Abstract
This article aims at contributing to the current literature on poverty data limitations and measurement by discussing the process for producing the first multidimensional poverty measure based on the consensual approach for the City of Buenos Aires. The results show a remarkable level of consensus about the necessities of life in the twenty-first century, underline the importance of generating more suitable indicators of deprivation and show that unmet basic needs-type variables are no longer adequate for measuring poverty in countries like Argentina. According to the valid and reliable poverty index, 20.3% of the city’s population live in households in multidimensionally poor households, this being the social dimension that shows the highest deprivation rate.
Introduction
Multidimensional poverty (MDP) measurement research has a long track record in Latin America (LA) (Boltvinik, 2013). During the 1980s, the unmet basic needs (UBN) approach has underpinned most indices in the region. Using census data, estimations were produced based on indicators that proxied lacking a limited group of basic needs (particularly access to water, sanitation and conditions of the dwelling and, in some cases, access to primary education).
More recently, since the 2000s, various countries began to produce MD poverty measures based on other conceptual frameworks; they considered indicators of deprivation on wider sets of dimensions, using data from household surveys. In (almost) all cases the measures resorted to existing data, that is, indicators built from the variables regularly included in censuses and surveys.
A poverty index is, however, as good as the indicators it uses to identify the diverse kinds of social and material deprivations. Specifically, employing ready-to-use data that was not explicitly collected to measure poverty is a feature whose impact was not sufficiently discussed. For many years the connection between basic needs, poverty and existing social statistics was supposedly clear. Altimir (1979) noted that this was not necessarily the case and pointed out that suitable data was a necessary condition for the valid measurement of poverty.
There is increasing evidence that the available data is not completely appropriate to capture MDP in the region and around the world. Nájera and Gordon (2019) showed that a regional measure that includes many UBN-type indicators leads to unreliable and invalid scores for six of the biggest LA countries and they suggest that part of the reason is that the existent data is no longer useful to measure poverty.1 Similarly, Beccaria and Fernandez (2021) show the degree of measurement error of non-official poverty indices for Argentina and reflect on the need for better data. Similarly, Nájera (2020) finds that the measurement error of the official Mexican measure has increased in recent years due to the lack of suitability of some UBN-type indicators. Villatoro and Santos (2019), drawing upon Altimir’s (1979) reflection, underline the need of more adequate data for LA.
These studies agree, not only on the limitations of the available data, but also on a potential solution. They suggest that the use of the consensual approach (CA), also referred as the consensual deprivation method, pioneered by Mack and Lansley (1985), drawing upon Townsend’s (1979) theory of relative deprivation, would help to produce more suitable deprivation indicators. Similarly, Lanau et al (2020) propose using the CA as a way to improve poverty data in both developing and developed countries by enhancing the measurement of access to basic services. This potential has been explored in some empirical studies. Guillen (2017) and Nájera (2016) showed that the reliability of the Mexican index increases when the socially perceived standards derived from the CA, implemented by CONEVAL in 2007, are used to set the thresholds for several variables of the index (CONEVAL, 2007). However, these exercises focused on the definition of thresholds for already existent indicators. The aim of this article is to contribute to the current literature on poverty data limitations and measurement by discussing the recent experience carried out in the City of Buenos Aires (CBsAs) where the local statistical office recently began to produce its official measure of MD poverty designed for the city2 based on the CA. It represents another example of how this approach that uses deprivation indicators elaborated from data specifically collected – one of the possible methods for generating robust poverty measures – could be implemented and included in the regular set of social indicators of a statistical office.
The article is organised as follows. The second section provides a brief review of LA experiences in MDP measurement. The third section provides the conceptual basis of the CA. The fourth section illustrates how this approach was linked with the existent UBN perspective to collect and produce poverty data. The final section concludes and discusses the findings.
A brief review of Latin American experiences with multidimensional poverty measures
LA pioneered the production of official MDP measures as, probably, the first of such efforts was carried out by the statistical office of Argentina, with the support of ECLAC, in 1984, using the UBN approach (INDEC, 1984) with the 1980 population census data. This approach was based on work done by the Planning Office of Chile to estimate a ‘poverty map’ employing 1970 census information (ODEPLAN/Universidad Católica, 1975).
The measure considered a set of variables capturing the degree of fulfilment/deprivation of certain ‘basic needs’. The number of indicators considered was low – five – as the selection was restricted by the data collected in the census. Among these, only those comparable and relevant for different regions of the country, were considered. A validity analysis was also performed to keep only those indicators with a given degree of association to income based on household survey data. The components were the following: i) type of household; ii) overcrowding; iii) sanitation; and iv) education of children. However, a fifth component was considered: one that was intended to reflect a household’s income, called the ‘income capacity’. This was subject to criticism, as the consideration of income does not clearly fit into the UBN method and, in general, the MDP measure approach.
The definitions of the thresholds for each indicator denoting deprivation and of the aggregation or combination criteria were related. The idea was to use ‘low’ thresholds, but a ‘demanding’ aggregation condition: the union approach (that is, a household is poor if deprived in at least one indicator).
Despite some methodological weaknesses and the lack of a clear conceptual basis, a measure that employed census data was viewed as a valuable tool for assessing poverty at a very high level of geographical disaggregation estimated with existing information, that is, poverty maps.3
The same approach was employed after a few years by several LA countries to produce UBN measures using data from the 1990 round of the population censuses. Some of them generated information with data from the subsequent censuses. Some of the indicators were adapted, especially with items that referred to housing and services. Changes in indicators were also introduced for the estimates with the subsequent censuses as their questionnaire requested information on more and/or different variables. Despite these initial efforts to produce official poverty measures, LA countries shifted their approach and focused on the production of monetary measures. Following the experience of using the UBN method with census data, no effort was made for almost 20 years in Latin America to employ household surveys to generate official MDP measures considering deprivation in other dimensions or other indicators referring to aspects of these dimensions.4
During the 2000s, some countries began to design and regularly produce MDP measures with a broader perspective than the UBN and using household survey data. Mainly, regular multipurpose surveys were employed and these indices used the variables usually collected by these surveys. Only in some cases were specific variables for the poverty measure included in the survey’s questionnaire.
