1 – A review of monetary poverty indicators

6 The two main monetary poverty indicators have been criticized in feminist research work for inaccurately measuring and thus contributing to the underestimation of the poverty of women and children. [3]

7 The best known and most used indicator is that of the World Bank, which considers as poor any person with daily income lower than the IPL. The World Bank calculates this threshold based on national poverty thresholds calculated through household surveys administered by statistics institutions and government ministries. [4] The point of departure for establishing the IPL is the poverty threshold of the 15 poorest countries, converted using purchasing power parity (PPP) to simplify international comparisons. This calculation results in a threshold in US dollars, standing today at $1.90 per person per day, based on 2011 PPP conversion rates.

8 Eurostat has established a common methodology for calculating European indicators, which are then calculated in a decentralized manner for each country. [5] The Statistics on Income and Living Conditions (SILC) survey collects the data used to calculate harmonized monetary poverty rates at the European level. The Eurostat definition of poverty is a person living in a household with equivalized income [6] lower than 60% of the median income of the national population. Consequently, the income level corresponding to 60% of median income is considered the poverty threshold, which means that an individual is considered poor based on their household. This poverty rate is referred to as ‘at risk of poverty’ (AROP).

9 These two indicators have several limitations. They determine poverty levels (or AROP levels) purely based on income and fail to take account of wealth, which nevertheless plays a major role in inequalities and gender inequalities more specifically, both in Northern countries (Bessière and Gollac, 2020) and Southern countries (Deere and Doss, 2006). Also, to measure poverty in terms of income, they are based on the concept of the household. Since the Treaty of Amsterdam and the Luxembourg Summit in 1997, gender mainstreaming calls for all indicators to be produced by sex (Fouquet, 2003). Regarding poverty, as the calculation was originally based on households, and because most households include men and women, the breakdown by sex unsurprisingly gives similar poverty rates for both sexes. [7] This would suggest that men are almost as concerned as women by poverty, or, stated differently, that poverty is not gendered.

2 – Feminist criticism: the problem of household-based measurements

10 Feminist criticism was quick to highlight the lack of realism of these official poverty measurements. It noted the paradox in wealthy countries surrounding the ‘working poor’, the official measurement of which failed to specifically indicate whether they were female workers, which contradicts knowledge on the general situation of women on the labour market (Ponthieux, 2004; Oxfam, 2018). Feminist research identified the theoretical and methodological root of the problem, namely that female poverty is underestimated because measurements are made based on households (Ponthieux and Meurs, 2015), considered in Beckerian or ‘unitary’ household approaches (Becker, 1981) as the elementary unit of society. This approach models household behaviour as the maximization of a utility function common to all household members.

11 This approach is problematic because it assumes the complete pooling of income within the household and equal access to income by all household members (Bennett, 2013). Yet the total income of a household is not always pooled. Ponthieux (2012) showed that just 64% of the couples in her French sample pooled their income (2,349 couples having lived together for at least a year responding to a special module, ‘Decision-making in couples’, in the 2010 Time Use survey). Cantillon et al. (2016) estimated this proportion at 59% for couples in Ireland (2,419 couples responding to a special module, ‘Intra-household sharing of resources’, in the 2010 SILC survey). Access to the resources themselves is not a guarantee against poverty since it does not always imply the possibility of making free use of the resources, for oneself or for one’s children (Delphy, 1971; Sen, 1984). Consequently, attributing to women some of the income of men by assuming that this income is shared equally within households and that each member benefits from it in the same way distorts the actual situation of women by failing to measure internal household inequalities. Haddad and Kanbur (1990) have shown in the southern Philippines that the measurement of total inequalities would be increased by 30% –40% if internal household inequalities were taken into account. [8]

12 Until now, this criticism has not led to a revision of official poverty indicators. But it has opened the way to extensive feminist research exploring the possibilities for measuring the poverty of women in a different way by proposing alternative indicators purported to reveal the true scale of female poverty. These alternative approaches include acknowledging that poverty is not strictly monetary (Fukuda-Parr, 1999) and individualizing income and monetary poverty indicators (Meulders and O’Dorchai, 2011; Corsi et al., 2016). Some have combined these two approaches by devising individual poverty indicators addressing the non-monetary aspects of poverty (Cantillon and Nolan, 2001). This article reviews these initiatives, looking first at alternative indicators of monetary poverty proposed for Southern countries (the poverty of households headed by a woman) and then at those developed for European countries (individual indicator of monetary poverty). Ultimately, research on the accurate measurement of female poverty has led across the board to the conclusion that such a measurement is not possible without addressing internal household aspects.

