Unpacking multidimensional poverty in Brazil through a gendered perspective

Desvendando a pobreza multidimensional no Brasil sob uma perspectiva de gênero

Abstract

This paper analyzes multidimensional poverty in Brazil by gender, race, and location. A Multidimensional Poverty Index was constructed using eleven indicators (years of schooling, literacy, employment, income, electricity, sanitation, water, garbage collection, cooking fuel, assets, and overcrowding) grouped into four dimensions: education, employment, income, and living standards. Data from the National Household Sample Survey (PNAD) for the period 2004-2015 were used. The results show that women experience higher levels of deprivation than men. They represent most of the poor and face more intense poverty. The situation is particularly severe for Black women, those living in rural areas, and those residing in the Northeast region. Employment and income are the main drivers of multidimensional poverty, highlighting women’s difficulties in transforming education into income.

Keywords:
Gender; Multidimensional Poverty; Race; Brazilian Regions; Rural Areas

Resumo

Este artigo analisa a pobreza multidimensional no Brasil considerando gênero, raça e localização. Foi construído um Índice de Pobreza Multidimensional com onze indicadores (anos de estudo, alfabetização, emprego, renda, eletricidade, saneamento, água, coleta de lixo, combustível para cozinhar, bens e lotação domiciliar), organizados em quatro dimensões: educação, emprego, renda e condições de vida. Utilizaram-se dados da Pesquisa Nacional por Amostra de Domicílios (PNAD) entre 2004 e 2015. Os resultados indicam que as mulheres apresentam maiores níveis de privação que os homens, sendo maioria entre os pobres e enfrentando pobreza mais intensa. A situação é ainda mais severa para mulheres negras, residentes em áreas rurais e na região Nordeste. As dimensões de emprego e renda são as que mais contribuem para a pobreza multidimensional, evidenciando dificuldades das mulheres em transformar educação em renda.

Palavras-chave:
Gênero; Pobreza Multidimensional; Raça; Regiões Brasileiras; Áreas Rurais

1 Introduction

Gender and poverty are two related phenomena. In fact, cultural and economic attributes lead women to face higher risk of poverty than men. Behind this fact, the primary origin of gendered poverty is the inequality between men and women, which exists in different aspects, such as income distribution, access to credit, control over income and assets, and command over property (Cagatay, 1998; Bennett; Daly, 2014).

In Brazil, not only are women poorer than men, but the proportion of poor women compared to men has increased in the last decades. According to the Femininity Index of Poverty built by the Economic Commission for Latin America and the Caribbean (ECLAC), the Brazilian index in 2001 was 105.5, while in 2022, it was 120.1, higher than the average for Latin American Countries (which is 117.7). This means that in 2022, for every 100 men (between 20 to 59 years old) living in poor households, there were 120 women in the same situation, 15 more women than in 2001 (ECLAC, 2024). It follows that poverty as a gendered phenomenon and its evolution over the years in Brazil is a relevant research theme.

However, a gender-sensitive approach to poverty analysis involves more than making a statistical breakdown by gender (Nussbaum, 2001). Besides being predominantly female, poverty in Brazil is also multidimensional (Fahel et al., 2016; Silva et al., 2016; Serra et al., 2020). People experience different deprivations, such as low consumption, inadequate living conditions, poor health, low life expectancy, lack of access to education, limited knowledge and information, and absence of power in several areas (Ferreira, 2011).

Women, in particular, are at a disadvantage in several of these dimensions. They receive lower wages for similar occupations, supply fewer hours at paid jobs due to childcare and household responsibilities, do not receive support as household heads, and are more vulnerable to shocks (Klasen et al., 2015; Batista; Costa, 2019; Batista; Costa, 2022). In Brazil, the female participation rate in the labor market was only 54.5% in 2019, against 73.7% for males. Their participation in managerial positions in the same year was only 37.4%. In contrast, the number of hours women spend in unpaid work doubled that of men (IBGE, 2021). When intersecting gender and race, the disadvantages are even more pronounced. In 2018, black women were paid less than half of white men's wages (IBGE, 2019).

In fact, adopting a gender-sensitive, multidimensional approach to evaluating poverty is particularly valuable for Brazil, given its current history of unstable macroeconomic and political conditions, reversing the downward trend in poverty that it had achieved since the 2000s. According to ECLAC (2021), poverty¹ decreased in Brazil, reaching its lowest point in 2014 (16.5%), rose to 20.3% in 2017, and decreased to 19.2% in 2019. These data illustrate the instability of poverty levels in Brazil in recent years, reflecting, according to Vegh et al. (2017), the country's recession and underscoring the impact of the business cycle on social indicators. Behind these consummated data on poverty is the reasoning that not all individuals respond to such shocks the same way (Glewwe; Hall, 1998; Kosec; Song, 2018; Berniell et al., 2020). Indeed, according to Batista and Costa (2022), economic shocks can alter the household's vulnerability status, being more pervasive for those headed by a woman in Brazil.

Not all households headed by women are equal, and there is also a variation among different ethnic groups. In fact, race stratifies people's lives and can lead to different opportunities. Ethnic minority people experience discrimination in the labor market, are segregated into low-income occupations, receive smaller wages for similar work, and have fewer opportunities to achieve higher positions (Reskin, 2012). In Brazil, 32.9% of black people were below the income poverty line in 2018, against 15.4% of white people(IBGE, 2019), and represented only 29.9% of managerial positions.

The location where women live is also important in a multidimensional poverty perspective, since regional characteristics may impact their sources of deprivation (Brady; Burton, 2016). Brazil shows significant local heterogeneities that must be accounted for when evaluating poverty: in 2019, 33.5% of the rural population was below the poverty line, an incidence 15.5 percentage points higher than that of urban areas (ECLAC, 2021). Likewise, states of the North and Northeast have been consistently more deprived than the other regions in Brazil (Rodrigues, 2014).

In this context, this research aims to assess multidimensional poverty in Brazil, by constructing and analyzing a Multidimensional Poverty Index and breaking it down by gender, with attention to race and regional disparities. Specifically, we seek to evaluate whether women are multidimensionally poorer than men, which dimensions of multidimensional poverty they are more deprived of, if there are race and regional differences in their deprivation, and how their poverty evolved between 2004 and 2015. This period was marked by the most significant reduction in monetary poverty in recent Brazilian history (Rocha, 2013; Rocha, 2019) and decrease in inequality, with the Gini coefficient of household incomes falling from 0.57 to 0.52 (Skoufias et al., 2017). Nonetheless, gender inequalities rose: the femininity index went from 106.7 to 116.6 (ECLAC, 2021), which brings an interesting scenario to evaluate.

We follow the Multidimensional Poverty Index (MPI), proposed by Alkire and Foster (2011a), with modifications to accommodate subgroups’ situations and our gender perspective. MPI has been widely used in literature due to its simplicity and effectiveness in measuring multidimensional poverty. In fact, it can be used to target the poorest population, track Sustainable Development Goals, and design policies that directly address the deprivations poor people experience (Alkire; Santos, 2014). There is a recognition that it provides a better measure of poverty than the income-based approach (Alkire; Foster, 2011a; Ravallion, 2011; Bourguignon; Chakravarty, 2019), which is specifically true for Latin American countries, as stated by Santos and Villatoro (2018).

Several authors have measured multidimensional poverty and its heterogeneities worldwide with important insights for our study (as Aguilar and Sumner (2020); Bastos et al. (2009); Batana (2013); Trani et al. (2016)). In general, they attest the importance of the recognition of the multidimensional nature of poverty (Santos and Villatoro (2018) for Latin America); the need to account for the differences among regions within the individual countries (Battiston et al. (2013), Dedecca et al. (2012) for Brazil); which dimensions and indicators should be considered (Barros et al. (2006) for Brazil) and the relative differences concerning the situation of women (Oliveira; Lima, 2023).

Nonetheless, our focus on gender with race and location intersections is a new approach for multidimensional poverty for Brazil, in which we differentiate and contribute to the literature on four main points. First, we assess the different experiences men and women have of poverty and its multiple dimensions. Second, we evaluate the temporal evolution of multidimensional poverty between 2004 and 2015. Third, we analyze the intersection of race and gender to provide a more detailed analysis of poverty. Lastly, we analyze how multidimensional poverty is spread throughout the Brazilian territory considering the differences between rural/urban areas and among the 26 states plus the Federal District. With our rich analyses of Brazil`s multidimensional poverty situation from a gendered perspective, we aim to provide input for a better understanding of the poverty phenomenon and its particularities in Brazil, thus enabling us to provide a solid foundation for formulating effective public policies.

The paper is organized as follows: Section 2 discusses the theoretical framework used to understand the multidimensional nature of poverty. Section 3 describes our methodology for the construction of our index. Our results are discussed in Section 4, and Section 5 outlines our conclusions.

2 The capabilities approach and the multidimensional poverty index

The one-dimensional approach to poverty, widely used in literature, aggregates all of an individual's achievements into a single variable, usually income or well-being, and uses a cutoff point (the poverty line) to determine who is deemed poor (Kageyama; Hoffmann, 2006; Mclanahan; Kelly, 2006). This method, although an important initial measure, does not distinguish the most widely deprived, underestimating the number of poor individuals (Alkire; Foster, 2011a). Income as a single indicator of well-being is considered limited since it does not incorporate other critical dimensions of poverty (Thorbecke, 2008).