Specifically, nine countries in the region now produce official multidimensional estimates on a regular basis, as shown in Table 1. Mexico led this process, although its indicator is, in fact, based on the combined method.5 Those national measures rested on different conceptual approaches. Mexico explicitly based its measure on the idea of poverty as deprivation of social rights.6 The indicators of the other countries, were influenced by the Oxford Poverty & Human Development Initiative (OPHI) which supported the work done in many of them. Even if the framework of ‘capabilities’ is mentioned as the conceptual basis of these most recent measures, the selection of dimensions and indicators appears to rest on more general criteria: certain ‘common sense’, normative criteria, views of the population, and/or what has been done in other similar exercises.7 OPHI also influenced the Ecuador measure, although the dimensions, as in México, were intended to reflect rights.8,9
Characteristics of the official MDM in Latin America
Chile | Colombia | Costa Rica | Ecuador | El Salvador | Honduras | Mexico | ECLAC | ECLAC-UNICEF (child poverty) | |||
---|---|---|---|---|---|---|---|---|---|---|---|
MDM | Income poor | Poor | |||||||||
Dimensions | 5 | 4 | 5 | 4 | 5 | 4 | 6 | 5 | 6 | ||
Indicators | 15 | 15 | 20 | 12 | 20 | 15 | 12 | 13 | 16 | ||
Weights | |||||||||||
Dimensions | Equal except one | Equal | Equal | Equal | Equal | Equal | Equal except one | Equal | |||
Indicators in dimensions | Equal | Equal | Equal except in one | Equal | Equal | Equal | Equal except in two | ||||
Cut off to identify poor units (% of weights of indicators) | 33.5 | 33 | 20 | 33 | 35 | 25 | One dimension. One indicator per dimension | Intersection | 22 | Extreme poverty: extreme deprivation in one dimension/Child poverty: moderate deprivation in one dimension | |
Unit of analysis | Household | Household | Household | Household | Household | Household | Individual | Household | Individual | Household | Individual (child) |
Source: Prepared by authors based on national official publications and publications from international agencies.
Dimensions are similar in the different measures; they go beyond the restricted set of the employed in the UBN measures and include some of those which are traditionally employed in multidimensional welfare analysis: education, health, food security and housing. Cut-offs for each indicator are defined according to expert opinion and, in the case of Mexico at least, considering normative criteria established by law.
Regarding weights, in general, each of the dimensions (and each of the indicators within a given dimension) are considered equally important. A unit (household or person) is identified as poor if deprived in some weighted proportion of the indicators. For Mexico, a person is denied one right (dimension) if he/she is deprived in one of the indicators of the dimension.10,11 Poverty results when a person is deprived of at least one right and has an income below the national welfare poverty line. The unit of analysis is the household in all cases, except for Mexico, where the individual is the relevant unit.
Apart from national official measures, it is worth mentioning two one-time exercises carried out by international agencies. ECLAC produced a multidimensional measure for 17 countries, also using the national household surveys, which is explicitly based on OPHI’s methodology (ECLAC, 2014). UNICEF produced, with ECLAC, a comparable measure for LA countries based on the idea of poverty as denial of rights. A child is poor if he/she is deprived in at least one dimension (CEPAL-UNICEF, 2010). In all these measures the head count ratio, an intensity measure and the adjusted head count ratio, based on the Alkire-Foster methodology, are calculated and disseminated. The exception is the integrated method of Mexico (only prevalence).
Therefore, regular official multidimensional measures have spread out in the region during the last ten years, enriching the analysis of MDP. Despite this clear advance, the more recent measures share with UBN one of its limitations as they also used indicators built from the variables included in the regular household surveys, whose questionnaires are not specifically designed to assess poverty. Mexico is the only exception, where a special module is introduced in the (bi-annual) Income and Expenditure Survey to capture information. Furthermore, even considering this restriction, the conceptual bases considered are not always sufficient to support the selection criteria of dimensions and indicators.
Relative deprivation and the consensual approach
A number of authors have argued that the measurement of poverty in Latin America demands better deprivation data. This raises questions about both the theoretical and the methodological options to frame a procedure to collect and produce new and sensible information on social and material deprivation. In this section, we discuss the theory of relative deprivation and the CA as a potential alternative to successfully tackle this data production challenge.
individuals, families and groups in the population can be said to be in poverty when they lack the resources to obtain the types of diet, participate in the activities and have the living conditions and amenities which are customary, or are at least widely encouraged or approved, in the societies to which they belong. (1979: 31)
Hence, Townsend argued that the relevant space to define the deprivations through which poverty is evaluated is society. Deprivations are relative in two senses: what is regarded as necessary in one society might not be fully representative of the set of needs of a different society and, second, what was regarded as necessary in the mid-twentieth century might not be the same as what is regarded as necessary in the twenty-first century.13
As indicated by Gordon and Pantazis (1997: 15), Townsend’s relative deprivation is scientific, which means that it is logically internally consistent and fulfils several minimum criteria. This, of course, does not mean that Townsend’s theory is correct or incorrect, it only tells that it can be used under the standard scientific method.
The implementation of Townsend’s theory has led to a series of methodological changes and technical and statistical enhancements over the last 40 years (Gordon and Pantazis, 1997; Pantazis et al, 2006; Dermott and Main, 2017). In his pioneering work Townsend (1979) put forward a list of the necessities of life that he regarded as customary or approved in the UK in the 1960s, that is, representative of the widely agreed needs of the population. Nonetheless, Townsend’s list was arguably unsatisfactory in that it represented what he regarded were the implicit standards of UK society, but he did not explicitly measure the socially perceived needs of the UK population. A second limitation was that Townsend was unable to distinguish between enforced lack and other reasons explaining lacking an item. Therefore, two key statements of his theory remained untested and was subject to several critiques (Townsend, 1981; Piachaud, 1987; Callan and Whelan, 1993).
In poverty measurement, value judgements permeate measurement and, even though Townsend’s theory focuses on society’s living standards, the challenge is both to define and include a relevant set of needs to capture poverty, and to be able to distinguish deprivation due to enforced lack from other possible causes that might not be connected to poverty – in Townsend’s sense. Mack and Lansley (1985) provided a methodological basis to identify socially perceived standards in society (what is necessary) and deprivation due to enforced lack (deprivation due to poverty) using a simple survey module to survey consensus with regard to what is necessary to live with dignity in society and what are the reasons for the deprivation of such needs.
Mack and Lansley’s approach has been successfully implemented in more than 50 countries to collect and produce poverty indices with very low levels of random and systematic error. From the perspective of measurement theory, that means that it is necessary to have instruments that lead to demonstrably reliable and valid (deprivation) scores (Gordon, 2006). It has been implemented in developed, developing, highly unequal and somewhat equal big and small countries. For example, for the European Union (Guio et al, 2016; 2017), Australia and Hong Kong (Lau et al, 2019; Saunders and Naidoo, 2020) and for countries like Uganda, Benin, South Africa and the islands in the South Pacific (Wright and Novel, 2013; Nandy and Pomati, 2015; Nájera et al, 2020; Pomati and Nandy, 2020). In Mexico, the CA was conducted in 2007 by CONEVAL to inform the selection of the thresholds of some of the indicators (Guillen, 2017).
Most of the existing poverty measures based on the relative deprivation theory and the CA have operationalised the Townsend concept of resources via income. Recently, Lanau et al (2020) point out that this approach clearly also involves identifying deprivation due to other reasons beyond insufficient incomes, such as a lack of access to basic services. The theory of relative deprivation offers a coherent framework for the measurement of poverty based on data that is explicitly collected for such a purpose using the CA. The question is the extent to which the underlying assumptions of the theory of relative deprivation hold in newer implementations and contexts.
Measuring multidimensional poverty using the consensual approach: the case of the City of Buenos Aires
The statistical office of City of Buenos Aires (the General Directorate of Statistics and Censuses, GDSC/DGEC) developed several indicators on the population’s living conditions and welfare including a monetary poverty measure.
