varying degrees of poverty acuity, and, therefore, degree of entrapment. Figure 2 introduces the proportion of multidimensionally poor people with between 33 and 39% of weighted deprivations, as well as those between 40 and 49%, 50 and 59%, 60 and 69% and over 70%.

This figure clearly shows similarities in the distribution of intensity between Argentina and Brazil and Bolivia and Paraguay. In the former two, the largest proportion of the poor had between 33 and 39% deprivations. However, 10% were between 40 and 49%, 6% between 50 and 59% and the rest with 60% or more of weighted deprivations. In Brazil, a more populous country, 20% of the population with 40% or more deprivations accounted for nearly 9 million people, constituting the poorest of the poor. In Bolivia and Paraguay, there was a greater proportion of poor people with higher poverty intensity than in Argentina and Brazil. Essentially, 63% of the poor in Bolivia had 40% or more deprivation of the weighted indicators, while this figure was 57% in Paraguay. This reveals that among those in poverty traps, a significant fraction is stuck in a vicious cycle that is particularly hard to overcome. In Uruguay, where the MPI is very low, virtually all of the poor people (97%) had intensity between 33 and 39%.

Figure 2. Distribution of Poverty Intensity Among Multidimensionally Poor People in Five Southern Cone Countries

Source: Estimates from Alkire et al. (2010).

Deprivations in Multiple Dimensions

As mentioned in the second section, one of the advantages of the MPI structure is that it allows for the analysis of the structure of deprivations the poor experience. This information is relevant when analyzing poverty traps because it helps determine which deprivations tend to occur together. It should be noted that although there must be deprivation in at least 33.33% of weighted indicators to be considered poor by the MPI, these indicators may all derive from the same dimension. It is therefore interesting to evaluate whether the poor in the Southern Cone, as defined by the MPI, tend to have deprivations in multiple dimensions, or occurring together.

Figure 3 displays the proportion of poor people with deprivations in the various combinations of dimensions for each country, whether that be health, education or living standards, some mix of two or three all at the same time. Figure 4 shows a radar chart of the so-called recount rates censored by each indicator. This describes the proportion of the population that is poor and has a deprivation in each of the indicators. It is nearly irrelevant in Uruguay, where the MPI is very low, such that analyzing its composition is of little significance.

Figure 3. Poverty Composition by Dimension in Five Southern Cone Countries

Source: Prepared by the author based on Alkire et al. (2010).

Figure 4. Poverty Composition by Indicator in Five Southern Cone Countries

Source: Estimates from Alkire et al. (2010). The diagrams reflect the censored recount rate defined
as the percentage of the population that is poor and deprived of each indicator.

For Argentina, Figure 3 shoes that the majority of the multidimensionally poor (71%) are deprived in education and living standards, where 10% have deprivations in health and education, 5.6% in health and living standards and 12% have deprivations in all three dimensions simultaneously. In other words, 99% of the poor, as defined by the MPI, in Argentina have deprivations in two or more dimensions. Figure 4 shows the breakdown by indicator, where it is clear that the multidimensionally poor tend to be especially deprived in asset ownership, flooring, cooking fuel source and sanitation services and the indicator for household members with less than five years of schooling. Fortunately, the indicator for children attending school presents low deprivation, which suggests favorable intergenerational evolution over time, where children coming from households with low education may manage to reverse this path.9 However, deprivations in living standards combined with low schooling for adult members of the household make it unlikely that these households will be able to improve their living conditions on their own.

In Bolivia, similar to Argentina, over 96% of the multidimensionally poor have deprivations in two or more dimensions – 42% in all three at the same time, followed by 37% in the education and living standards combination and 17% with the health and living standards combination. In terms of indicators, as shown in Figure 4, in the living standards dimension, precarious sanitation facilities were the indicator with the greatest relative weight, followed by using unclean cooking fuel, flooring and lack of access to electricity. In terms of education, Bolivia's pattern was the inverse of Argentina; the children attending school indicator saw greater deprivation than the indicator for household members with less than five years of schooling. In other words, in this situation, there may be an inverse intergenerational dynamic, obviously headed on a downward trend, by which children obtain less education than their parents. In addition, nearly 20% of the population is poor and lives in households where at least one child has perished. There is a possibility that deprivations composed of residential features are associated to this mortality, which would also show how poverty traps persist among the multidimensionally poor.