1 – Female poverty in development policies and discourse

14 Until the 1990s, the discourse and actions of international institutions were informed by the ‘Washington consensus’. [9] In the 1990s, the failure of structural adjustment plans led these institutions, along with numerous NGOs, to establish a new paradigm of analysis and action against poverty (Prévost, 2011), based notably on the idea of ‘social development’ as a new discourse on poverty, accompanied by the concepts of vulnerability, human capital, risk, and empowerment. This new discourse turned the spotlight on female poverty, as seen at the Fourth World Conference on Women held in Beijing in 1995 (Prévost and Palier, 2007), where it was asserted that 70% of poor people worldwide were women (according to an estimate proposed by the UN at the time and later challenged). This observation, along with that of the ‘feminization of poverty’, whereby not only do women account for most poor people but more and more poor people are women, became the new orthodoxy of the discourse on poverty (Chant, 2010b). This led to the specific targeting of women in policies on the fight against poverty, as it appeared that combating female poverty reduces adult poverty in general, along with that of children. Hence the emphasis placed on the development of microfinance, aimed at simplifying access to loans for women and seen as a major source of female empowerment (Prévost, 2011). Women are specifically targeted by the microcredit projects encouraged by international institutions as part of the promotion of state disengagement programmes (Guérin, 2015), especially since they are considered particularly reliable borrowers (Garikipati, 2010).

15 This new approach of empowerment has been criticized in several respects. It is seen as a philosophy that considers women as individuals who should be raised to the economic level of men through equally individual resources (Chant, 2010a). It is criticized as having been designed for women but without the involvement of feminists ‘on the ground’ and always under the official supervision of the United Nations, the institution that organized both the Conference on Women in Beijing and the NGO Forum (Falquet, 2003). The approach has also been criticized for focusing on promoting the education of girls rather than improving the situation of women in general, with a view to developing ‘human capital’. Kabeer (2005) has shown that the attention paid to women’s empowerment has led international institutions to place the emphasis on a limited number of systems and indicators with no consideration of the social relations to which these systems were to be applied, which has led to the relative failure of empowerment policies. Broadly speaking, the neoliberal approach to the empowerment of women has reinscribed logics of domination, albeit through new mechanisms, more than it has eliminated or even attenuated them (Guérin, 2015).

16 This discourse has been reflected in the increased targeting of women in development programmes, chief among them the conditional cash transfers promoted by international institutions (Debonneville and Diaz, 2013). These programmes (including Bolsa Familia in Brazil and Progresa/Oportunidades in Mexico) condition the eligibility of women for financial aid on the education of children and the respect of health and nutrition conditions. They lend women a central role in the fight against poverty as both the recipients of the policy and vectors of social development, as it is through women that action is taken for children (Coelho de Souza Lago et al., 2014). Ultimately, the aim of these programmes is the empowerment of children through the development of their human capital rather than the reduction of poverty in the short term (Bradshaw et al., 2019). They have been criticized because they saddle women with new responsibilities but only moderately improve their situations (González de la Rocha, 2010).

2 – A new indicator quick to attract criticism

17 In scientific terms, the instrument developed to measure the poverty of women initially consisted in measuring the phenomenon in single-adult households led by a woman, referred to as female-headed households, or FHHs. This status is interpreted as evidence of the lack of a man in the household, in turn seen as a logical factor of extreme poverty (Buvini et al., 1978; Buvini and Gupta, 1997).

18 Initial empirical work on specifically measuring the poverty of FHHs concerned Southern countries, generally using the World Bank’s IPL as an indicator. Extensive research has explored this approach, notably comparing households led by a woman with those led by a man (Marcoux, 1998; Quisumbing et al., 2001). But this work was quick to raise objections as to the relevance of the method.