In 2019, with the aim of better understanding the city’s living conditions, GDSC/DGCE embarked on a project to develop a MDP measure based on the CA. The objective was to design a measure based on specific indicators of deprivation to be collected in a module of one of the GDSC/DGEC’s regular household surveys (the Annual Household Survey) (EAH).14 The results were to be representative of the whole population of the CBsAs, including children.15 It was also considered that this experience could contribute to the discussion on methodologies for MDP measurement in Argentina, as it could be replicable for other cities or provinces (or for the country as a whole).
The main features of the adopted methodology, consistent with the CA approach, were that: i) the evaluation of the items of the multiple dimensions for its suitability, that is, those included in the measure should be regarded as necessary by the population of the CBsAs; ii) deprivation in one indicator implies that the person/household could not access the good or is unable to do the activity/take part in the activity in question, due to lack of resources; iii) the measure and the individual items should be evaluated according to its suitability, reliability and validity; and iv) a household would be considered poor according to this multidimensional criteria if it is multiply deprived across a number of the different dimensions.
Regarding the issue of the lack of resources for identifying deprivation, a wide definition is considered. Hence, not only insufficient monetary income was taken into account but also other reasons for not accessing to goods or services, in particular, the non-existence (or serious difficulty in the provision) of those provided by the State.16
Given that the EAH is answered by only one member of the household (who responds on behalf of all household members), it was not possible to evaluate deprivation at the individual level but for the household as a whole. This might generate some bias in the number of people defined as poor and does not allow the analysis of intra-household inequality measurement.
The activities that were followed to obtain the MDP measure involved following steps:
identification of dimensions and indicators: local adaptation of items and preliminary definition of necessities;
discussion in focus groups (FGs);
development of the questionnaire for the pilot survey and data collection;
evaluation of the results from the pilot survey;
design of the questionnaire for the survey and data collection;
evaluation of the results and final selection of indicators;
aggregation and measurement.
First selection of necessities through discussion in focus groups
Initially, a first (and long) list of potential indicators concerning different necessities was elaborated (see Appendix 1) considering past experiences in other countries,17 as well as previous measurements carried out in Argentina (such as the UBN). The indicators include goods, services and activities that account for present living conditions in the CBsAs. Consequently, they exclude those that are universally available or very exclusive. Each item defines simultaneously one aspect of deprivation and the threshold indicating deprivation. These indicators have been defined with precision and the wording has been chosen to be easily understood by people of different social backgrounds. For this, the experience of the GDSC/DGEC in designing and implementing household surveys was an important input, as well as other experts’ experience.
To have an initial view on the suitability of the items included in the long list, they were tested in three focus groups following the methodology of the Poverty and Social Exclusion projects (extensively explained in Gordon et al, 2000; Pantazis et al, 2006; Dermott and Main, 2017). The participants were selected from inhabitants of the CBsAs, pertaining to three different socioeconomic strata (low-middle, middle, middle-high) in terms of family income, education, occupation and neighbourhood (comuna). Additionally, to reflect diversity, participants were of different genders, ages, position in household and differed as to the presence of children in their households. Each of the groups was relatively homogeneous in terms of socioeconomic status.18 The most important activity in the FGs was to reach consensus about which of the items of the long list were considered as necessary for all those living in the CBsAs and about any items that could have been missing from the list. See Appendix 1 for details on the results.
The design of the questionnaire
The pilot survey
In August 2019, a pilot survey was implemented to evaluate a module of the EAH that included the questions needed to produce the selected list of indicators resulting from the FGs to a sample of 250 households. The inclusion of specific questions on deprivation indicators in a larger survey provided further relevant information about each household and its members such as composition, gender, age, occupation and income, among others.
For each item, three questions were asked: i) if it is considered necessary for those living in the CBsAs; ii) whether or not the household, or households’ members, has or does it; and iii) for those answering negatively to the previous questions, whether or not it was due to lack of resources.
The interviewers of the pilot survey suggested that the items were well-understood during the face-to-face interviews. The results obtained from the pilot survey were the main inputs to define the final list of indicators and to formulate the questions to be asked in the actual survey. The answers showed that all items were suitable, as they were considered to be necessary by 70% or more of households. Internet connection in the household was the only exception.
Testing for validity and reliability of the measure
Gordon (2006), drawing upon measurement theory and methods, proposes relying on the scientific principles of reliability and validity to put under scrutiny the amount of random and systematic error of a poverty index. Following Guio et al (2016; 2017), the results were evaluated for validity and reliability to select the optimal set of items to be included in the full survey,
Validity is all the available evidence in favour of the desired interpretation of an index (Bandalos, 2018). When validity holds, it is possible to claim that it measures what it is intending to capture. There are different types of validity: face, criterion, concurrent and construct. Face validity refers to the extent to which the respondents identify the indicators as manifestations of poverty. The CA is devised to guarantee face validity in that it asks the population to classify items as necessary or not necessary. Criterion validity assesses whether an index as a whole or an indicator correlates with measures that capture causes or consequences of the phenomenon of interest: each item can be considered valid if they are correlated with a set of variables known to be associated with poverty. Perry (2002) finds a mismatch between poverty measured by direct indicators and income poverty, but Guio et al (2016) use low income, economic strain and self-reported health-status to test for validity. Also, Townsend (1979) and Mack and Lansley (1985) selected the relevant items based on the correlation between deprivation and income. In the pilot survey, items were considered valid if they were related to low income and low educational level of the household head. Construct validity holds when both the items and the structure of the scale have a common underlying cause. Confirmatory factor methods were used to assess this type of validity (Bandalos, 2018).
Reliability refers to the extent to which a set of indicators are manifestations of a latent variable and result in an internally consistent index (Zinbarg et al, 2005). The ideal way to test for reliability would be to compare at least two independent measures and see if they lead to the same population rankings. If not possible, Cronbach (1951) suggests the use of the square of the correlation between the variability of the measured scale (the sum of individual item scores) and the variability of the underlying factor, the so-called α coefficient (Nunnally, 1967; Nunnally and Bernstein, 1994). The coefficient has been subject to several criticisms so the proportion of the variance of the underlying phenomenon (poverty) that is explained by the variances and covariances between items was proposed as a better statistic: the Omega coefficient (Zinbarg et al, 2005). Data drawn from the pilot survey showed values of alpha over 0.85 and omega over 0.9, suggesting that the resulting index is internally consistent as they are considerably larger than those considered as the minimum desirable for each criterion (Revelle and Zinbarg, 2009; Nájera, 2019).
The testing of the reliability of each of the observed items on measuring the ‘latent trait’ or unobservable phenomenon (that is, poverty) can be achieved using Item Response Theory (IRT) which measures two aspects of each deprivation indicator: its discrimination and severity (Guio et al, 2017; Nájera, 2019). The discrimination parameter describes how fast the probability of success (or failure) changes for different levels of the latent variable, represented by the standardised correlation coefficient between each item and MDP. If the parameter takes values below 0.4 standard deviation from the mean of the latent variable, it suggests low discrimination (Guio et al, 2016). The second parameter indicates how severe latent poverty is likely to be for individuals lacking each item, measured in units of standard deviation from the mean. It is desirable to include indicators with different severity scores in a composed measure. For this parameter, Guio et al (2016) suggest a threshold of three standard deviations from the mean of the latent variable and that any item with higher values would be too severe for use in social surveys. Consequently, only very poor individuals will fail to accomplish that need so that those variables would be unreliable as poverty indicators.