In Brazil, there were only eight of the ten indicators, limiting the analysis. Figure 3 reveals that there was a greater proportion of the poor with deprivation in only one dimension, 48.5% in education and nearly 14% in health. However, 32% had combined deprivations in education and living standards. Among the indicators considered, the absence of household members with at least five years of schooling stands out: 18% of the population is poor and lives in a household where no members have completed five years of schooling. Among living standards, the indicators with the greatest deprivation rates were sanitation services and cooking fuel.

Finally, in Paraguay, as can be seen in Figure 3, 87.4% of the poor had deprivations in two or three dimensions, and the most frequent combination was – as in Argentina and Brazil – education and living standards (47% of the poor), followed by health and living standards (30%). In terms of the indicators, Figure 4 reveals that there was relatively high deprivation among all living standards indicators, except electricity, which are frequently combined with deprivation in household member years of schooling.

In summary, in three of the five countries analyzed, the majority of the poor, as defined by the MPI, had deprivations in at least two dimensions, where the most frequent combination was education and living standards, followed by health and living standards. This suggests that in the absence of outside help, this population will face severe difficulties in trying to overcome poverty.

Poverty Traps in Rural and Urban Areas

Aggregate level poverty estimates tend to conceal important inequalities among population groups. Specifically, differences between urban and rural areas are significant. The data source for Argentina only included urban areas, and therefore cannot be included in this analysis. Of the four remaining countries, Uruguay and Brazil are highly urbanized. In Uruguay, 92% of people in the WHS sample live in urban zones, while this same figure is 83% in Brazil. By contrast, Bolivia and Paraguay are less urbanized: 62% of the DHS sample for Bolivia lived in urban areas, while the number was 55% in the WHS Paraguay sample.

Figure 5 presents the values of the MPI and its components for rural and urban zones in each of the four countries. In all cases, the rural MPI was greater than the urban, but the gap between the two varied. In Bolivia and Paraguay, the urban-rural discrepancy was more noticeable: the rural MPI was 5.6 times greater than the urban. This divergence is fundamentally determined by different incidence rates, which are 4.8 times higher than urban areas in Bolivia, and 5.2 times higher in Paraguay. In Brazil, the rural MPI is three times that of the urban, with incidence 2.5 times higher. Uruguay had the lowest discrepancy: the rural MPI was only 1.76 times higher than the urban. In Bolivia, Brazil and Paraguay, rural zones also had greater poverty intensity, but the difference was much less accentuated than for incidence, at only 1.1 and 1.2 times higher. In Uruguay, poverty intensity is similar among rural and urban areas.

Figure 5. The MPI, Incidence and Intensity in Rural and Urban Zones of Four Southern Cone Countries

Source: Alkire et al. (2010).

This outcome reflects the fact that the probability of falling into a poverty trap is substantially higher for people living in rural zones than for those in urban areas. In absolute values, Bolivia is the most notable case: seven out of ten people living in rural zones are poor, defined by the MPI. In Brazil, this number is four out of ten, and, in Paraguay, nearly 2.5 out of ten. In summary, geographic location would appear to constitute a significant factor conditioning the formation of poverty traps.

Quantifying the Traps: Multidimensional Poverty vs. Income Poverty

A natural question would be to compare the MPI with income poverty metrics, and determine why income poverty metrics are less effective in quantifying poverty traps. The discrepancy among population groups captured by income poverty measures and the MPI can be verified, initially, by observing aggregate statistics. Figure 6 introduces estimates of poverty incidence according to the MPI as compared to those of income poverty, using two international poverty lines defined by the World Bank, 1.25 USD/day and 2 USD/day, as well as each country's national poverty line. As can be expected, national poverty lines are more demanding, as they have been adapted to the specific context of each country and therefore provide estimates of income poverty that are higher than those made using internationally defined poverty lines. Comparing the MPI incidence rate with the income poverty rate in Argentina and Paraguay, multidimensional poverty is significantly lower than income poverty. In Uruguay, MPI incidence is greater than that of 1.25 USD/day, but less than that of 2 USD/day. By contrast, in Bolivia and Brazil, the proportion of poor according to the MPI is greater than that of people living on less than 1.25 USD/day, and similar to the proportion of those living on less than 2 USD/day.