19 First, it soon became clear that FHHs are not necessarily poorer than other households. The phenomenon is not uniform geographically, with FHHs more likely to be over-represented among the poor in Asia and Latin America than in Africa (Kabeer, 2003; Onu and Abayomi, 2009). Several studies have shown that, in Latin America, despite FHH targeting in new policies on the fight against poverty, the poverty of women has continued to increase (Chant, 2009; Bradshaw et al., 2019).

20 Secondly, the assumption that FHHs are poorer than other households is based on the idea that the presence of a female household head mechanically implies the absence of a man. This reflects what has long been the default norm of statistical surveys, whereby a household is classified as an FHH if and only if it does not include a man. But more and more women, unprompted, say they are the household head, even when they live with a male partner, especially in Latin America (Villarreal and Shin, 2008), to the extent that UN Women introduced the new ‘Female Only Household’ classification in its Progress of the World’s Women 2015–2016 report to refer to households not including a man. Similarly, FHHs have been wrongly considered single-parent families when in truth they are households in which several family units cohabit, and sometimes as part of multi-generation households in which the grandmother is the sole provider of income (for example, see Schatz, 2007 for South Africa).

21 In addition, studies of poverty based on FHHs are not representative of female poverty in general as they account for only a limited share of the population. Safa (2010) estimates that FHHs account for one-third of households in Cuba and around 27% of households in Puerto Rico and the Dominican Republic. Focusing on these households has even been interpreted as a way of not addressing the situation of all women, considering that only women living without a man were at risk of poverty. This criticism underlines the existence of female poverty among other households, resulting from the unequal sharing of income and unequal access to resources, the reasons for which are to be found in social relations.

22 Lastly, while the income of women living alone is often lower than that of women living with a partner, the former may offset the situation through better access to and control of the income of her household. That women have access to a household’s resources does not necessarily mean they can use those resources as they like. A distinction needs to be made between budget management and the possibility of spending income freely (Bradshaw et al., 2017). In the poorest households, women are often responsible for managing the budget but unable to spend income on their own consumption or that of their children, even where they contribute to the household’s income through salaried work (Kabeer, 1997). This shows that the social relations within heterosexual couples can prove decisive, including in terms of monetary poverty, sometimes depriving women of access to resources available to the household. In other words, the concept of power is essential to understanding household dynamics (Kabeer, 1999).

23 Because of this last criticism, the FHH-focused measurement tool was abandoned in favour of a more qualitative approach to poverty in Southern countries not exclusively centred on income. The most damaging criticisms, then, were not those concerning the low representativeness of FHHs but those challenging the relevance of a purely monetary approach to poverty. This opened the way to a more qualitative and multidimensional definition of poverty, encompassing aspects of free time, education, leisure, and restricted access to property and inheritance. The problem stems not just from the bias resulting from a unitary approach to the household, and changing the calculation method for accurately measuring female poverty will not suffice. Instead, greater light needs to be shed on the effect of gender relations on the sharing of and access to internal household resources.

24 Work based on the concept of FHHs and research on the impact of programmes uniquely targeting single women have ultimately shown that this method is insufficient, as it concerns a limited part of the population and fails to address the poverty experienced by women in households where their access to and use of resources are controlled by a male partner. In contrast, we need to develop an indicator that includes the measurement of the income of the entire female population and is not blind to internal household dynamics.

1 – An individual indicator of monetary poverty in European countries

26 A few recent studies on European countries have proposed a different method for clearly showing the poverty and income levels of women, consisting of the individualization of poverty rates (Belgian Gender and Income Analysis [BGIA], 2011; Meulders and O’Dorchai, 2011; Corsi et al., 2016; Statbel, 2019). This work draws on data from SILC surveys and recalculates poverty rates by individual. The resulting poverty rate provides the proportion of the population that would be poor if it were required to live purely on its own income (see BGIA, 2011), thus removing the hypothesis of the pooling of income within households. Individuals in situations of individual poverty are those whose net individual income is lower than 60% of individual median income. In most of the research, this rate is called the ‘financial dependence rate’. It is preferable to use the term ‘individual poverty risk’ as the reference to financial dependence (as made by Corsi et al., 2016) implies that the absence of individual income is offset by dependence on the income of others. But this is not necessarily the case, and individuals in situations of individual poverty may simply be poor regardless of other members’ presence or income. ‘Individual poverty risk’ is therefore more general than ‘financial dependence’. In addition, the methodology and data used to calculate individual poverty risk rates are the same (with disaggregated income) as those used to calculate the poverty risk of the household. [10]

27 In studies using the individual poverty rate, the calculation of individual rates is considered a satisfactory method because it produces results for the entire population, unlike the FHH method, which is limited to a small proportion of the population. Corsi et al. (2016) observed that single-adult households account for just 18% of the EU-27 population, so the number of FHHs would have been low for these countries.