The results of the pilot survey showed that three items were too severe: having a fridge in the household, taking yearly vacations outside the city for one week and having piped water in the household. Vacations and piped water also proved to have low discrimination potential.19
Estimation of multidimensional poverty
Selecting the items
The list of items that was finally included in the module for measuring MD poverty based on the CA reflects the consensus reached in the FGs, the evaluation of suitability derived from the pilot survey and the analysis of validity and reliability of the results of the pilot survey. The module was applied with the EAH corresponding to the last quarter of 2019 on a sample of 9,570 homes, representative of the city’s population. Households were asked about their opinions on 28 items (see Appendix 1), for which they had to assess if they were necessary for living in the city.
An endorsement threshold of 50% is generally used in the CA approach, but given that this is the first implementation of the CA in Argentina, a higher threshold (that is, >70% instead of >50%) was used to reduce the possibility of accepting items with large between group variability. However, only one item (internet) showed endorsement rates below 70% (but actually higher than 50%). For the rest of the items more than 80% of the households considered them as a necessity and over 90% of respondents for most items (Figure 1). Results hold for different types of households (data for households classified as income poor or living in shanty towns are included). Internet connection remains below the threshold as only just over 62% of households identified this service as a necessity. However, the proportion rises when only income poor households (63%) or residents in shanty towns (approximately 75%) are considered. Based on the discussions in the FGs, and comments provided by households, this might be associated to the use of internet through mobile phone service or the presence of elder people who do not know how to use it.

Socially perceived necessities (% of households)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAs.
Socially perceived necessities (% of households)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAs.Socially perceived necessities (% of households)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAs.For each of these items and for three additional questions on food security, respondents were asked if they have the items or could do the activities (and if not, if it was because of lack of resources or other reasons).20 Households were considered to be deprived of an item if they answered that they did not have access to it because of lack of resources (including insufficient/inadequate provision in the case of public services granted by the government). The proportions of households deprived in individual items were very low in the case of traditional UBN indicators, that is, those related to housing and schooling but large in items referring to social activities (Figure 2).

Households with deprivation in individual items (%), City of Buenos Aires, 2019
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAs.
Households with deprivation in individual items (%), City of Buenos Aires, 2019
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAs.Households with deprivation in individual items (%), City of Buenos Aires, 2019
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAs.The validity was tested by correlating each item with both monetary poverty and belonging to the first quintile of income distribution. Only five items showed a non-significant relation, as detailed in Appendix 1. For global and item reliability, analysis was made both for indicators corresponding to households with children (in which case, all items were considered) and for households without children. The results reported here correspond to households without children, but the conclusions hold for both.
As for global reliability, alpha (0,82), omega (0,96) and hierarchical omega (0,90) showed good performance of the group of indicators. The results provided by IRT point to floor and roof material, medical controls and kindergarten and secondary school attendance that proved to be too restrictive or difficult to be useful indicators of poverty. Additionally, getting medical attention when needed and access to contraceptives were also too severe. This means that only very poor households are deprived of those items so that their inclusion in the MDP indicator would not be useful. Households showing deprivation for those will probably also suffer deprivation in other items associated with lower levels of deprivation.21
The questions about food were taken as necessities from the beginning. The results showed, however, that deprivation in one of the questions (about changes in food variety) was unusually high when compared with those drawn from other surveys that asked a very similar question. Moreover, a consistency analysis of the answers of the three questions suggests that that item was not clearly understood during the interviews and was thus discarded.
Considering the results of these tests – after eliminating the item referring to balanced food – four items were left out for households without children and six for households with children (Appendix 1). Consequently, 17 indicators were selected for the first group and 23 for the second (Appendix 2). For both groups of households, the indicators were grouped in five dimensions: food, health and care, housing and services, equipment, social deprivation and education.
All items proved to be valid indicators of poverty and globally and individually reliable both for households with children and for those without children. However, each household was evaluated considering the items according to their composition (with or without children – that is, members aged under 18).22
The aggregation criteria
Having defined the set of items that would be included in the measure, the last methodological aspect to be decided upon was the aggregation procedure or how to define whether a household is poor or non-poor. The poverty situation of each unit was evaluated with the vector of indicators corresponding to its type (with or without children). A household was considered as poor if it showed deprivation in two or more dimensions. This cut-off resulted from following the criterion employed by Gordon et al (2000). It explores the relationship between household income and the number of deprivations; it was precisely at two deprivations where the relationship falls sharply.23 The alternative of using only one would not account for the idea of multidimensionality associated with poverty according to our conceptual framework.
To assess whether each dimension is met or not by the household and, given that the total number of indicators in each of them is different, a proportional threshold has been set at 33% of the indicators included in the dimension. If the household is deprived in at least 1/3 of the total items from a dimension, it will be considered as suffering deprivation in that dimension. There is some degree of arbitrariness in using this – in general, any – proportion of indicators, but it is the case in most examples of MDM that consider two thresholds (Table 1).
The incidence of MDP would be estimated, at least, for two aggregates: i) the whole population of the city and ii) households with children.
Results: multidimensional poverty in the City of Buenos Aires
We estimated MDP rates for total households and for those with children. Of all households, 15.3% are poor, which means that 20.2% of the population lives in that situation (Table 2). As for households with children among their members, the resulting poverty rate is 25.7% and 30.6% of all children are poor. Poverty rate for children from 3 to 5 years old is 32.7%, while for children from 6 to 17 the rate is lower but still higher than for the total population, reaching 31.2% for children from 15 to 17. Poverty incidence for elder adults is lower the average (13.8%). All these figures are worse for people living in households where the reference person is a woman: average rate reaches 26.1% for that group and child poverty reaches 39.9%. As expected, the proportion of people living in shanty towns who are multidimensionally poor is substantially higher than average (72.7% and 79.2% for children).
Multidimensional poverty rates (% of households and % of population)
Poverty rate | Total | Reference person is woman | Shanty town |
---|---|---|---|
Total households | 15.3 | 18.8 | 72.7 |
Households with children | 25.7 | 32.9 | 74.0 |
People | 20.2 | 26.1 | 73.8 |
Children | 30.6 | 39.9 | 79.2 |
3–5 years old | 32.7 | 42.0 | 75.0 |
6–4 years old | 30.1 | 40.1 | 84.5 |
15–17 years old | 31.2 | 36.8 | 74.9 |
Adults more than 64 years old | 13.8 | 17.0 | 66.4 |
Source: MDP Module – EAH – DGEC-CBsAs
Table 3 shows poor people classified according to the number of deprivations suffered, as a proportion of poor people; 49.4% of the poor are deprived in two dimensions, 29.1% are deprived in three and 13.6% in four dimensions. Only 7.9% are deprived in all dimensions. The proportion of poor households with children suffering deprivation in only one dimension is lower (45.2%), while higher proportions are deprived in more dimensions (31.2% in three, 14.2% in four and 9.4% in five dimensions).