With a new view of poverty, Amartya Sen introduced the capabilities approach, in which 'real poverty' arises from the deprivation of capabilities and freedom. In that sense, the income approach may not be sufficient to compare different realities since it does not account for poverty's relative components. Several variables, such as gender, race, and location, enhance the differences between income and capabilities. Despite their differences, the two approaches are interconnected and complementary. The capabilities perspective sheds light on the fact that an increase in an individual's capabilities heightens their potential to be productive and raise income, which can be especially important to reduce monetary poverty. Also, an increase in income should give the individual means to achieve higher capabilities (Sen, 1976; 1999).

To achieve development, defined as the expansion of freedoms, it is important to remove sources of freedom deprivation, such as poverty, the lack of economic opportunities, public sector negligence, and excessive state intervention. Without such freedom, individuals live in a critical condition, lacking prospects for the future. By arguing that development is the expansion of freedom, Sen (1999) claims that reducing deprivations is one way to achieve such freedom. Therefore, poverty must be seen as an absence of elementary capabilities instead of only low income (Sen, 2018). Capabilities are considered the basic needs of an individual, such as education, health, well-being, freedom, political engagement, and others.

According to Alam (2011), the capabilities approach contributes to the gender and poverty debate in three main aspects. It helps monitor the differences in fundamental achievements between men and women over space and time, draws attention to regional differences in gender inequality, and helps uncover aspects that persist regardless of economic growth. In turn, Nussbaum (2001, 2003) suggested that the capabilities have a tight/close relationship with human rights, making it valuable to approach gender inequality issues. Gender inequality, as stated by her, when added to a poverty scenario, results in a severe failure of the central human capabilities. Women lack opportunities to play, to cultivate cognitive faculties, do not have proper bodily integrity, and live with fear, all of which come from unequal social and political circumstances. Thus, women tend to be the most likely to not live a dignified, full, quality life (Nussbaum, 2001; 2003).

There are also important connections between the poverty situation of women from race and location perspectives. Minority ethnic women have less access to the central capabilities than men, and as black, they experience more failure in their capabilities than white people. The magnitude of the gender disadvantage is amplified by the race disadvantage, resulting in double jeopardy for these women (Hardy; Hazelrigg, 1995). Besides that, structural constraints shaped by gender and racial discrimination affect people's access to resources, impacting their chances to escape poverty (Cabaniss; Fuller, 2005). The location where women live is also important. Indeed, women in rural and remote areas face unequal gender relations, leading to failure in central human capabilities. They are not prioritized when receiving inheritances, have fewer employment opportunities, and tend to be the only persons responsible for the housework and childcare (Luther; Gerhardt, 2018; Ramundo Staduto et al., 2013).

Although the capabilities approach can help tackle the deprivation suffered by different population subgroups, shedding light on their specific needs, it cannot provide helpful guidance to reduce gender inequality and poverty if there is no definition of the central capabilities (Nussbaum, 2001; 2003). In that sense, Martha Nussbaum went beyond Sen. She introduced a list of central human functional capabilities, providing a threshold level to citizens' demand from their governments, which is the absolute minimum a person needs to live a quality life (Nussbaum, 2001).

To understand the dimensions of poverty experienced by individuals in Brazil, we used the Multidimensional Poverty Index (MPI), proposed by Alkire and Foster (2011a), which has been widely used in the literature due to its simplicity and effectiveness in measuring multidimensional poverty. The MPI can be used to target the poorest population, track Sustainable Development Goals, and design policies that directly address the deprivations that poor people experience (Alkire; Santos, 2014). Like any method, the MPI has limitations since it cannot capture all sides of poverty in one index. However, there is an understanding that the index provides a better measure of poverty than the income-based approach, as discussed above (Alkire; Foster, 2011a; Ravallion, 2011; Bourguignon; Chakravarty, 2019).

That said, several authors have measured multidimensional poverty and its heterogeneities worldwide. Aguilar and Sumner (2020) show that global multidimensional poverty is predominantly rural and concentrated among children and young adults, with rural deprivations being largely associated with deficits in infrastructure, education, and living conditions. Gender disparities are also recurrent. In Portugal, Bastos et al. (2009) find that women experience higher and more severe multidimensional poverty than men. This result is particularly true among older, isolated, and single-parent women. Batana (2013) highlights education as the main source of deprivation among women in Sub-Saharan Africa. Evidence from fragile and conflict-affected settings reinforces these patterns, as Trani et al. (2016) show that women, rural populations, and people with disabilities face the worst multidimensional outcomes in Afghanistan.

For Latin America, Battiston et al. (2013) document higher multidimensional poverty in rural areas of El Salvador, Mexico and Brazil, emphasizing sanitation and education as priority dimensions for policy intervention. Santos and Villatoro (2018) estimate that nearly 28% of the Latin American population was multidimensionally deprived in 2012, with living standards contributing most to overall deprivation. The authors also highlight the weak correspondence between monetary and multidimensional poverty (Santos; Villatoro, 2018).

In the Brazilian context, several studies identify persistent spatial and gender inequalities. Barros et al. (2006) and Dedecca et al. (2012) show that rural households, especially those headed by women, face deeper deprivations due to worse living and working conditions. Subsequent evidence confirms strong regional and gender disparities, with higher multidimensional poverty among women in the North and Northeast (Ferreira; Marin, 2016; Oliveira; Lima, 2023), among rural populations (Silva et al., 2016), and in remote municipalities (Serra et al., 2020).

Despite this extensive body of work, analyses jointly addressing gender, race, and spatial dimensions at the individual level remain scarce in Brazil. This study contributes to literature by explicitly examining gender differences in multidimensional poverty across multiple dimensions and population subgroups, allowing for a more nuanced understanding of women’s vulnerability and supporting the design of more targeted public policies.

3 Methodology

In this section we discuss the construction of the Multidimensional Poverty Index (3.1), the definitions made (3.2), and the data used (3.3).

3.1 The multidimensional poverty index

The Multidimensional Poverty Index (MPI), developed by Alkire and Foster (2011a) introduces an intuitive approach to identify who is poor through two cutoff points. The first one identifies whether the person is deprived of any dimension, while the second delimits how many dimensions the person must be deprived of to be considered poor. Thus, the MPI methodology identifies people experiencing poverty and determines which dimensions drive multidimensional poverty among different groups of people (Alkire; Seth, 2012).

The construction of the MPI must follow nine steps (Alkire; Santos, 2014): 1) Define the dimensions and their indicators; 2) Define the cutoff points, z, for each indicator; 3) Apply the cutoff points to define whether each individual, in each indicator, is deprived or not; 4) Select the weights for each indicator, which should add up to 1; 5) Create each person's weighted proportion of deprivation, which is called deprivation score; 6) Determine the poverty cutoff, k, and identify the poor individuals; 7) Compute the proportion of people identified as poor, which is the headcount ratio, H; 8) Compute the average portion of weighted indicators in which each poor person is deprived, which is done by adding up the deprivation scores of the poor individuals and dividing it by the total number of poor, which will be the intensity of poverty, A; and finally, 9) Calculate the poverty measure $M_0$, multiplicating H by A.

The MPI was constructed by assigning equal weights to each dimension and to each indicator within each dimension. This choice was made since equally weighting the dimensions facilitates the interpretation and the use for public policies (Alkire et al., 2010; Alkire; Santos, 2014).

By accounting if each person is deprived or not in each indicator, we provide each person's weighted proportion of deprivation, the deprivation score. The score allows us to identify the dimensions in which individuals are deprived and the number of deprivations they experience. It allows a comparison of deprivations between subgroups of the population, such as gender, race, and location. To identify who is poor, Alkire and Foster (2011a) consider the cutoff level, k, to be between two extremes: 1, when the person is deprived in only one dimension, and d, if the person is deprived inall dimensions.

The MPI, thus, can be considered a dual method of identifying poverty as it depends both on the deprivation cutoff, z, and the poverty cutoff, k. Our cutoff point, k¸ is as proposed by Alkire and Santos (2014). So, the person must be deprived of 1/3 (33%) of all the indicators to be considered poor. The dimensions and capabilities they refer to, the indicators, cutoff points, descriptions of what we intend to capture with them, and the reference from the literature are presented in the following section.

The next step is to identify the proportion of poor individuals, H, or the incidence of multidimensional poverty, defined by $H=q/n$, where q is the number of individuals considered poor by the dual cutoff, and n is the total number of individuals in the sample. Even though H cannot be broken down to show how much each dimension contributes to poverty, it can be split into subgroups. Hence, it is possible to assess differences in poverty incidence across gender, race, place of residence (rural versus urban), and regions in Brazil.

The intensity of multidimensional poverty, A, provides additional information on the breadth of deprivation experienced by people in poverty. It is the average deprivation share across people experiencing poverty and is given by $A=c/q$ where c represents the portion of possible deprivations that a poor person experiences. The indicator A captures the share of dimensions in which the average poor individual is deprived. It can be disaggregated across population subgroups, enabling the identification of deprivation patterns by gender, race, place of residence (rural versus urban), and region.