28 The methodology used systematically to calculate individual poverty rates was as follows. First, gross and net income [11] were disaggregated at individual level, attributing to each household member the income that can be identified as their own. Non-individualizable income, i.e. that common to the household members, was then shared between them (for example, income from assets and transfers between households were shared evenly between the adults of the household, and family allowances were shared evenly between the parents). These individual incomes having been established, an individual poverty rate may be determined.

29 This methodology generates higher rates of female poverty than the official indicators. This results from the generally unfavourable situation of women on the labour market [12] (women are less often on permanent contracts than men, have lower income than men for the same job position, and work part-time more often; see Maruani, 2017), which means that their own wage incomes are lower down the scale. However, in 2007 as in 2017, all the incomes of women were on average lower than those of men, including income from assets and social transfers. In Belgium, men received on average 1.58 times more net income than women in 2007 and 1.39 more in 2017 (Statbel, 2019). Table 1 summarizes the results for Belgium included in the Statbel report (2019), which provides individual poverty rates disaggregated by sex as well as conventional rates calculated per household by Eurostat for Belgium, based on the 2007 and 2017 SILC surveys. [13] These rates are comparable, even though they do not take account of household size (the individualized rates are not equivalized), providing we assume children have the same effect on the poverty of men and women. [14]

Table 1. Household poverty rates and individual poverty rates for Belgium (in %)

Household poverty rate (Eurostat) Individual poverty rate Men Women Men Women SILC 2007 19.9 23.1 11 36 SILC 2017 14.9 16.9 13 28

Table 1. Household poverty rates and individual poverty rates for Belgium (in %)

30 The figures show an underestimation of female poverty in the Eurostat measurement. These results are particularly striking given that Gradin et al. (2010) have demonstrated that Belgium is one of the European countries with the least differences in individual characteristics between men and women. Corsi et al. (2016) arrived at similar results on the underestimation of female poverty, obtained by calculating the individual poverty rate based on the SILC survey, but this time for all the EU-27 countries, from 2007 to 2012 (see Figure 2 in Corsi et al., 2016).

31 Individualized indicators can thus be harnessed to highlight the suspicions of feminist criticism, namely that the official indicators underestimate female poverty. But the individual indicator also has its limits. As this methodology has (to my knowledge) yet to be critiqued, the following section will do just that.

2 – A critique of individual indicators

32 Poverty indicators, like all indicators, are conventional constructions underpinned by various hypotheses (Laderchi et al., 2003). The individual indicator is based on a different perception of poverty than that underlying the Eurostat indicator. This is because it considers as poor any adult unable to live off their own income, while the Eurostat indicator assumes that dependence on the access to the income of another household member is not a factor in poverty risk. However, while the Eurostat indicator contains a bias that shrouds the poverty of women, the individual indicator also skews the measurement of poverty by limiting the definition of the latter to purely individual income (wages, income from capital, and individual share of income from social transfers), i.e. by assuming that individuals never have access to any part of the income contributed by another household member. This indicator accentuates the trend of purely taking account of income without considering the broad range of social factors conditioning the effective access to resources. The pooling of resources in a household is one such factor, as are the possession of capital and the existence of family transfers from without the household. Furthermore, an individual measurement of poverty leads to paradoxical results concerning the poverty of women. For example, a single-parent household composed of a woman working full-time on minimum wage with a child is not considered poor (owing notably to an approach to poverty in relative terms, with the woman’s income, while modest, placing her in the upper reaches of the rankings of the individual income of women and above the general poverty threshold, which is much lower with individual indicators because of the large number of women with very low income). In contrast, a woman not working and living with a salaried partner with comfortable income will be considered extremely poor.