Number of deprivations (% of poor people)
Number of deprivations | Average | People in poor households | |
---|---|---|---|
without children | with children | ||
2 | 49.4 | 61.4 | 45.2 |
3 | 29.1 | 23.1 | 31.2 |
4 | 13.6 | 12.1 | 14.2 |
5 | 7.9 | 3.4 | 9.4 |
Source: MDP Module – EAH – DGEC-CBsAs
The indicators with the higher rates of deprivation among the poor households are those relating to food, and two of the three included in the social dimension – vacations and invite friends/family. Instead, those corresponding to housing (including services) and household equipment are the lowest (Table 4).
Households with deprivation in each item (% of poor households and % of poor households with deprivation in each dimension)
Dimension | Deprivation | Poor households with deprivation | Poor households with deprivation among households with deprivation in each dimension |
---|---|---|---|
Food | Skipping one meal | 51.5 | 67.7 |
Eating less | 76.1 | 100.0 | |
Health and care | Assistance for care | 14.6 | 31.2 |
Medicines | 34.5 | 81.3 | |
Clinic studies and analysis | 16.8 | 48.0 | |
Dental treatments | 43.9 | 86.3 | |
Housing and services | Repair leakages | 39.5 | 74.9 |
Hot water | 17.9 | 34.0 | |
Electricity | 15.0 | 28.5 | |
Household equipment | Fridge | 4.6 | 17.9 |
Blankets | 6.9 | 33.9 | |
Adequate clothes | 16.1 | 80.1 | |
Replace clothes and shoes | 44.4 | 96.4 | |
Social deprivation and education | Personal expenditure | 82.6 | 93.8 |
Vacation | 91.1 | 97.1 | |
Inviting | 44.5 | 51.7 | |
Public transportation | 7.3 | 8.6 |
Source: MDP Module – EAH – DGEC-CBsAs
As expected, there is a clear relationship between the poverty index and incomes (Figures 3a and 3b); 46% of households in the first quintile are multidimensionally poor and 18.2% of the second are in that situation. Alternatively, only 1.5% of households that belong to the higher income 20% are poor. This means that almost 60% of all MD poor households belong to the first quintile of the distribution of household incomes.

Distribution of poor households according to quintiles of the distribution of per capita household income
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304

Distribution of poor households according to quintiles of the distribution of per capita household income
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Distribution of poor households according to quintiles of the distribution of per capita household income
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304

Multidimensional poverty rate (5 of households)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAs
Multidimensional poverty rate (5 of households)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAsMultidimensional poverty rate (5 of households)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Source: MDP Module – EAH – DGEC-CBsAsSuch association also emerges when households are simultaneously classified according to MDP and to monetary poverty.24 Its results show that precisely 59% of all persons considered as poor when resorting to the first of these approaches are members of households whose incomes are lower than the poverty line. Much of the other 40% of MD poor households – that is, the non-income poor – correspond to a large extent to low-income households, as could be deduced from Figure 3. Many of them frequently ‘enter’ and ‘exit’ a situation of income poverty mainly due to the instability of the labour market.25 That proportion of MDP households that are also income poor, is larger among children, indicating that their families are more structurally poor (Table 5).
Monetary poverty and multidimensional poverty (% of total population)
Monetary poverty | |||
---|---|---|---|
Multidimensional poverty | Total | Poor | Not poor |
Total population | 100 | 21.5 | 78.5 |
Poor | 20.3 | 11.9 | 8.4 |
Not poor | 79.7 | 9.6 | 70.1 |
Children | 100 | 30.2 | 69.8 |
Poor | 30.6 | 22.2 | 8.3 |
Not poor | 69.4 | 14.6 | 54.8 |
Source: MDP Module – EAH – DGEC-CBsAs
Conclusion and discussion
There are many different approaches to measuring MD poverty and a debate is ongoing about their strengths and limitations. The main objective of this article is to contribute to such discussion in LA showing the advantage of generating indicators specifically designed for measuring multidimensional poverty as many of the measures existing in the region only resort to variables included in regular household surveys. For this, it presents the results of a survey carried out in the CBsAs to illustrate how the CA and the theory of relative deprivation can be implemented to produce more robust poverty data on material and social deprivation. Drawing upon previous international experiences in both developed and developing countries, the article shows how the use of qualitative and quantitative methods (focus groups, pilot survey and representative surveys) leads to a set of deprivation indicators that result in highly reliable and valid poverty scores.
This result echoes the findings and conclusions from previous studies in the region and underlines the importance of generating more suitable indicators of deprivation to measure housing adequacy and education (Nájera and Gordon, 2019; Beccaria and Fernandez, 2021). The exercise also shows how the CA approach permits the derivation of indicators for population groups. In this particular application, a child poverty index has been put forward for the CBsAs.
The method applied for this study can be replicated in virtually any other country interested in improving its poverty data and measures. In countries where the standard UBN-type indicators are no longer suitable to capture the relevant needs of the population in the twenty-first century, the CA offers a sensible alternative to measure more accurately the many dimensions of poverty. However, its implementation requires changes to current data collection instruments and a better understanding of the relevance and suitability of relative deprivation theory for the region.
Among the shortcomings of the method, literature has focused on the effect of adaptive preferences in contexts of deprivation (Nussbaum, 2001) but no evidence of this problem was identified in this case (similar results can be found, for example, in Nandy and Pomati, 2015; UNICEF, 2019; Lau et al, 2019; Saunders and Naidoo, 2020).
Regarding empirical issues, difficulties might be associated with the need for frequent revisions of the set of indicators, as well as the use of a relatively longer, and to some extent, more complex questionnaire, than those usually employed. However, the experiences here reported showed that an official statistical office could include a special module in its continuous household survey without a substantial alteration of the regular process.
More specifically to the experience reported in this article, and consequently not inherent to the method, all indicators on deprivations refer to the household, even if they are of an individual nature. As a result, it is not possible to study differences among household members and draw conclusions at the individual level.
Notes
This conclusion of Nájera and Gordon (2020) was contested by Santos and Villatoro (2020) and then Gordon and Nájera (2020) wrote a response to this latter work. Part of the disagreement seems to rest on different understandings of the meaning of measurement and on discrepancies about the adequacy of adopting contemporary model-based perspectives in measurement.
Buenos Aires is the capital city of Argentina, a middle-income country with 45 million inhabitants, whose per capita GDP was, in 2019, US$ 9.900 (ECLAC). The city concentrates 7.8% of the total population (INDEC) and 21% of total GDP (INDEC, 2018 and DGEyC CABA, 2019a). For further socioeconomic information, see DGEyC-CABA, 2019b.
The activities that led to the production of the UBN in Argentina began with a requirement of a social affairs ministry that needed criteria to geographically focus some of its programmes.
Some exercises (non-official) were also carried out combining the UBN and the income methods (the integrated approach), that is, each unit is classified as poor/non-poor simultaneously according to both criteria (Beccaria and Minujin, 1985; Kaztman, 1989).