Lastly, we are interested in the multidimensional poverty measure that provides information on the prevalence of poverty and the average extent of a poor person's deprivation. Alkire and Foster (2011a) call it the adjusted headcount ratio, which is given by $M_0=H\times A$. Therefore, $M_0$ is the product of the headcount ratio and the average deprivation share. This measure is sensitive to multidimensional poverty's frequency and breadth (Alkire; Foster, 2011a). Also, it satisfies dimensional monotonicity, meaning that if a person becomes deprived in an additional dimension, $M_0$ will increase. Overall, we find that $M_0$ reflects the proportion of weighted deprivations that the poor experience out of the total potential deprivations the society could experience (Alkire; Santos, 2014).

Finally, decomposability is the most significant advantage of this multidimensional poverty measure for our purpose. $M_0$ can be broken down by indicator, which means that it is possible to evaluate the contribution of deprivations in each indicator to overall poverty. And it can be decomposable by subgroups. Hence, calculating such an index allows us to evaluate the temporal evolution, the geographical differences, the profile, and the various characteristics of multidimensional poverty in Brazil.

3.2 Selected dimensions, indicators and cutoffs, and variable definitions

The poverty index proposed here aims to maximize the available information from the household survey conducted in the country. Therefore, we constructed an index with 11 indicators within 4 dimensions which can be seen in Table 1. They were derived from those proposed by Rippin (2016), based on Nussbaum's (2001) central human functional capabilities. Accordingly, the dimensions and their corresponding capabilities are: (i) education, referring to senses, imagination and thought, and practical reason; (ii) employment, referring to affiliation and control over one’s environment; (iii) living standards, referring primarily to bodily health; and (iv) income, included as a proxy for control over one’s environment.

The dimensions and indicators shown in Table 1 are:

Education: This dimension aims to capture deprivations associated with access to and the consolidation of basic cognitive capabilities. According to Sen (1999) and Nussbaum (2001), the ability to think, imagine, and reason is essential for individuals to live a full and dignified life. Therefore, two indicators are included in this dimension: years of schooling, and literacy, following Costa et al. (2018). The first indicator captures situations of early interruption of the educational cycle, which may compromise individuals’ ability to participate socially and economically in society. The literacy indicator, in turn, measures whether the individual has at least the minimal ability to read and write, which is considered essential for accessing information, communicating, and exercising basic rights.

Living Standards: The objective of this dimension is to capture deprivations related to basic living conditions that directly affect individuals’ physical well-being, dignity, and social integration. Within the Capabilities Approach, adequate living standards are closely associated with the capabilities of bodily health and affiliation, as emphasized by Sen (1999) and Nussbaum (2001). To operationalize this dimension, a set of indicators related to housing quality and access to essential infrastructure is employed, including electricity, sanitation, access to safe water, garbage collection, cooking fuel, asset ownership, and overcrowding, following Fahel et al. (2016) and Costa et al. (2018). Together, these indicators capture whether individuals live in environments that ensure minimum conditions of health, safety, and comfort. Deprivations in this dimension reflect structural deficits that are often beyond individual control and tend to be spatially concentrated, particularly in rural areas and economically disadvantaged regions. Consequently, inadequate living conditions constitute a central aspect of multidimensional poverty, as they not only represent failures of basic capabilities in themselves but also exacerbate vulnerabilities in other dimensions, reinforcing persistent and overlapping forms of deprivation.

Table 1:
Dimensions, indicators, and cutoffs

Dimension	Indicator	Weights	Deprived if (cutoff)	Descriptions	Reference
Education	Years of schooling	1/8	Has under nine years of schooling	Captures school dropouts.	Costa et al. (2018)
	Literacy	1/8	Cannot read or write	Captures if the person has minimal education.	Costa et al. (2018)
Employment	Employment	1/4	Has at least one of the following: 1. Activity status as unemployed 2. Wage below minimum 3. Time-poor (works over 10 hours a day, counting paid and unpaid work)2	1. captures individuals who want to work but are not currently employed. 2. captures the precariousness of work. 3. recognizes the importance of leisure time.	Rippin (2016)
Income	Income	1/4	Has an income below the national poverty line3	Measures the ability to get along financially.	Rippin (2016)

Dimension	Indicator	Weights	Deprived if4 (cutoff)	Descriptions	Reference
Living Standard	Electricity	1/28	Has no electricity	Together, they capture adequate shelter conditions.
	Sanitation	1/28	Where a toilet is not connected to the sewage collection network or is shared with other households
	Water	1/28	Does not have piped water available in at least one room, regardless of the source of water supply.
	Garbage collection	1/28	Does not have access to an appropriate garbage collection service		Fahel et al. (2016)
	Cooking fuel	1/28	Cooks with dung, wood, or coal
	Assets	1/28	Has three or fewer assets: radio, television, telephone, fridge, stove, and computer
	Overcrowding	1/28	Has three or more people per bedroom		Costa et al. (2018)

• Source: Constructed by the authors

Employment: The employment dimension seeks to capture deprivations associated with job quality and labor market insertion. Within the capabilities approach, employment can be understood as a socially relevant functioning, as it encompasses social participation, recognition, autonomy, and the ability to enjoy leisure time (Sen, 2018; Nussbaum, 2001). The indicator adopted in this dimension accounts for situations of unemployment, inadequate remuneration, and time poverty, thereby capturing different forms of labor precariousness that may directly compromise individual well-being. These aspects are particularly relevant in contexts marked by gender inequality, in which unequal divisions of paid and unpaid work, occupational segregation, and labor market discrimination tend to exacerbate employment-related deprivations (Sá et al., 2025; Bertrand et al.; 2019; Bobilev et al., 2020).

Income: The last dimension captures deprivations related to individuals’ ability to meet basic needs and maintain an adequate standard of living. In this study, income is operationalized as the capacity to financially sustain oneself and access essential goods and services. The indicator adopted identifies individuals whose income falls below the established poverty line, capturing situations of material insufficiency that may constrain the realization of other capabilities and increase vulnerability to economic shocks. This dimension is included following the proposal of Rippin (2016).

The inclusion of income and employment in multidimensional poverty measures has been the subject of sustained debate in literature. From a theoretical point, rooted in the Capabilities Approach, income is conventionally treated as a means rather than an end. This has motivated the construction of “pure” multidimensional poverty indices that deliberately exclude monetary indicators to focus exclusively on deprivations in achieved functioning (Sen, 2018; Alkire; Foster, 2011). This argument generally assumes a degree of redundancy between monetary and non-monetary indicators, because income can, at least in principle, be used to satisfy multiple needs simultaneously. Therefore, excluding income helps avoid conceptual overlap between resources and outcomes, preserving the normative coherence of capability-based measurement.

However, the empirical separation between monetary and non-monetary measures has been questioned. According to Santos and Villatoro (2018), the argument for excluding income from multidimensional indices is often weak. In contexts characterized by data limitations, income may act as a surrogate for some missing dimensions, such as mobility, nutrition, and health. Also, income can complement information provided by the non-monetary indicators that are prone to measurement error. Moreover, income deprivation may carry independent informational value for public policy design, especially in the context of income-targeted interventions such as conditional cash transfer programs. Along similar lines, ECLAC (2014) argues that incorporating a monetary indicator into a multidimensional poverty index can be justifiable as it facilitates the identification of the poor and enhances the policy relevance of the measure. Reflecting this perspective, several studies have combined monetary and non-monetary indicators in multidimensional poverty measurement across different contexts (Alkire et al., 2014; ECLAC, 2014; Santos, Villatoro, 2018; Santos et al., 2015). A strategy that we also adopt in this study.

A similar tension arises with respect to the inclusion of employment. Some authors argue that labor market status should be conceptualized as a determinant or predictor of poverty rather than as an outcome and therefore should be excluded from poverty indices (Nájera; Gordon, 2020). Others emphasize that employment conditions encompass more than income generation alone (Santos; Villatoro, 2018). Employment status and job quality are closely linked to social integration, economic security, access to social protection, and autonomy, representing dimensions that are particularly salient in societies marked by high levels of informality and gender inequality, such as Brazil. In these contexts, precarious employment and time poverty may constitute direct manifestations of deprivation. Especially for women, whose labor market participation is often shaped by unequal care responsibilities and occupational segregation.

Overall, the selection of dimensions and indicators made in this study reflects a deliberate balance between theoretical consistency and empirical relevance. While acknowledging the conceptual debates surrounding the inclusion of income and employment, this study adopts a multidimensional framework that explicitly accounts for labor market vulnerabilities that are central to gendered poverty in Brazil.

Beyond the choice of dimensions and indicators, multidimensional poverty measurement also requires careful consideration of the level at which deprivation is identified. Even theoretically well-justified dimensions may fail to capture relevant inequalities if the unit of analysis does not reflect how deprivations are experienced. This issue is particularly salient in the presence of gender disparities, as household-level approaches implicitly assume homogeneity within families.

A growing body of literature highlights longstanding methodological debates regarding indicator selection, aggregation strategies, and the empirical validity of multidimensional indices, especially those relying exclusively on household-level variables (Santos et al., 2010; Nájera; Gordon, 2020). Recent advances further stress the limitations of traditional approaches focused on acute poverty and households as homogeneous units, emphasizing the importance of intrahousehold inequalities, gendered dimensions of work and economic autonomy, and moderate forms of deprivation that remain invisible to conventional measures (Tavares; Betti, 2024; Alkire et al., 2023).