Figure 6. Incidence of MPI Poverty Compared with Income Poverty Incidence (Various Poverty Lines)

Source: Prepared by the author. MPI incidence estimates are from Alkire et al. (2010). Estimates for 1.25 USD/day and 2 USD/day poverty come from the World Development Indicators and were taken from the year closest to the MPI estimate for which there was available information. The years were 2005 for Argentina, 2002 for Bolivia and Paraguay and 2003 for Brazil and Uruguay. Incidence values using the national poverty line were taken from World Development Indicators in 2003 for Bolivia, Brazil and Paraguay, and for Argentina and Uruguay, from the SEDLAC database (http://sedlac.econo.unlp.edu.ar/eng/statistics-detalle.php?idE=34) and come from the first half of 2005 for Argentina and 2003 for Uruguay.

These aggregate indicators do not allow for greater deductions, because the degree of overlap among different groups of poor people is unknown. The estimates simply suggest that, assuming that the MPI captures the group of people in a poverty trap, this figure may not necessarily match the number of people identified as income poor, even when using rather undemanding poverty lines, such as internationally defined lines. A lack of income does not have a one-to-one relationship with specific deprivations, nor with multiple deprivations.

To analyze the degree of overlap among multidimensional poverty groups in greater depth, we must have information on the ten MPI indicators and income information from the same data source. The World Health Survey used to calculate the MPI for Brazil, Paraguay and Uruguay includes a brief income module, which helps identify those that are income poor and those that are poor as defined by the MPI, as well as those that are income poor but not MPI poor, MPI poor but not income poor, and not poor at all. To estimate income poverty, the 1.25 USD/day line was used, adjusted for purchasing power parity in 2002 (World Bank, 2004). Figure 7 presents the estimates for Brazil and Paraguay, because the MPI in Uruguay is too low to analyze in this way. The estimates are shown in a Venn diagram.

Figure 7. MPI Poverty Index Compared to the 1.25 USD/Day Poverty Index on the Disaggregate Level,
Brazil and Paraguay in 2003

Source: Prepared by the author based on Alkire et al. (2010) estimates.

According to the sample, in Brazil, 9.1% of people live on less than 1.25 USD/day, whereas 21.6% are MPI poor and 4.1% are poor according to both criteria. This would suggest that if we assume that the MPI accounts for people in a poverty trap, using a proxy metric, like income poverty, to identify this segment of the population, has a 55% chance of inclusion error (55=5/9), because 55% of the income poor group are not MPI poor. In addition, there is a 19% exclusion error (19=17.5/90.9), because 19% of the people that are not income poor are indeed MPI poor. In Paraguay, the overlap between these two types of poverty is even less, and the potential exclusion error rate runs up to 70% (70=6.7/9.6), with an inclusion error rate of 11.5% (11.5=10.4/90.4).


This work proposed using the multidimensional poverty index (MPI) (Alkire et al., 2010, 2014; PNUD, 2010) to approximate and quantify poverty traps in various countries. This index is a way to complement income estimates of chronic poverty, which require panel data. When panel data is not available, the MPI may be a viable alternative to periodically estimate the presence of poverty traps. The arguments in favor are twofold: first, the MPI is made up of indicators related to basic human functions. Second, the MPI identifies those with multiple deprivations among these indicators. This work therefore explored the potential of the MPI as a poverty trap metric in five Southern Cone countries: Argentina, Bolivia, Brazil, Paraguay and Uruguay.

It found that significant portions of the Bolivian and Brazilian populations are in poverty traps; more than one-third in the former country and one out of five in the latter. Logically, because Brazil has much higher population density, it is there where the most people trapped in poverty in the Southern Cone are located. Paraguay, with a 13% incidence, is lower, but still significant. Argentina and Uruguay had much lower estimates and small numbers in absolute terms. Although Bolivia and Brazil had a greater proportion of people in poverty traps, poverty intensity was greater in Bolivia and Paraguay, where the average poor person is deprived of 48% of the weighted indicators. This situation can be interpreted as a deeper trap than what exists in Brazil, where, on average, the poor are deprived of 38% of the weighted indicators.

When analyzing multiple dimension deprivation, that is, deprivation among indicators from different dimensions, in Bolivia, and even Argentina and Paraguay, where the average incidence and intensity is lower, 87% or more of the poor have deprivations in two or three dimensions. This number is only 37% in Brazil. In all countries, the most frequent combination of deprivations is in education and living standards. The fact that deprivations exist in more than one dimension is precisely what suggests that these households will unlikely be able to overcome poverty on their own. In addition, this study found that the residents of rural areas had a much greater probability of becoming trapped in poverty than those in urban zones, an alarming fact for Bolivia.