The integrated measure combines the ‘space of social rights’ (the specific MD part) and the ‘economic welfare’ space (that identifies those with incomes below a monetary poverty line).
Rights could be understood as the expression of needs, values, interests and goods that, given their relevance, have been considered as fundamental for all persons (CONEVAL, 2011).
In Costa Rica, the ‘dimensions were selected following sectorial or functional approach of public policy’ (INEC, 2015). In Chile, the selection was based on normative judgements, data availability and consensus reached in participatory processes (MDS, 2016). In Colombia, they were defined taking into account the thematic possibilities of the survey and a review of dimensions frequently used in multidimensional indexes, discussion with experts, social rights established by the Constitution, qualitative studies, the Millennium Development Goals and the Government’s social policy to reduce poverty (Angulo Salazar et al, 2011).
Ecuador’s method is based on indicators reflecting the ‘Buen Vivir’ (‘Good Living’) rights as established in the Constitution.
Ad hoc measures based on OPHI approach were also produced in Argentina. For example, see: Ignacio-Gonzalez and Santos (2020).
In Mexico, the equal importance attached to each dimension and indicator and the identification criteria, is clearly related to the right framework. This decision is based on the principles of indivisibility and interdependence of human rights (CONEVAL, 2011).
A multidimensionally poor person, in terms of the Mexican indicator, is one who does not meet at least the threshold of one indicator and his/her income is below the monetary poverty line.
Part of these different sources are often considered by the Canberra Group’s (2001) international standards for the measurement of monetary and non-monetary income.
This conception of relative deprivation detonated a series of famous exchanges between Sen and Townsend about the relative or absolute meaning of needs in relation to poverty. Gordon (2006) offers an overview of the debate and concludes that many differences were semantic and its possible to find common ground among both perspectives.
The survey collects information on labour and socio-demographic aspects. A description can be found at https://www.estadisticaciudad.gob.ar/eyc/?page_id=702
The Annual Household Survey sample is based on two master samples: one of private households excluding shanty towns or slums and a second master sample including households in shanty towns or slums. Primary sampling units (PSU) are selected in each master sample (with a probability of selection estimated as the inverse of the number of households in each unit). In the first master sample the second stage of selection is based on the stratification of the 912 PSU in five groups according to income and household information provided by the population census and previous surveys, to randomly select 9.120 households. The 45 PSUs of the second master simple are considered to pertain to a sixth stratus and 450 households are randomly selected in the second stage. Through a fix margin calibration technique, the sample weights are calculated based on population estimations for each neighbourhood or comuna. For 2019, 5,848 households were surveyed (14,319 people) who represent 1,305,988 households and 3,071,892 people (DGEyC-CABA, 2020).
The sample size for the pilot survey (250 households) was determined by the operational capacity of the statistics office. Households were randomly selected by interviewers based on a scheme provided by the statistics office, parting from the sample of the Quarterly Employment and Income Survey (Encuesta Trimestral de Ocupación e Ingresos).
The interviewer explained to respondents the wide meaning of the word ‘resources’.
The large number of countries includes, among others, 28 countries of the European Union, Uganda, the Kingdom of Tonga and South Korea.
These sessions took place during March 2019.
Piped water resulted too severe according to IRT and therefore was not included in the final survey as it would overlap with other indicators. Even more, it is not usually included in the regular household survey questionnaire. Data from the 2010 Population Census indicate that just 2.5% of all households of the City of Buenos Aires lack this service.
Deprivation questions on eight items were only asked in households with children.
High values of the severity parameter of the IRT tests indicate that deprivation in the item only appears in cases of severe poverty. This means that that item would only be useful to identify severely deprived units, who would also be deprived (and identified) in items with lower results in the severity parameter.
The tests were run for all for households with children and for the reduced group of items (excluding child-related necessities) for households without children.
Annex 3 includes the results of a higher-order confirmatory factor model with the five dimensions. The overall fit was acceptable (TLI and CFI>.95) and all indicators correlated with the overall latent construct (Multidimensional Poverty). The housing dimension might be ill-specified but as previously noted this has to do with the fact that the indicators of this dimension are too severe. Hence, the amount of random noise added by these indicators is quite low as very few people are deprived of these items. These indicators could be in any case regarded as redundant but there is no evidence (high reliability and good fit) that they affect the ordering of the population.
The DGEyC also regularly estimates the poverty index (head count ratio) using the monetary (or ‘poverty line’ approach. See DGEyC (2016) Construcción de las líneas de indigencia (LI) y pobreza (LP) para la Ciudad de Buenos Aires. Síntesis metodológica, Buenos Aires: DGEyC.
For evidence about income mobility in Argentina, see, for example, Luis Beccaria; Roxana Maurizio; Martin Trombetta; Gustavo Vázquez (2021) ‘Short-term income mobility in Latin America in the 2000s: intensity and characteristics’. Socio-economic Review, doi:
Funding
The qualitative fieldwork of the pilot study was partially funded with funds from the Global Challenges Research Fund from the ESRC and the University of Bristol (H100004-104).
Acknowledgements
We would like to thank Professor David Gordon (University of Bristol) and three anonymous referees for their comments. All the potential mistakes remain the authors’ responsibility.
Conflict of interest
The authors declare that there is no conflict of interest.
References
Altimir, O. (1979) La Dimensión de la pobreza en América Latina, Cuadernos de la CEPAL, (Nº27).
Angulo Salazar, R., Diaz Cuervo, Y. and Pardón Pinzón, R. (2011) Índice de Pobreza Multidimensional para Colombia, (IPM-Colombia) 1997–2010, in Archivos de Economía, Doc. Nº 382, Bogota: Departamento Nacional de Planeación.
Bandalos, D. (2018) Measurement Theory and Application for the Social Sciences, New York: The Guilford Press.
Beccaria, L. and Minujin, A. (1985) Alternative methods for measuring the evolution of poverty, in Proceedings of the International Statistical Institute, Amsterdam: International Statistics Institute.
Beccaria, L. and Fernández, A.L. (2021) Measuring multidimensional poverty using household surveys, Problemas del Desarrollo, 51(200): 129–56.
Boltvinik, J. (2013) Medición multidimensional de la pobreza. AL de precursora a rezagada, Revista Sociedad y Equidad, 1(5): 4–29.
Callan, T., Nolan, B. and Whelan, C. (1993) Resources, deprivation and the measurement of poverty, Journal of Social Policy, 22(2): 141–72. doi: 10.1017/S0047279400019280
Canberra Group (2001) Expert Group on Household Income Statistics. Final Report and Recommendations, Otawa: Canberra Group.
CEPAL-UNICEF (2010) Pobreza infantil en América Latina y el Caribe, Santiago de Chile: CEPAL – UNICEF, https://repositorio.cepal.org/handle/11362/1421.
CONEVAL (2007) Encuesta para determinación de umbrales multidimensionales de pobreza 2007, Consejo Nacional para la Evaluación de la Política de Desarrollo Social, México.
CONEVAL (2011) Metodología Para la Medición Multidimensional de la Pobreza en México, México: CONEVAL.