The capabilities approach offers a strong theoretical foundation for addressing these limitations. As argued by Nussbaum (2003), this framework has important advantages over other approaches to poverty because it centers on individuals’ real opportunities to be and do, rather than solely on resources or observed outcomes. Building on Sen’s contributions, this perspective shifts the unit of analysis from the household to the individual and moves the analytical focus from access to resources toward control over commodities and the effective conversion of resources into well-being (Jackson, 1998). Consequently, individuals living in the same household may experience distinct forms and intensities of deprivation.

This distinction is particularly relevant for the analysis of gender inequalities. Women may reside in households classified as non-poor according to income-based or household-level indicators, while lacking access to, or control over, household resources, leading to deprivations that extend beyond the monetary dimension (Budlender, 2005; Brady; Burton, 2016). Social and cultural gender norms further reinforce these disparities, implying that even within relatively affluent households women may face restricted agency and well-being (Jackson, 1998). As emphasized by Cagatay (1998), beyond biases in intrahousehold resource allocation, women often face greater obstacles than men in transforming capabilities into income and well-being.

Consequently, poverty cannot be adequately understood as a household-level phenomenon either. Reliance on household variables may obscure substantial gender deprivations and lead to incomplete or biased assessments of multidimensional poverty. Following Sen’s and Nussbaum’s framework, to construct a gender-sensitive multidimensional poverty measurement, one needs to use variables at an individual level instead of the typical household level. Which is what we do in this study. Understanding this point enhances our understanding of multidimensional poverty and inequality, thereby providing a stronger foundation for policy-making.

3.3 Data

We use data from the Pesquisa Nacional por Amostra de Domicílios (PNAD) provided by the Instituto Brasileiro de Geografia e Estatística (IBGE). The PNAD was⁵ an annual survey of a probability sampling of households. It is a complex and self-weighted sampling plan that ensures that all households have the same probability of selection (Silva et al., 2002). We used information on the sample’s weights, strata, and primary sampling unit (PSU) to obtain statistics that represent the population.

Our unit of analysis is the individual; however, household variables were used. Additionally, we restrict the sample to people of age 15 or more , considering the minimum active age defined by IBGE. Including children could bias our results due to their incomplete education and lack of employment and income, which are dimensions of our index. Additionally, observations with null information about the multidimensional poverty indicators were dropped. The period we analyzed is 2004 to 2015, representing the first year in which the PNAD sample included all Brazilian states' rural areas and the last year with available data (except for 2010 when the Demographic Census was carried out instead of this research). Finally, all income variables were brought to 2015 levels to eliminate the impact of inflation.

4 Results

In this section, we present our results divided into 5 subsections: Section 4.1 describes the sample and analyzes the profile of poverty and deprivation in Brazil. Sections 4.2 to 4.4 present the multidimensional poverty index results aggregated into subgroups: first, we present the gender analysis, followed by the race assessment with a gender focus, and third, the location evaluation with a gender emphasis. Finally, section 4.5 presents a sensitivity analysis assessing the robustness of the results.

4.1 The deprivation profile in Brazil

In this section we focus on data from 2015 to describe the sample, the most recent year of the analysis. Table 2 presents the proportion of people from each population subgroup in the sample. There are 243,149 observations, most of which are women, black people, and in urban areas, which is consistent with the Brazilian demographics (IBGE, 2021).

The column “deprivation score” presents the average weighted proportion of deprivation considering our indicators (years of schooling, literacy, employment, income, electricity, sanitation, water, cooking fuel, garbage collection, assets, and overcrowding) for each group of people.

On average, in 2015, among all the deprivations the Brazilian population could experience, they were deprived of almost one-third (29.93%). Women are deprived of 34.90% of the deprivation indicators, while men are only deprived of 24.92%. The deprivation score for black people is of 53.99%, expressively higher than that of white people (26.07%). Also, people living in rural areas are much more deprived than their urban counterparts: 45.39% versus 27.51% of the indicators.

Table 2 also shows that Black women are the most deprived group, with an average deprivation of 38.46% of the indicators, while White men are the least deprived (20.99%). Notably, even White women (30.89%), the least deprived among women, exhibit higher deprivation levels than Black men (28.13%), the most deprived among men. Similarly, when considering place of residence, women living in rural areas present the highest deprivation score (51.53%). Therefore, we find that women exhibit higher levels of deprivation than men across the board in Brazil, with this pattern consistently observed across racial groups and between rural and urban areas. This suggests that gender-based inequalities in social, economic, and political conditions may constrain women’s capabilities (Nussbaum, 2001), ultimately leading to higher levels of deprivation.

Table 2:
Population, deprivation, and poverty profile - Brazil, 2015

	Population	Deprivation Score	Multidimensional Poverty	Income Poverty6
	%
Women	50.19	34.90	61.46	44.77
Men	49.81	24.92	38.54	22.37
White	46.01	26.07	40.41	29.78
Black	53.99	33.21	59.59	36.88
Rural	13.54	45.39	19.27	44.14
Urban	86.46	27.51	80.73	31.96
White women	23.61	30.89	25.84	40.19
Black women	26.58	38.46	35.62	48.83
White men	22.40	20.99	14.58	18.81
Black men	27.41	28.13	23.97	25.28
Rural women	06.13	51.53	10.30	60.99
Urban women	44.06	32.58	51.16	45.51
Rural men	07.41	40.31	08.97	30.20
Urban men	42.40	22.23	29.57	21.00
Total	243,149	29.93	44.17	33.61

• Source: Research results

The multidimensional poverty column in Table 2 presents the percentage of individuals deprived of 33% or more of the indicators. Overall, 44.17% of Brazil’s population was multidimensionally poor in 2015, among which 61.46% are women, which reinforces the existence of a female overrepresentation in poverty, as proposed by the literature (Barros et al., 1997; Liu et al., 2017; Batista; Costa, 2019). We also see that the majority of multidimensionally poor are black (59.59%). It is essential to highlight that being black does not cause poverty but rather the discriminating attitudes towards black individuals enhances their poverty (Hardy; Hazelrigg, 1995). Hence, being poor is not an intrinsic characteristic of Afro-Brazilian people. However, the poverty phenomenon in Brazil is fundamentally black (Carneiro, 2015). Regarding the intersection of gender and race, we find that Black women represent the largest share of the multidimensionally poor population in Brazil (35.62%), indicating that race intensifies women’s poverty, as highlighted by Hardy and Hazelrigg (1995). Additionally, most multidimensionally poor individuals live in urban areas (80.73%), where the majority of the population is concentrated (IBGE, 2021). Nevertheless, this does not mean that individuals living in urban areas are poorer than those living in rural ones. The deprivation score, as shown before, can confirm the inverse, indicating that failing to account for the national context could mislead/distort the poverty analysis, as Costa et al. (2018) also stated.

The proportion of people below the monetary poverty line (defined as half the minimum wage per capita, R$ 394 per month, as proposed by Hoffmann (2000)) was calculated for comparison purposes (Table 2). As expected, the subgroups considered the poorest remain the same. However, the proportion of multidimensionally poor is higher than that of individuals below the income poverty line, which was 33.61% in 2015. These results suggest that income poverty measurement is limited in Brazil and underestimates the number of poor individuals (Alkire; Foster, 2011; Thorbecke, 2008; Sen, 2018).

Figure 1 reports the evolution of the percentage of multidimensional poverty for the overall population, men, and women, from 2004 to 2015. The proportion of people experiencing multidimensional poverty dropped over the years: from 75.91% in 2004 to 44.17% in 2015, a reduction of 41.81%. According to Rocha (2013; 2019), Brazil's new economic growth cycle was established in 2004, which decreased the income poverty rates up to2014. The period was marked by the minimum wage valorization, the expansion of public income transfers, and the labor market behavior, all contributing to poverty reduction (Rocha, 2013; 2019).

Figure 1:
Time evolution of the proportion of multidimensional poverty - Brazil, men, and women, 2004 to 2015

Despite these facts, gender inequality seems to have increased. Figure 1 shows that women have higher poverty rates than men throughout the period analyzed, and their poverty has been reducing more slowly than men's. While men's multidimensional poverty has reduced by more than half, women's has reduced by only nearly 29 percentage points from 2004 to 2015. These combined data show that women are multidimensionally poorer and more deprived than men in Brazil and that their status has evolved slowly, which suggests that the possibility of a convergence between men's and women's situations is far distant. In fact, according to the World Economic Forum's (2021) global gender gap report, considering a series of dimensions, it will take 135.6 years to achieve gender equality worldwide. Therefore, it is safe to suppose that the gender gap in Brazil will not end soon.

Table 3, in turn, presents the proportion of people deprived in each indicator. The dimensions that stand out in Brazil are employment, with more than 50% of the population deprived of it, followed by years of schooling (45.20%), sanitation (40.76%), and income (33.61%). Even though the order of deprivation in which these indicators appear in the country is curious, it corroborates the need for a multidimensional evaluation embracing more than the income aspect of deprivation. Additionally, the incidence of individuals deprived of almost every indicator is greater in rural areas. The only aspect where rural individuals are better off is overcrowding, indicating that they tend to have more living space. As for black people, there are more of them deprived of each indicator. This result validates the data in Table 2 that show that most multidimensionally poor in Brazil are black.