However, the way in which those identified as MPI poor and the income poor overlap (using the 1.25 USD/day poverty line) in Brazil and Paraguay indicates that each metric is identifying different groups, and that the income poor in one time period do not necessarily constitute the group of people in a poverty trap. However, better databases will be needed for further in-depth analysis.

More exhaustive analysis would also be required for policy recommendations, in conjunction with cost-benefit evaluations. However, preliminary analysis suggests a few lines of actions to explore, such as comprehensive plans to distribute basic services (sanitation, drinking water, electricity, natural gas), together with educational measures for adults in certain countries and, in Bolivia, measures to universalize access to primary school and prevent mortality. The fact that these human capabilities seem to complement each other means that isolated interventions will be ineffective in helping households to overcome poverty.

The MPI has been designed to measure acute poverty. Although the definition of acute poverty responds to the features of a poverty trap, using the MPI to evaluate poverty traps deserves further evaluation. For example, it would be useful to compare the MPI with income poverty metrics over time with panel data and some information on MPI indicators, and it would also help to evaluate how the households defined as poor by the MPI evolve over time.


The author would like to thank ANPCyT-PICT 1888 CONICET-PIP 11220110100363 for financing this research.


Alkire, Sabina, and James Eric Foster (2011), “Counting and multidimensional poverty measurement”, Journal of Public Economics, vol. 95, no. 7-8, Elsevier, August, pp. 476-487.

Alkire, Sabina, and María Emma Santos (2010), “Acute multidimensional poverty: A new index for developing countries”, OPHI working paper, no. 38. http://www.ophi.org.uk/wp-content/uploads/ophi-wp38.pdf. ______ (2014), “Measuring acute poverty in the developing world: Robustness and scope of the multidimensional poverty index”, World Development, vol. 59, Elsevier, July, pp. 251-274.

Azariadis, Costas, and John Stachurski (2005), “Poverty traps”, in Philippe Aghion, and Steven Durlauf (eds.), Handbook of economic growth, vol. 1, Part A, Chapter 5, Amsterdam, Elsevier, pp. 295-384.

Banco Mundial (World Bank), World development indicators, http://data.worldbank.org/data-catalog/world-development-indicators (consulted March 2012).

______ (2004) 2004, World development indicators, pp. 278-280, available at: http://siteresources.worldbank.org/ICPINT/Resources/Table5_7.pdf (consultado el 10/03/10).

Barret, Christopher B., and John G. Peak (2006), “Poverty traps and safety nets”, in Alain de Janvry, and Ravi Kanbur (eds.), Poverty, inequality and development, essays in honor to Erik Thorbecke, Chapter 8, New York, pp. 131- 154.

Dasgupta, Partha, and Debraj Ray (1986), “Inequality as a determinant of malnutrition and unemployment: Theory”, Economic Journal, vol. 96, no. 384, Blackwell Publishing, March, pp. 1011-1034.

Demographic Health Survey (DHS) http://www.measuredhs.com/start.cfm. (consulted between July 2009 and July 2010).

Encuesta Nacional de Nutrición y Salud (2005), Ministerio de Salud, Argentina.

Glewwe, Paul; Hanan Jacoby, and Elizabeth King (2001), “Early childhood nutrition and academic achievement: A longitudinal analysis”, Journal of Public Economics, vol. 81, no. 3, Elsevier, September, pp. 345-368.

Leibenstein, Harvey (1957), Economic backwardness and economic growth, New York, Wiley and Sons.

London, Silvia, and Mara Rojas (2013), Tópicos de trampas de pobreza, Ediuns, Bahía Blanca, Argentina.

Myrdal, Gunnar (1957), Economic theory and underdeveloped regions, Gerald Duckworth, London.

Naciones Unidas (2000), United Nations millennium declaration. http://www. un.org/millennium/declaration/ares552e.pdf

______ (2010), Department of Economic and Social Affairs, Population Division, World population prospects: The 2010 revision, http://esa.un.org/ wpp/Excel-Data/population.htm (consulted October 2011).

Nurske, Ragnar (1952), “Some international aspects of the problem of economic development”, American Economic Review, vol. 42, no. 2, American Economic Association, May, pp. 571-583.

PNUD (2010), Informe sobre desarrollo humano 2010: la verdadera riqueza de las naciones: caminos al desarrollo humano.