Cronbach, L.J. (1951) Coeffcient alpha and the internal structure of tests, Psychometrika No. 16, 297–334. doi: 10.1007/BF02310555
Dermott, E. and Main, G. (eds) (2017) Poverty and Social Exclusion in the UK, Vol 1, Bristol: Policy Press.
DGEyC-CABA (2010) Hogares y población en viviendas particulares y distribución porcentual por condición de Necesidades Básicas Insatisfechas (NBI), Áreas Programáticas de Salud de Hospitales Generales de Agudos, Buenos Aires: DGEyC, https://www.estadisticaciudad.gob.ar/eyc/?p=61434.
DGEyC-CABA (2019a) Producto Geográfico Bruto a precios básicos. Valor Bruto de Producción, Buenos Aires: DGEyC, in https://www.estadisticaciudad.gob.ar/eyc/?p=70385.
DGEyC-CABA (2019b) Revista Cuidad Estadística, Año 2, número 2, octubre, https://indd.adobe.com/view/c893d592-007e-47b8-a4a0-71549f0c755f.
DGEyC-CABA (2020) Encuesta Anual de Hogares 2019. Ciudad de Buenos Aires. Síntesis metodológica, https://www.estadisticaciudad.gob.ar/eyc/wp-content/uploads/2022/07/2019_sintesis_metodologica.pdf.
ECLAC (2014) Social Panorama of Latin America 2014, Santiago de Chile: ECLAC.
Gordon, D. (2006) The concept and measurement of poverty, in C. Pantazis, D. Gordon and R. Levitas (eds) Poverty and Social Exclusion in Britain: The Millennium Survey, Bristol: Policy Press, chapter 2, pp 29–69.
Gordon, D. and Pantazis, C. (1997) Measuring poverty: breadline Britain in the 1990s, Breadline Britain in the 1990s, Aldershot: Ashgate, pp 5–47.
Gordon, D. and Nájera Catalán, H.E. (2020) Reply to Santos and colleagues ‘The importance of reliability in the multidimensional poverty index for Latin America (MPI-LA)’, The Journal of Development Studies, 56(9): 1790–94. doi: 10.1080/00220388.2019.1663178
Gordon, D., Adelman, L., Ashworth, K., Bradshaw, J., Levitas, R., Middleton, S., Pantazis, C., Patsios, D., Payne, S., Townsend, P. and Williams, J. (2000) Poverty and Social Exclusion in Britain, York: Joseph Rowntree Foundation.
Guillén, Y. (2017) Multidimensional poverty measurement from a relative deprivation approach, Doctoral dissertation, University of Bristol.
Guio, A.C., Marlier, E., Gordon, D., Fahmy, E., Nandy, S. and Pomati, M. (2016) Improving the measurement of material deprivation at the European Union level, Journal of European Social Policy, 26(3): 219–333. doi: 10.1177/0958928716642947
Guio, A.C., Gordon, D., Najera, H. and Pomati, M. (2017) Revising the EU Material Deprivation Variables, Luxembourg: European Union.
Ignacio-Gonzalez, F. and Santos, M. (2020) Pobreza multidimensional urbana en Argentina. ¿Reducción de las disparidades entre el Norte Grande Argentino y Centro-Cuyo-Sur? (2003–2016), Cuadernos de Economía, 39(81): 795–822, [Multidimensional poverty in urban Argentina: Reduction of disparities between the Norte Grande Argentino and Centro-Cuyo-Sur? (2003–2016)].
INDEC (1984) La Pobreza en Argentina, Buenos Aires: INDEC.
INDEC (2019) Informe de avance del nivel de actividad. Cuarto trimestre de 2018, Informe Técnico, 3(50), https://www.indec.gob.ar/uploads/informesdeprensa/pib_03_19.pdf.
INDEC (n/d) Proyecciones de población por sexo y grupo de edad 2010–2040, para cada provincial, in https://www.indec.gob.ar/indec/web/Nivel4-Tema-2-24-85
INEC (2015) Índice de Pobreza Multidimensional (IPM) Metodología, San José (Costa Rica): INEC.
Kaztman, R. (1989) The heterogeneity of poverty: the case of Montevideo, in CEPAL Review, 37: 131–42. doi: 10.18356/b98eec41-en
Lanau, A. and Fifita, V. (2020) Do households prioritise children? Intra-household deprivation a case study of the South Pacific, Child Indicators Research, 1–21.
Lanau, A., Mack, J. and Nandy, S. (2020) Including services in multidimensional poverty measurement for SDGs: modifications to the consensual approach, Journal of Poverty and Social Justice.
Lau, M.K., Gordon, D., Zhang, M.F. and Bradshaw, J. (2019) Children’s and adults’ perceptions of child necessities in Hong Kong, Social Policy & Administration, 53(6): 835–53.
Mack, J. and Lansley, S. (1985) Poor Britain, London: Allen & Unwin.
MDS (Ministerio de Desarrollo Social) (2016) Nueva metodología de medición de la pobreza por ingresos y multidimensional, Serie Documentos Metodológicos, Nº28, Santiago de Chile: MDS.
Nájera, H. (2016) Youth Poverty and Social Inequalities in Mexico, Doctoral Dissertation, Bristol: University of Bristol.
Nájera, H. (2019) Reliability, population classification and weighting in multidimensional poverty measurement: a monte carlo study, Social Indicators Research, 142(3): 887–910.
Nájera, H. (2020) La confiabilidad estadística de la medición oficial multidimensional de la pobreza en México: 2008-2018, in F. Cortés (ed) Medición Multidimensional de la Pobreza en México, FLACSO-México, pp 77–102.
Nájera, H. and Gordon, D. (2020) The importance of reliability and construct validity in multidimensional poverty measurement: an illustration using the Multidimensional Poverty Index for Latin America (MPI-LA), Journal of Development Studies, 56(9): 1763–83.
Nájera, H., Fifita, V.K. and Faingaanuku, W. (2020) Small-area multidimensional poverty estimates for Tonga 2016: drawn from a hierarchical Bayesian estimator, Applied Spatial Analysis and Policy, 13(2): 305–28.
Nandy, S. and Pomati, M. (2015) Applying the consensual method of estimating poverty in a low income African setting, Social Indicators Research, 124(3): 693–726. doi: 10.1007/s11205-014-0819-z
Nunnally, J. (1967) Psychometric Theory, New York: McGraw-Hill.
Nunnally, J. and Bernstein, I. (1994) Psychometric Theory, New York: McGraw-Hill.
Nussbaum, M. (2001) Women and Human Development: The Capabilities Approach, Cambridge: Cambridge University Press.
Oficina de Planificación Nacional (ODEPLAN) and Instituto de Economía de la Universidad de Chile (1975) Mapa de la Extrema Pobreza, Santiago de Chile: ODEPLAN.
Pantazis, C., Gordon, D. and Levitas, R. (2006) Poverty and Social Exclusion in Britain: The Millennium Survey, Bristol: The Policy Press.
Perry, B. (2002) The mismatch between income measures and direct outcome measures of poverty, Social Policy Journal of New Zealand, 19.