It is interesting to note that men exhibit greater deprivation in education indicators than women: their incidence is 7.75 p.p. higher in years of schooling and 1.92 p.p. higher in the literacy indicator than that of women. In contrast, women are far more deprived of employment and income dimensions. There is a gap of 23.22 p.p. favoring men in the employment indicator and 22.40 p.p. in the income indicator. These results shed light on women’s difficulty in converting education into work and pay. The labor market requires women to study more than men to receive the same wage and have access to better opportunities. According to Melo and Thomé (2018), this is a characteristic of the Brazilian labor market. Women are more educated and work more with productive and reproductive labor, yet they are segregated into informal market segments, receive less than men for the same job, and have fewer opportunities for career growth (Melo; Thomé, 2018).

Table 3:
Proportion of people, by subgroups, deprived of each indicator - Brazil, 2015

	Years of schooling	Literacy	Employment	Income	Electricity	Sanitation	Water	Garbage collection	Cooking fuel	Assets	Overcrowding
	%
Women	41.34	05.17	61.93	44.77	00.09	39.46	03.51	08.25	03.16	02.20	10.29
Men	49.09	07.09	38.71	22.37	00.15	42.06	04.07	10.27	03.98	02.92	09.53
White	38.19	03.83	45.85	29.78	00.04	33.53	02.05	06.14	02.97	01.32	07.01
Black	51.17	08.08	54.20	36.88	00.19	46.92	05.28	11.92	04.08	03.62	12.38
Rural	70.60	15.89	62.94	44.14	00.78	93.93	19.10	62.19	21.87	09.30	09.72
Urban	41.22	04.60	48.39	31.97	00.02	32.43	01.40	00.97	00.70	01.50	09.94
White women	35.07	03.38	56.92	40.19	00.03	32.21	01.88	05.34	02.51	01.11	07.17
Black women	46.90	06.76	66.38	48.83	00.14	45.91	04.97	10.84	03.73	03.16	13.05
White men	41.49	04.30	34.19	18.81	00.05	34.92	02.23	06.99	03.45	01.53	06.84
Black men	55.30	09.36	42.40	25.28	00.24	47.90	05.58	12.95	04.41	04.06	11.73
Rural women	64.53	12.43	75.78	60.99	00.67	93.47	19.00	61.01	21.36	08.46	10.42
Urban women	38.11	04.16	60.00	42.51	00.01	31.95	01.36	00.91	00.63	01.33	10.27
Rural men	75.62	18.75	52.32	30.20	00.87	94.30	19.18	63.16	22.29	10.01	09.13
Urban men	44.45	05.05	36.33	21.00	00.03	32.93	01.43	01.03	00.78	01.68	09.60
Brazil	45.20	06.12	50.36	33.61	00.12	40.76	03.79	09.26	03.57	02.56	09.91

• Source: Research results

Furthermore, black women are in the worst situation in terms of employment and income indicators, 66.38% and 48.83%, respectively, followed by white women (56.92% and 40.19%). In contrast, black men are the most deprived of education and living standards indicators. Moreover, rural men and women are the most deprived in each indicator, except the overcrowding one, and rural men are the most deprived in the education and living standards indicators, while rural women are the most deprived in employment and income.

In summary, our results suggest that women experience higher levels of deprivation and multidimensional poverty than men, and that their poverty has been decreasing at a slower rate than that of men in Brazil. The situation is particularly dire for Black and rural women across all dimensions considered.

4.2 Gender and the multidimensional poverty index

In this section and the next two (4.3 and 4.4) we show the results of the Multidimensional Poverty Index (MPI), which is an interaction between the headcount ratio (H) and the intensity of poverty (A).

Figure 2 shows that 44.20% of the Brazilian population was considered multidimensionally poor (H) in 2015, equivalent to about 90 million people, and they were deprived in 56.33% of the indicators (A). Subsequently, the MPI for Brazil in 2015 was 24.9%, representing the share of the multidimensionally poor, adjusted by the intensity of deprivation they suffer. For that same year, the global MPI calculated by Alkire et al. (2015) was 15.7%, with a 29.8% headcount ratio and a 52.6% average intensity of poverty. Hence, according to our analysis, the Brazilian population was more multidimensionally poor and more intensively poor than the average global population. Comparing our results with those of the other countries in Alkire et al. (2015), we note that Brazil has similar poverty measures to lower-middle-income countries, even though it is an upper-middle-income⁷ country. Therefore, the poverty phenomenon in Brazil, considering all possible dimensions people may be deprived of, must be a focal point for public policy-making.

On the gender analysis, Brazilian women are multidimensionally poorer than men, (54.10% versus 34.20%), and slightly more intensively deprived (a difference of 1.67 p.p. favoring men). Therefore, women have a 11.9 p.p. larger MPI than men. Our results once more are consistent with the view that women are likelier to lack capabilities than men and to not live a dignified, full, quality life (Nussbaum, 2001, 2003; Alam, 2011). We also corroborate the findings in Avila et al (2012); Ferreira and Marin (2016) and Oliveira and Lima (2023) in which women in Brazil are more multidimensionally deprived than men. The gendered nature of multidimensional poverty in Brazil appears to be more of a product of the high count of poor women than that of the intensity of deprivation they suffer, since A was very similar for both sexes.

Figure 2:
Headcount ratio, intensity of poverty, and multidimensional poverty index - Brazil, men, and women, 2015

Figure 3, in turn, presents the evolution of the headcount ratio, the intensity of poverty, and the multidimensional poverty index, between 2004 and 2015. All three measures have been on a path to reduction in the period. The Brazilian headcount ratio experienced a reduction of 41.77%, while the intensity of poverty and the MPI fell by 7.65% and 46.22%, respectively. However, the percentage of multidimensionally poor (H) women has dropped less than men’s, a gap of 15.80 p.p. favoring the latter. The same happened to the MPI: a reduction of 50.22% for men versus a decrease of 39.73% for women.

The intensity of poverty was on a reduction path until 2006, became almost constant between 2007 and 2014, and then rose in 2015, corroborating the poverty instability in Brazil, as shown by ECLAC (2021). A possible explanation for this rise is that in 2015 Brazil entered an economic and social crisis, and social indicators tend to respond to the business cycle (Vegh et al.; 2017). In addition, according to Glewwe and Hall (1998), people's poverty status tends to not respond immediately to macroeconomics shocks so that the economic crisis would take longer to reflect on the poor count (H), but it could suddenly affect the level of deprivation of the poor (A).

Figure 3:
Time evolution of the headcount ratio (H), the intensity of poverty (A), and the multidimensional poverty index (MPI) - Brazil, men, and women, 2004 to 2015

Figures 4 and Table 4 show the contribution of each dimension and indicator to the MPI. It is worth mentioning that these data differ from the ones presented before, since Table 3 displayed the percentage of people deprived in each indicator, while this shows how much each indicator and, therefore, each dimension contributed to the MPI calculation considering its weights.

The dimension/indicator that contributed most to the Brazilian MPI in 2004 was employment. Its contribution has increased to 2015's MPI primarily due to the reduction of the contribution of education and living standards from 2004 to 2015. In addition, the dimension that contributed the least was the living standards, with the electricity indicator being the one with the lowest contribution. Electricity's modest relevance to the MPI is due to the proportion of the Brazilian population's access to electricity, which in 2019 was 99.80% (IBGE, 2023). According to Freitas and Oliveira (2017), this achievement is due mainly to the "Luz para Todos" program created in 2003 to bring electricity to every Brazilian home, no matter how remotely located.

Figure 4:
Contribution of each dimension to the multidimensional poverty index - Brazil, men, and women, 2004 and 2015

Table 4:
Contribution of each indicator to the multidimensional poverty index - Brazil, men, and women, 2004 and 2015

Indicator	Weight	Contribution to the MPI
		2004			2015
		Brazil	Men	Women	Brazil	Men	Women
K=33%		%
Years of Schooling	0.125	15.40	16.90	14.30	14.00	15.90	12.80
Literacy	0.125	02.70	02.90	02.50	02.40	03.40	01.80
Employment	0.250	41.00	41.90	40.30	44.00	44.50	43.70
Income	0.250	33.70	30.20	36.30	33.80	29.60	36.40
Electricity	0.036	00.10	00.10	00.10	00.00	00.00	00.00
Sanitation	0.036	03.20	03.50	03.10	03.00	03.30	02.90
Water	0.036	00.50	00.60	00.50	00.40	00.50	00.30
Garbage Collection	0.036	00.90	01.10	00.80	00.90	01.20	00.80
Cooking Fuel	0.036	00.50	00.60	00.40	00.40	00.50	00.30
Assets	0.036	00.90	01.10	00.80	00.30	00.40	00.20
Overcrowding	0.036	01.00	01.10	01.00	00.80	00.70	00.80

• Source: Research results

The importance of the dimensions and indicators remains in the same order when considering men's and women's MPI. However, it is worth mentioning that education, employment, and living standards contribute more to men's measurement of poverty than women's, while income contributes more to women's MPI than men's. These results suggest that reducing the deprivation in employment would be the most effective way to reduce people's multidimensional poverty in Brazil, regardless of gender. Also, improving women's income would have a higher impact on the MPI than improving men's.

4.3 Gender, race, and the multidimensional poverty index

Figure 5 shows our analysis of race, presenting the headcount ratio, the intensity of poverty, the MPI by gender and race, as well as their interaction for the last year of analysis, 2015. Black people have a higher count of multidimensional poor (H), 48.70%, than white people, 38.80%. They also experience a more intensive form of deprivation (A), 57.49% against 54.64%. Therefore, Afro-Brazilian people had greater MPI values than white people, 28% versus 21.20% highlighting, once again, that Brazil's poverty phenomenon is black (Carneiro, 2015).