33 This effect can be seen in the results of the Statbel report (2019). Part 3 in the report shows that the individual indicator and the Eurostat indicator arrive at different conclusions on poverty (individual but not household, or household but not individual) for 17.8% of individuals. But the two definitions lead to a difference in category for just 12.5% of men compared with 22.8% of women. Men are less affected by individual poverty (12.8%) than by household poverty (13.9%), but overall, the two definitions of poverty generate roughly the same figures for men. Women, however, are more affected by individual poverty (27.6%) than by household poverty (15.9%). The proportion of women not considered as individually poor but as living in a poor household is almost identical to that of men who are not poor individually but belong to a poor household (5.5% vs. 6.8%). But the proportion of women in situations of individual poverty in a non-poor household is 3 times higher than that of men (17.3% vs. 5.7%). The concealment of individual poverty through the measurement of households mainly impacts female poverty. The underestimation of the latter stems from women who are not considered poor because they have access to income other than their own. Unlike what the schematic example above suggests (non-working woman whose husband has a comfortable income), this conclusion in no way predicts the economic level of the household in which an individually poor woman lives. The latter may escape individual poverty because her husband’s income makes up for her lack of income or because her own income is too low to exceed the poverty line but is supplemented by the similarly low income of her partner.

34 Ultimately, the individual poverty indicator serves more to assess the extent to which an individual may be financially independent through their own income (including social transfers ‘owed’ to the individual). Financial independence can reasonably be considered an aspect of poverty, with the impoverishment of women following a separation having been documented (Jarvis and Jenkins, 1999; BGIA, 2011; Bessière and Gollac, 2016, 2020), but it does not suffice to qualify all the aspects of monetary poverty.

1 – Current research on monetary poverty indicators

36 Work is now combining the two approaches mentioned above, with the development of multidimensional and individualized indicators. These indicators generally exclude the monetary aspect of poverty and thus fall outside the scope of this article. But they can also be seen as a synthesis of the contributions of the two feminist approaches discussed earlier. Notable examples include the indicators derived from the United Nations Development Programme’s Multidimensional Poverty Index (Alkire and Foster, 2011), aimed at developing an individualized version rather than a household version. Harnessing available individual data on important aspects of deprivation (mainly education and health) and building individual data on living standards (access to common household property, living space in a dwelling, ability to take financial decisions, etc.), these indicators provide estimations of multidimensional individual poverty by sex. This work has generally focused on Southern countries, with Klasen and Lahoti (2016) on India and Espinoza-Delgado and Klasen (2018) on Nicaragua. The Australian National University is currently hosting a research project (together with the Australian government and the International Women’s Development Agency) aimed at developing a new deprivation indicator. The ‘Individual Deprivation Measure’ is both individual and multidimensional and measures deprivation by sex based on 15 aspects, excluding income. [15] The indicator is still in the experimental phase, and the data collected thus far concerns just a handful of Southern countries, but the protocol will come into wider use once finalized and validated. Rather than monetary poverty, these indicators focus on gender inequalities understood through the multiple dimensions of poverty.

37 Returning to monetary poverty, one of the results of feminist research, as seen earlier, is that a monetary poverty indicator that does not underestimate female poverty should include factors on how income is shared and accessed in the household (or the related consumption expenditure). Several studies have explored how the unequal sharing of resources within a household affects inequalities (for example, Chiappori, 1988; Davies and Joshi, 1994; Kanbur and Haddad, 1994; Crespo, 2017). However, owing to the lack of available data, this work is based on alternative theoretical hypotheses or simulations of women’s access to household resources, or both. As with the indicators presented in Sections II and III, this research has sought to respond to the critique of the unitary perception of the household by circumventing the issue of data availability. To take things further, the reality of income sharing in households now needs to be studied. While generating greater complexity, this reality needs to be addressed so that it can be factored into official measurements of poverty. As stated by Bradshaw et al. (2019), information on resource sharing in households must be collected rather than modelled. The question, then, is how to explore what happens inside households and how to use the results of this exploration to build a better indicator of monetary poverty.