Piachaud, D. (1987) Problems in the definition and measurement of poverty, Journal of Social Policy, 16(2): 147–64. doi: 10.1017/S0047279400020353
Pomati, M. and Nandy, S. (2020) Measuring multidimensional poverty according to national definitions: operationalising target 1.2 of the sustainable development goals, Social Indicators Research, 148(1): 105–26. doi: 10.1007/s11205-019-02198-6
Revelle, W. and Zinbarg, R. (2009) Coefficients alpha, beta, omega and the GLB: comments on Jitsma, Psychometrika, 74(1): 145–54. doi: 10.1007/s11336-008-9102-z
Santos, M.E. and Villatoro, P. (2018) A multidimensional poverty index for Latin America, Review of Income and Wealth, 64(1): 52–82. doi: 10.1111/roiw.12275
Saunders, P. and Naidoo, Y. (2020) The overlap between income poverty and material deprivation: sensitivity evidence for Australia, Journal of Poverty and Social Justice.
Townsend, P. (1979) Poverty in the United Kingdom, London: Penguin Books.
Townsend, P. (1981) Rejoinder to pichaud, new society. Reproduced, in P. Townsend (1993) The International Analysis of Poverty, London: Harvester/Wheatsheaf.
UNICEF (2019) Multidimensional child poverty and deprivation in Uganda: Volume 1. The Extent and nature of multidimensional child poverty and deprivation, UNICEF, UBOS, Cardiff University, Bristol Poverty Institute, The Republic of Uganda.
United Nations (1995) The Copenhagen Declaration and Programme of Action: World Summit for Social Development, New York: United Nations Department of Publications.
United Nations Economic Commission for Europe (2011) Canberra Group Handbook on Household Income Statistics, 2nd edn, Geneva: United Nations Economic Commission for Europe.
Villatoro, P. and Santos, M.E. (2019) ¿Quiénes son pobres? Análisis de su identificación en América Latina, Problemas del Desarrollo. Revista Latinoamericana de Economía, 50(199): 3–29.
Wright, G. and Noble, M. (2013) Does widespread lack undermine the socially perceived necessities approach to defining poverty? Evidence from South Africa, Journal of Social Policy, 42(1): 147–65. doi: 10.1017/S0047279412000530
Zinbarg, R.E., Revelle, W., Yovel, I. and Li, W. (2005) Cronbach’s α, Revelle’s β, and McDonald’s ω H: their relations with each other and two alternative conceptualizations of reliability, Psychometrika, 70(1): 123–33. doi: 10.1007/s11336-003-0974-7
Appendix
Appendix 1
Process of selection of indicators (in black items that were included in the measure)
List of items | Reasons for exclusion | ||||||
---|---|---|---|---|---|---|---|
Focus groups | Pilot survey | Survey | |||||
Not necessary | Difficult to understand | Low discrimination potential | Difficult to understand | Not considered necessary | Validity tests | Reliability tests | |
Having floors of solid material inside the house (for example: tiles, wood or ceramic) | x | x | |||||
Roofs made of solid materials like: cement, tile, slab, membrane, roof tiles or tin (with interior ceiling) | x | x | |||||
Being able to repair leaks in the roof | |||||||
Being able to keep the house warm in winter | x | ||||||
Being able to keep the house cool in summer | x | x | |||||
Having running water inside the house | x | ||||||
Having hot water in bathroom and kitchen | |||||||
Having one of the following fuels for cooking: piped gas, gas cylinder, electricity | x | ||||||
Having electricity in the house with a legal connection | |||||||
Having broadband internet connection in the house | x | ||||||
Having enough rooms (excluding bathroom, kitchen, hallways) for children and adults to sleep separately | |||||||
Having one bedroom for each child/adolescent | |||||||
Having an adequate place in the household for children and adolescents to do their homework | |||||||
Having a refrigerator to preserve food | |||||||
Being able to replace or repair damaged furniture or appliances/make small repairs in the house | x | ||||||
Having enough blankets for winter | |||||||
Having adequate and enough clothes for the whole year | |||||||
Being able to replace damaged shoes and clothes of household members | |||||||
Being able to buy clothes in fancy shops | x | ||||||
Being able to change the car each 5 years | x | ||||||
Being able to travel in public transport | |||||||
Being able to take a taxi in case of need (urgency) | x | ||||||
Going on vacation out of the city at least one week a year | |||||||
Having support for the adequate care for children or adults | |||||||
Eating at least two meals a day | |||||||
Eating what you think you should (quantity) | |||||||
Eating varied and balanced | x | ||||||
Receiving medical attention when needed | x (hh without children) | ||||||
Obtaining medication prescripted by the doctor when sick | |||||||
Being able to do all the medical tests (X-rays, blood tests, etc.) suggested by the doctor | |||||||
Do all the necessary dental treatments | |||||||
Being able to obtain contraceptives | x | ||||||
Bring children up to 12 years old to annual medical controls (even if healthy) | x | x | |||||
Children from 3 to 5 attend kindergarten | x | x | |||||
Children from 15 to 17 attend school | x | x | |||||
Children who attend school have all the required school items | |||||||
Children who attend school have acces to notebook/netbook/tablet for studying at home | |||||||
Being able to spend a small amount of money per week on personal expenses | |||||||
Being able to invite friends/family for eating or having coffe/drinks at least once a month | |||||||
Children and adolescents can be member of a club to practice sport or other leisure activities | x |
Appendix 2: Selected items employed in the poverty measure
Households without children | Households with children |
---|---|
1. Food (2 indicators) | 1. Food (2 indicators) |
Skipping meals | Skipping meals |
Eating less | Eating less |
2. Health and care (4 indicators) | 2. Health and care (5 indicators) |
Medicines | Doctor |
Medical studies | Medicines |
Dental treatment | Medical studies |
Assitance for care | Dental treatment |
Assitance for care | |
3. Housing and services (3 indicators) | 3. Housing and services (6 indicators) |
Repairing leaks | Repairing leaks |
Hot water | Hot water |
Electricity | Electricity |
Separate rooms | |
Beds for each child | |
Place to do homework | |
4. Household equipment (4 indicators) | 4. Household equipment (4 indicators) |
Refrigerator | Refrigerator |
Blankets | Blankets |
Adequate clothes | Adequate clothes |
Replace clothes | Replace clothes |
5. Social deprivation and education (4 indicators) | 5. Social deprivation and education (6 indicators) |
Personal expenses | Personal expenses |
Vacations | Vacations |
Invite friends/family | Invite friends/family |
Public transport | Public transport |
School items | |
Computer for studying |
Appendix 3: Analysis of dimensions

Confirmatory Factor Model. Adults. (TLI = .98; CFI = .98; RMSEA<.005)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304

Confirmatory Factor Model. Adults. (TLI = .98; CFI = .98; RMSEA<.005)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304
Confirmatory Factor Model. Adults. (TLI = .98; CFI = .98; RMSEA<.005)
Citation: Journal of Poverty and Social Justice 31, 1; 10.1332/175982721X16644668262304