Nevertheless, gender appears to influence people's deprivation status more than race. The headcount ratio for white women is 48.30%, while the percentage of black women in multidimensional poverty is 59.20%, a difference of 10 p.p. However, when comparing white men and women, the difference in incidence is approximately 20 p.p. The same is valid when comparing black men and women. Hence, the gender effect on deprivation is twice the effect of race. This result aligns with the ones in Table 2, which shows that regardless of race, women have a higher deprivation score than men. As Carneiro (2015) expressed, structural racism affects every aspect of black individuals' lives in Brazil, which would lead us to expect race to have a more significant influence on people's poverty status than gender.

Still, the double jeopardy black women experience in Brazil is unquestionable and consistent with the literature (Santos, 2009; Silva, 2013; Carneiro, 2015). They have a headcount ratio of 59.20%, while white men's is 28.70%, a count of 30.5 p.p. favoring the last. This high count leads black women to have the highest MPI value in the country, 34.30%, and white men to have the lowest one, 15.30%. Black men and women experience a very similar percentage of deprivation, being slightly higher for women.

Figure 5:
The headcount ratio, the intensity of poverty, and the multidimensional poverty index - Brazil, gender, and race, 2015

Finally, regarding the contribution of each dimension to the MPI value⁸, for the white people in Brazil, regardless of their gender, the importance of income and employment dimensions to their MPI is greater than that for black individuals. It does not mean that white people are worse off in these dimensions than black people. The only reason is that for black individuals, the dimensions of education and living standards are more important. This finding is consistent with Gradín (2009), which shows that the discrepancy in poverty levels between white and black people in Brazil is mainly explained by the gap in education and the labor market.

In sum, women and black people are the most multidimensionally poor in Brazil, a fact which sheds light on the greater disadvantages black women suffer. Hence, their specific situation should be considered when designing policies to alleviate poverty.

4.4 Gender, location, and the multidimensional poverty index

Figures 6, 7, and 8 report the headcount ratio, the intensity of poverty, and MPI across the country in 2015, considering the 27 Federal Units. Figure 9 shows results separated for the rural-urban areas.

Our results are consistent: multidimensional poverty (H) is mainly concentrated in the Northeast and Northern states, both for men and women. Alagoas is the state with the highest headcount ratio for the total population (61.7%) and for women (70.9%). Its incidence of multidimensional poverty for the total population is almost twice that of Santa Catarina, where we found the lowest value (31.8%). Among men, the incidence of multidimensional poverty ranges from 21.5% in Santa Catarina to 53.5% in Ceará. Notably, even at its highest level, poverty among men is only 10 percentage points higher than the lowest level observed among women (43.1% in the Federal District), reinforcing the extent of gender disparities.

When it comes to the intensity of poverty (A) (Figure 7), Maranhão and Alagoas are where the poor people experienced the highest percentage of deprivation, with a value of 60.45%. The lowest percentage of deprivation was found in São Paulo: 53.51%. Men living in Maranhão are the ones experiencing the greatest deprivation in Brazil, 60.57%, while the ones living in São Paulo experience the least, 52.09%. In turn, women who live in Pará experience the highest percentage of deprivation in Brazil, 61.08%, and those who live in the Federal District experience the lowest, 53.92%.

Figure 6:
The Headcount Ratio (H) across the country - Total, Women, and Men 2015

Figure 7:
The intensity of poverty (A) across the country - Total, Women, and Men, 2015

As for the MPI (Figure 8), the state with the highest MPI value for the total population and for women was Alagoas (37.30% and 43.10%, respectively). On the other hand, Maranhão is where MPI was highest for men (32.10%). Santa Catarina was the state with the lowest MPI value for the total population and for men (17.40% and 11.30%, respectively) and the Federal District is where we found the lowest value for women, 22.70%. It is interesting to note that the lowest level for women is double that of men (22% versus 11%).

Figure 8:
The multidimensional poverty index (MPI) across the country - Women, Men, and Total 2015

Figure 9 shows results for individuals living in rural and urban locations in 2015. We see that the proportion of multidimensionally poor individuals (H) living in the Brazilian rural areas (62.90%) was significantly higher than for those living in the urban areas (41.20%). Also, poor individuals in rural areas were intensively poorer, with A being 62.16% for the rural, against 54.85% for the urban. Therefore, MPI is substantially higher for rural individuals (39.10%) than for urban ones (22.60%). Rodrigues (2014) also found that Brazilian rural areas present the highest poverty incidence, regardless of the poverty measurement used. We also corroborate Fahel et al. (2016) findings that the country's rural areas are expressively multidimensionally poorer than the urban ones.

Figure 9:
The headcount ratio, the intensity of poverty, and the multidimensional poverty index - Brazil, gender, and rural/urban location, 2015

Our gendered analysis shows an even wider difference between rural and urban individuals, since the proportion of multidimensionally poor (H) rural women was 43.40 p.p. larger than that of urban men in Brazil in 2015. Rural women were also experiencing the most significant deprivation. They had a poverty intensity (A) of 63.88% against 60.37% for rural men, 55.56% for urban women, and 53.90% for urban men. Therefore, the highest value of MPI was for rural women, 47.40%, which suggests that living in rural areas also acts as an additive effect on women's poverty. Silva et al. (2016) and Serra et al. (2020), also suggest that living in remote areas of the country contributes to higher multidimensional poverty.

When it comes to each dimension’s contribution to the MPI⁹, we see a disadvantage for rural women. Their education dimension has significantly less relevance to their MPI than rural men's, while their income dimension has substantially more relevance than men's. Furthermore, these gender differences are higher for rural women than for urban ones. Thus, rural women tend to have even more difficulty transforming education into payment than men and their urban counterparts. The difficulties women face in these areas reflect rural Brazil's patriarchal structure, which increases women's disadvantages inside the family and in the labor market (Ramundo Staduto et al., 2013).

To summarize, we see important heterogeneity among poverty situations throughout the Brazilian territory. Women living in Alagoas, a northeastern state, are worse off than in other states. Rural women have a double jeopardy since they suffer from the disadvantage of bothgender and location.

4.5 Sensitivity analysis

The methodology section discusses the inclusion of the employment and income dimensions. We report the Multidimensional Poverty Index for the previously analyzed subgroups to illustrate the relevance of incorporating these dimensions. To enhance clarity and conciseness, the analysis focuses exclusively on results for the final year of the period considered, namely 2015.

The MPI is constructed under three alternative specifications: one including all dimensions (education, employment, living standards, and income); a second excluding the income dimension; and a third excluding both income and employment. By doing so, we acknowledge that, within the Capabilities Approach, the inclusion of income raises the question of whether it should be treated as a means rather than an end. Nevertheless, we argue that incorporating both income and employment is essential for capturing key aspects of vulnerability in the Brazilian population, as the results presented below demonstrate.

Figure 10 shows that excluding the income dimension does not substantially alter the gender differences observed in the MPI, as women remain more multidimensionally poor than men under both specifications. This result is consistent with the argument put forward by Santos and Villatoro (2018), who emphasize that the inclusion of income does not necessarily bias multidimensional poverty measures and may instead complement non-monetary indicators. By contrast, excluding the employment dimension leads to a reversal of the gender vulnerability to poverty, with women appearing slightly less vulnerable than men. This finding resonates with the concerns raised by Nájera and Gordon (2020), who highlight the sensitivity of multidimensional poverty profiles to the treatment of labor market-related indicators. In the Brazilian context, where gender inequalities in employment, job quality, and time use are well documented, omitting the employment dimension risks obscuring a central source of vulnerability and producing a poverty profile that is less consistent with observed labor market disparities.

Figure 10:
Multidimensional poverty index calculated under alternative specifications - Brazil, men, and women, 2015

Additional sensitivity checks were performed for other subgroup analyses presented in the paper, including race, rural-urban residence and state-level comparisons. The results consistently indicate that excluding income does not substantially alter gender differences in multidimensional poverty, while excluding employment tends to attenuate or reverse these differences. Given the similarity of these patterns across subgroups, detailed results are not reported to preserve conciseness and are available from the authors upon request.

5 Conclusion

This research evaluated the gendered nature of multidimensional poverty in Brazil and its evolution between 2004 and 2015. We constructed the Multidimensional Poverty Index (MPI) with data from households surveys (PNAD’s), considering eleven indicators within four dimensions (education, employment, income, and living standards). We extensively explore the index from a race and location perspective.

Our results suggest that poverty has, in fact, multidimensional and gendered components in Brazil. First, the relevance of the dimensions evaluated, excluding income, sheds light on the lack of capabilities experienced by Brazilian individuals, highlighting the need to account for poverty's multidimensional nature. There is also an interesting connection between the capabilities and the income approaches since the income and employment dimensions were the most relevant for multidimensional poverty in Brazil.

We found that women have higher deprivation scores, meaning they have, on average, a superior proportion of deprivation than men, which culminates in a more intensive form of poverty. Additionally, women's poverty status has been improving more slowly than men's, a result that should guide the actions aimed at reducing gender gaps in the country. We also found that women have more difficulty transforming education into employment and income, and their index is more affected by the income dimension than men's.