2 – Two avenues of research

38 Information on what actually happens in households (or couples) is not entirely lacking. A burgeoning literature at the nexus of sociology and economics is studying the issue of resources, access to resources, and how their use is controlled. The poverty or material deprivation of the various members of the household differs according to whether they have individual income and their ability to control its use. Similarly, access to the income of the other household members is not guaranteed, and even when it is the case, we need to know whether this access involves the limited management or entirely free use of this income, for oneself or for others (mainly children). Extensive research has taken the direction initiated by Pahl (1989), exploring what Cantillon et al. (2016) dubbed the ‘financial regime’ of households, i.e. an arrangement characterized by type of income, type of income pooling, and the way access to this income is organized (p. 462). This work has served to empirically invalidate the Beckerian assumption of the household as unit. In many cases, the pooling of resources is partial rather than total (Ponthieux, 2012; Cantillon et al., 2016), the sharing of resources does not always entail equal access to the household’s income (Vogler and Pahl, 1994; Cantillon et al., 2016), and not all household members benefit in the same way from common resources (Phipps and Burton, 1995; Lundberg et al., 1997). But it is difficult to go beyond these three consensual factors and deliver, through this literature, clear and unequivocal results on the scale of income pooling in households and the management and use of income (Cantillon, 2013). These studies also differ substantially in their theoretical roots, analysis method, and data, [16] to the extent that the results are difficult to synthesize, even harder to generalize, and sometimes contradictory. [17] Nevertheless, they have demonstrated that forging an in-depth understanding of how resources are used and shared in households necessarily hinges on the consideration of social relations. Neither having individual primary income nor being responsible for the daily management of household income ensures that this income is available for oneself and one’s children (Kabeer, 1997; Pahl, 1989; Bessière and Gollac, 2020).

39 Two options can thus be envisaged for building a monetary poverty indicator based on the empirical reality of access to resources in households.

40 The first is to take a ‘fully empirical’ approach by including questions on resource sharing and access in data collection. This approach has already been tested in a specific module in the 2010 SILC survey entitled ‘Intra-household sharing of resources’, designed to better understand the economic behaviour of couples, particularly as it relates to the sharing of resources, the choice of expenditure, and the use of time. This module has given rise to several studies on the financial regimes of households but until now has not been reproduced. This is because numerous technical obstacles render the collection of these data particularly difficult, both in itself and for establishing conclusions that can be used in the calculation of poverty. First, access to the creation of questionnaires is required, and the qualitative data obtained needs to be converted into quantitative data (Atkinson, 2019). It is also impossible to rule out perception biases depending on which household member responds to the questions. This may generate variability in the results according to the household member filling in the questionnaire, or it may entail increasing the number of respondents for the same household (Ponthieux and Meurs, 2015). But these obstacles are not insurmountable, and neither are they specific to this survey. Several surveys based on this type of reported data (such as INSEE’s Household Wealth and Time Use surveys) have proven to be of interest for scientific research and feminist research in particular. INSEE’s Time Use survey has been used to quantify housework and estimate its value to show that it involves more hours than occupational work and corresponds to at least one-third of gross domestic product (Chadeau et al., 1981; Roy, 2012). The obstacles to this survey type are not simply technical; they also stem from the need to work in close collaboration with data collection and processing institutes, the ease or difficulty of which varies according to the context (independent or non-independent institutes, with or without a research mission, etc.).

41 The second option consists in using non-exhaustive empirical results to devise (socio-demo-economic) categories of households for which a standard resource-pooling profile can be identified. The sole point of agreement on the financial regime of households currently emerging is that a single financial regime does not exist and that the ways household members share and access resources differ considerably across time and space, and by social category. Because practices diverge, there can be no single way of integrating the pooling of income at household level. Without going so far as to take an entirely empirical approach, investigations could be fine-tuned to establish socio-demo-economic patterns based on which households could be categorized according to their standard profile. Each type of household could then be assigned a resource-pooling coefficient used to distinguish the individual poverty of each household member. The coefficients would have an empirical basis, though without fully reflecting the reality of the practices of each household. As with today’s method of weighting income on an equivalized basis to take account of the economies of scale made by households, we could also weight individual income through a resource-pooling coefficient reflecting the household’s profile.