______ (2011), Informe sobre desarrollo humano 2011: sostenibilidad y equidad: un mejor futuro para todos.

______ (2013), Informe sobre desarrollo humano 2013: el ascenso del sur, progreso humano en un mundo diverso.

Rosenstein-Rodan, Paul (1943), “Problems of industrialisation of Eastern and South-eastern Europe”, Economic Journal, vol. 53, no. 210/211, Blackwell Publishing, June–September, pp. 202-211.

Santos, María Emma; María Ana Lugo; Luis Felipe López-Calva; Guillermo Cruces, and Diego Battiston (2010), “Refining the basic needs approach: A multidimensional analysis of poverty in Latin America”, Studies in applied welfare analysis: Papers from the Third ecineq Meeting, Bingley, Emerald, pp. 1-29.

Sen, Amartya (1992), Inequality re-examined, Oxford, Clarendon Press.

______ (2009), The idea of justice, Allen Lane, London.

Socio economic database for Latin America and the Caribbean, http://sed-lac.econo.unlp.edu.ar/eng/statistics-detalle.php?idE=34 (consulted 20/04/12).

World Health Organization (WHO) and United Nations Children’s Fund (UNICEF) (2000), Global water supply and sanitation assessment 2000 report.

World Health Survey (WHS) http://www.who.int/healthinfo/survey/en/index. html (consulted between July 2009 and July 2010).

Young, Allyn (1928), “Increasing returns and economic progress”, Economic Journal, vol. 28, no. 152, Royal Economic Society, December, pp. 527- 542.

* Universidad Nacional del Sur, Argentina. maria.santos@qeh.ox.ac.uk.

1 See Azariadis and Starchuski (2005) for a review of poverty trap models.

2 The term function is part of the Sen capabilities approach vocabulary (2009, most recent version) and refers to the different “beings and doings" that a person can value or have reason to value.

3 The other metrics in this family are all cardinal indicators.

4 The MPI cannot be constructed with macrodata.

5 The assumption is that achievements in the various dimensions can be represented with real, non-negative numbers.

6 The reason for working in this way is that there is no individual information for all of the indicators, especially for health indicators.

7 It should be clarified that the national households surveys conducted in Latin America generally do not collect information on health indicators such as nutrition and mortality, which is why they cannot be used in calculating the MPI. In Argentina, Paraguay and Uruguay, the sources of data used were the only ones available that contained MPI health indicators. For Bolivia and Brazil, there was a more recent data source available in each case (DHS 2008 for Bolivia and PNDS 2006 for Brazil), but the choice was made to use 2003 data to be comparable with the other countries.

8 Strictly speaking, the comparability of intensities is limited, because, as clarified in section 2, because there are missing indicators, Brazil, Paraguay and Uruguay have a relatively higher weight for the schooling indicator and nutrition had a greater weight for Brazil.

9 This is consistent with evidence from other data sources (Santos et al., 2010).

Licencia de Creative Commons  Problemas del Desarrollo. Revista Latinoamericana de Economía by Instituto de Investigaciones Económicas, Universidad Nacional Autónoma de México
is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License

Published in Mexico, 2012-2018 © D.R. Universidad Nacional Autónoma de México (UNAM).
PROBLEMAS DEL DESARROLLO. REVISTA LATINOAMERICANA DE ECONOMÍA, Volume 49, Number 194 July-September 2018 is a quarterly publication by the Universidad Nacional Autónoma de México, Ciudad Universitaria, Coyoacán, CP 04510, México, D.F. by Instituto de Investigaciones Económicas, Circuito Mario de la Cueva, Ciudad Universitaria, Coyoacán,
CP 04510, México, D.F. Tel (52 55) 56 23 01 05 and (52 55) 56 24 23 39, fax (52 55) 56 23 00 97, www.probdes.iiec.unam.mx, revprode@unam.mx. Journal Editor: Moritz Cruz. Reservation of rights to exclusive use of the title: 04-2012-070613560300-203, ISSN: pending. Person responsible for the latest update of this issue: Minerva García, Circuito Maestro Mario de la Cueva s/n, Ciudad Universitaria, Coyoacán, CP 04510, México D.F., latest update: August 29th, 2018.
The opinions expressed by authors do not necessarily reflect those of the editor of the publication.
Permission to reproduce all or part of the published texts is granted provided the source is cited in full including the web address.
Credits | Contact

The online journal Problemas del Desarrollo. Revista Latinoamericana de Economía corresponds to the printed edition of the same title with ISSN 0301-7036