Thus, failing to account for the differences between women's and men's situations in Brazil could lead to policies worsening gender inequalities. Brazil's public policies should focus on providing women more means to enter and develop in the labor market. Policies focused on school-aged children that encourage girls to pursue Science, Technology, Engineering and Math careers are an example. Women also need to have available childcare, and the caregiving they exercise should be accounted for since women's time poverty tends to worsen the discrepancies between theirs and men's deprivation situations.

From our race perspective, Brazil's black women are the ones who suffer the most deprivations, being at a disadvantage compared to white women. Another highlight of our results is that location matters with regard to poverty. Individuals living in Brazil's North and Northeast states and those living in the rural areas are the most multidimensionally deprived. Hence, given the country's size and heterogeneity, local public policies should consider each location's specificities. For example, extending and facilitating access to education for rural individuals can significantly alleviate their multidimensional poverty more than it would for the urban population.

In sum, poverty is a complex multi-attribute phenomenon, and, in Brazil, it has a gender, a color, and an address. Thus, neutral public policies might not have the desired effect when pursuing poverty alleviation; on the contrary, they could even widen the gaps. We believe that our findings help in the understanding of Brazil's poverty and can assist in the design of more focused public policies. However, the data available placed restraints upon our multidimensional poverty analysis, since other dimensions, such as health, are also relevant to a more detailed evaluation (providing, therefore, an opportunity for future research).

Acknowledgements

The authors acknowledge the financial support of FAPEMIG, CAPES and CNPq.

6 References

• AGUILAR, G. R.; SUMNER, A. Who are the world's poor? A new profile of global multidimensional poverty. World Development, v. 126, p. 104716, 2020.
• ALAM, A. Impact of gender discrimination on gender development and poverty alleviation. Sarhad Journal of Agriculture, v. 27, n. 2, p. 330-331, 2011.
• ALKIRE, S.; APABLAZA, M.; JUNG, E. Multidimensional poverty measurement for EUSILC countries. OPHI Research in Progress, n. 36c. Oxford: Oxford Poverty and Human Development Initiative, University of Oxford, 2014.
• ALKIRE, S.; FOSTER, J. Understandings and misunderstandings of multidimensional poverty measurement. The Journal of Economic Inequality, v. 9, n. 2, p. 289-314, 2011a.
• ALKIRE, S.; JINDRA, C.; AGUILAR, G. R.; SETH, S.; VAZ, A. Global multidimensional poverty index 2015. OPHI Briefing, n. 4. Oxford: Oxford Poverty and Human Development Initiative, University of Oxford, 2015.
• ALKIRE, S.; SANTOS, M. E. Measuring acute poverty in the developing world: Robustness and scope of the multidimensional poverty index. World Development, v. 59, p. 251-274, 2014.
• ALKIRE, S.; SANTOS, M. E.; SETH, S.; YALONETZKY, G. Is the multidimensional poverty index robust to different weights? OPHI Working Paper, n. 22. Oxford: Oxford Poverty and Human Development Initiative, 2010.
• ALKIRE, S.; SETH, S. Multidimensional poverty reduction in India between 1999 and 2006. World Development, v. 40, n. 11, p. 2288-2304, 2012.
• AVILA, R.; BAGOLIN, I. P.; COMIM, F. Heterogeneidades individuais versus intensidade da pobreza em Porto Alegre-RS. Economia (Brasília), v. 13, n. 3, p. 563-581, 2012.
• BARROS, R. P.; CARVALHO, M.; FRANCO, S. Pobreza multidimensional no Brasil. Texto para Discussão IPEA, n. 1227, 2006.
• BARROS, R.; FOX, L.; MENDONÇA, R. Female-headed households, poverty, and the welfare of children in urban Brazil. Economic Development and Cultural Change, v. 45, n. 2, p. 231-257, 1997.
• BASTOS, A.; CASACA, S. F.; NUNES, F.; PEREIRINHA, J. Women and poverty: A gender-sensitive approach. The Journal of Socio-Economics, v. 38, n. 5, p. 764-778, 2009.
• BATANA, Y. M. Multidimensional measurement of poverty among women in Sub-Saharan Africa. Social Indicators Research, v. 112, n. 2, p. 337-362, 2013.
• BATISTA, A. L.; COSTA, L V. The effects of income shocks on female headship and poverty in Brazil. Brazilian Review of Econometrics, v. 42, n. 1, p. 1-25, 2022.
• BATISTA, A. L.; COSTA, L. V. Domicílios chefiados por mulheres e pobreza no Brasil: Uma análise de 2011 a 2015. Revista Feminismos, v. 7, n. 3, p. 28-45, 2019.
• BATTISTON, D.; CRUCES, G.; LÓPEZ-CALVA, L. F.; LUGO, M. A.; SANTOS, M. E. Income and beyond: Multidimensional poverty in six Latin American countries. Social Indicators Research, v. 112, n. 2, p. 291-314, 2013.
• BENNETT, F.; DALY, M. Poverty through a gender lens: evidence and policy review on gender and poverty. Oxford: University of Oxford, Department of Social Policy and Intervention, 2014.
• BERNIELL, I.; LA MATA, D.; MACHADO, M. The Impact of a Permanent Income Shock on the Situation of Women in the Household: the case of a pension reform in Argentina. Economic Development and Cultural Change, v. 68, n. 4, p. 1389-1432, 2020.
• BERTRAND, M.; BLACK, S. E.; JENSEN, S.; LLERAS-MUNEY, A. Breaking the glass ceiling? The effect of board quotas on female labour market outcomes in Norway. The Review of Economic Studies, v. 86, n. 1, p. 191-239, 2019.
• BOBILEV, R.; BOSCHINI, A.; ROINE, J. Women in the top of the income distribution: what can we learn from LIS-data? Italian Economic Journal, v. 6, p. 63-107, 2020.
• BOURGUIGNON, F.; CHAKRAVARTY, S. R. The measurement of multidimensional poverty. In: BOURGUIGNON, F.; CHAKRAVARTY, S. R. Poverty, Social Exclusion and Stochastic Dominance New York: Springer, p. 83-107, 2019.
• BRADY, D.; BURTON, L. M. The Oxford Handbook of the Social Science of Poverty Oxford: Oxford University Press, 2016.
• BUDLENDER, D. Women and poverty. Agenda, v. 19, n. 64, p. 30-36, 2005.
• CABANISS, E. R.; FULLER, J. E. Ethnicity, race and poverty among single women: A theoretical synthesis. Race, gender & class, p. 142-162, 2005.
• CAGATAY, N. Gender and poverty. New York: UNDP, Social Development and Poverty Elimination Division, 1998.
• CARNEIRO, S. Racismo, sexismo e desigualdade no Brasil São Paulo: Selo Negro, 2015.
• COSTA, G. O. T.; MACHADO, A. F.; AMARAL, P. V. Vulnerability to poverty in Brazilian municipalities in 2000 and 2010: A multidimensional approach. EconomiA, v. 19, n. 1, p. 132-148, 2018.
• DEDECCA, C. S.; ROCHA, R.; TROVÃO, C. J. B. M. Uma abordagem multidimensional da pobreza rural segundo a perspectiva da política pública. In: MIRANDA, C.; TIBURCIO, B. (Org.). A nova cara da pobreza rural: desafios para as políticas públicas Brasília: IICA, p. 45-84, 2012.
• ECONOMIC COMMISSION FOR LATIN AMERICA AND THE CARIBBEAN (ECLAC). Femininity index of poor households Available at: https://oig.cepal.org/en/indicators/femininity-index-poor-households Accessed Feb. 2024.
  » https://oig.cepal.org/en/indicators/femininity-index-poor-households
• ECONOMIC COMMISSION FOR LATIN AMERICA AND THE CARIBBEAN (ECLAC). CEPALSTAT: Statistical databases and publications Available at: https://statistics.cepal.org/portal/cepalstat/dashboard.html Accessed Dec. 2021.
  » https://statistics.cepal.org/portal/cepalstat/dashboard.html
• ECONOMIC COMMISSION FOR LATIN AMERICA AND THE CARIBBEAN (ECLAC). Social Panorama of Latin America 2013. Santiago: United Nations, 2014.
• FAHEL, M.; TELES, L. R.; CAMINHAS, D. A. Para além da renda: Uma análise da pobreza multidimensional no Brasil. Revista Brasileira de Ciências Sociais, v. 31, n. 92, p. 1-21, 2016.
• FERREIRA, F. H. G. Poverty is multidimensional. But what are we going to do about it? The Journal of Economic Inequality, v. 9, n. 3, p. 493-507, 2011.
• FERREIRA, T. R. S.; MARIN, S. R. Pobreza multidimensional feminina: uma aplicação do método Alkire-Foster nas grandes regiões brasileiras nos anos de 2001 e 2011. Revista Pesquisa & Debate, v. 27, n. 1, 2016.
• FREITAS, G. F.; OLIVEIRA, M. L. R. DE. Uma análise do programa luz para todos do Governo Federal. Journal of Extension and Rural Studies, v. 6, n. 2, jul. 2017.
• GLEWWE, P.; HALL, G. Who is most vulnerable to macroeconomic shocks? Hypothesis tests using panel data from Peru. World Bank Policy Research Working Paper, n. 1234. Washington: World Bank, 1998.
• GRADÍN, C. Why is poverty so high among Afro-Brazilians? A decomposition analysis of the racial poverty gap. The Journal of Development Studies, v. 45, n. 9, p. 1426-1452, 2009.
• HARDY, M. A.; HAZELRIGG, L. E. Gender, race/ethnicity, and poverty in later life. Journal of Aging Studies, v. 9, n. 1, p. 43-63, 1995.
• HOFFMANN, R. Mensuração da desigualdade e da pobreza no Brasil. In: BARROS, R. P.; HENRIQUES, R.; MENDONÇA, R. (org.). Desigualdade e pobreza no Brasil Rio de Janeiro: IPEA, p. 81-107, 2000.
• INSTITUTO BRASILEIRO DE GEOGRAFIA E ESTATÍSTICA (IBGE). Desigualdades sociais por cor ou raça no Brasil. Estudos e Pesquisas - Informações Demográficas e Socioeconômicas, v. 41, p. 1-12, 2019.
• INSTITUTO BRASILEIRO DE GEOGRAFIA E ESTATÍSTICA (IBGE). Pesquisa Nacional por Amostra de Domicílios Contínua - PNAD Contínua 2019.
• INSTITUTO BRASILEIRO DE GEOGRAFIA E ESTATÍSTICA (IBGE). Estatísticas de Gênero: Indicadores sociais das mulheres no Brasil. Estudos e Pesquisas - Informação Demográfica e Socioeconômica, n. 38, p. 12p., 2021.
• JACKSON, C. Women and poverty or gender and well-being? Journal of International Affairs, v. 52, n. 1, p. 67-81, 1998.
• KAGEYAMA, A.; HOFFMANN, R. Pobreza no Brasil: uma perspectiva multidimensional. Economia e Sociedade, v. 15, n. 1, p. 79-112, 2006.
• KLASEN, S.; LECHTENFELD, T.; POVEL, F. A feminization of vulnerability? Female headship, poverty, and vulnerability in Thailand and Vietnam. World Development, v. 71, p. 36-53, 2015.
• KOSEC, K.; SONG, J. Aspirations and women's empowerment: evidence from Kyrgyzstan. Washington: International Food Policy Research Institute, 2018
• LIU, C.; ESTEVE, A.; TREVIÑO, R. Female-headed households and living conditions in Latin America. World Development, v. 90, p. 311-328, 2017.
• LUTHER, A.; GERHARDT, T. E. Educação obrigatória, êxodo rural e fechamento das escolas do campo no Brasil. Revista Saberes da Amazônia, v. 3, n. 7, p. 281-310, 2018..
• MCLANAHAN, S. S.; KELLY, E. L. The feminization of poverty. In: RISMAN, B. (ed.). Handbook of the Sociology of Gender New York: Springer, p. 127-145, 2006.
• MELO, H. P.; THOMÉ, D. Mulheres e poder. Rio de Janeiro: Editora FGV, 2018.
• NÁJERA, H. E.; GORDON, D. The importance of reliability and construct validity in multidimensional poverty measurement: an illustration using the Multidimensional Poverty Index for Latin America. The Journal of Development Studies, v. 56, n. 9, p. 1763-1783, 2020.
• NUSSBAUM, M. Capabilities as fundamental entitlements: Sen and social justice. Feminist economics, v. 9, n. 2-3, p. 33-59, 2003.
• NUSSBAUM, M. C. Women and human development: the capabilities approach Cambridge: Cambridge University Press, 2001.
• OLIVEIRA, N. S. M. N.; LIMA, J. F. Pobreza das mulheres chefes de família da região Nordeste do Brasil: uma análise multidimensional. Revista Brasileira de Gestão e Desenvolvimento Regional, v. 19, n. 3, 2023.
• RAMUNDO STADUTO, J.; NASCIMENTO, C. A.; SOUZA, M. Ocupações e renda das mulheres e homens no rural do estado do Paraná, Brasil: uma perspectiva de gênero. Cuadernos de Desarrollo Rural, v. 10, n. 72, p. 91-115, 2013.
• RAVALLION, M. On multidimensional indices of poverty. The Journal of Economic Inequality, v. 9, n. 2, p. 235-248, 2011.
• RESKIN, B. The race discrimination system. Annual Review of Sociology, v. 38, p. 17-35, 2012.
• RIPPIN, N. Multidimensional poverty in Germany: A capability approach. Forum for Social Economics , v. 45 , n. 2-3 , p. 230-255, 2016
• SÁ, F. J.; ALMEIDA, A. C.; BATISTA, A. L.; BRAGA, C.; COSTA JARDIM, S. Explorando a desigualdade de gênero nos estados brasileiros: construção e análise do Índice de Gap de Gênero. Revista Brasileira de Estudos de População, v. 42, p. 1-25, 2025.
• SANTOS, J. A. F. A interação estrutural entre a desigualdade de raça e de gênero no Brasil. Revista Brasileira de Ciências Sociais, v. 24, n. 70, p. 37-60, 2009.
• SANTOS, M. E.; LUGO, M. A.; LÓPEZ-CALVA, L. F.; CRUCES, G., BATTISTÓN, D. Refining the basic needs approach: a multidimensional analysis of poverty in Latin America. In: BISHOP, J.; SALAS, R. (eds.). Studies in Applied Welfare Analysis Bingley: Emerald, p. 205-233, 2015.
• SANTOS, M. E.; VILLATORO, P. A multidimensional poverty index for Latin America. Review of Income and Wealth, v. 64, n. 1, p. 52-82, 2018.
• SEN, A. Poverty: an ordinal approach to measurement. Econometrica, v. 44, n. 2, p. 219-231, 1976.
• SEN, A. Development as freedom Oxford: Oxford University Press, 1999.
• SERRA, A. S.; YALONETZKY, G. I.; MAIA, A. G. Multidimensional Poverty in Brazil in the Early 21st Century: Evidence from the Demographic Census. Social Indicators Research, v. 149, p. 1-36, 2020.
• SILVA, A. F.; ARAUJO, J. A.; CAMPELO, G. L.; VASCONCELOS, J. C.; SILVA, J. C. Análise da pobreza multidimensional no Brasil. Revista Espacios, v. 37, p. 17-25, 2016.
• SILVA, P. L. N.; PESSOA, D. G. C.; LILA, M. F. Análise estatística de dados da PNAD: incorporando a estrutura do plano amostral. Ciência & Saúde Coletiva, v. 7, p. 659-670, 2002.
• SILVA, T. D. Mulheres negras, pobreza e desigualdade de renda. In: IPEA. Dossiê Mulheres Negras: retrato das condições de vida das mulheres negras no Brasil Brasília: IPEA, p. 109-126, 2013.
• SKOUFIAS, E.; NAKAMURA, S.; GUKOVAS, RENATA. Safeguarding against a reversal in social gains during the economic crisis in Brazil. Washington: World Bank, 2017.
• TAVARES, F.F., BETTI, G. Gender differences in multidimensional poverty in Brazil: a fuzzy approach. Social Indicators Research, v. 175, p. 385-422, 2024.
• THORBECKE, E. Multidimensional poverty: conceptual and measurement issues. In: KAKWANI, N.; SILBER, J. (eds.). The many dimensions of poverty New York: Palgrave Macmillan, p. 3-19, 2008.
• TRANI, J. F.; KUHLBERG, J.; CANNINGS, T.; CHAKKAL, D. Multidimensional poverty in Afghanistan: who are the poorest of the poor? Oxford Development Studies, v. 44, n. 2, p. 220-245, 2016.
• VEGH, C. A.; VULETIN, G.; RIERA-CRICHTON, D.; PUIG, J.; CAMARENA, J. A.; GALEANO, L.; MORANO, L.; VENTURI, L. Effects of the business cycle on social indicators in Latin America and the Caribbean: when dreams meet reality. LAC Semiannual Report. Washington: World Bank, 2017.
• WORLD ECONOMIC FORUM. Global Gender Gap Report 2021 Geneva: World Economic Forum, 2021.

• Research data availability
  Data will be made available upon publication of the article.
• JEL Codes:
  I32, J16, J15, 015
• JEL Codes:
  I32, J16, J15, 015
• 1
  Considering the international poverty line of people living with less than US$5.5 a day.
• 2
  This dimension is essential when considering the gender-sensitive approach since women tend to be more responsible for housework and caregiving than men (Buvinić; Gupta, 1997).
• 3
  We considered the one proposed by Hoffmann (2000), which considers as poor the individual with a monthly income below half a minimum wage. Due to data available, this variable is a per capita variable, representing a household income instead of an individual one.
• 4
  Live in a household that...
• 5
  The survey was replaced in 2016 by another survey, named PNAD Contínua, the characteristics of which did not allow for its use in this study.
• 6
  Note: Income poverty line: R$ 394.00 per month (equivalent to half the Brazilian minimum wage in the reference year).
• 7
  Classification made by the World Bank (2023).
• 8
  Results not shown. Available upon request.
• 9
  Results not shown. Available upon requests.

• Responsible editor
  Lucas Resende de Carvalho (Associate Editor)
  Center for Regional Development and Planning, Federal University of Minas Gerais, Belo Horizonte, MG, Brazil
  ORCID: https://orcid.org/0000-0002-3618-3967

Data availability

Data will be made available upon publication of the article.

Publication Dates

• Publication in this collection
  10 Aug 2026
• Date of issue
  2026

History

• Received
  05 Sept 2025
• Accepted
  16 Jan 2026