Population People

Posted June 1997

The Feminization of Poverty:
Facts, Hypotheses and the Art of Advocacy

by Alain Marcoux
Senior Officer, Population and Environment
Population Programme Service (SDWP)
FAO Women and Population Division


The world's population of poor is commonly estimated at 1300 million persons (UNDP 1996: 20; ICQL 1996: 18). Women, especially in developing countries, bear an unequal share of the burden of poverty; an oft-repeated statement in this respect is that 70 percent of the world's poor are women (e.g. UNDP 1995: iii; United Nations 1996a: 6).[1]

No scientific study, however, is cited to document this exceptionally high ratio, and a statistician or demographer cannot help harbouring doubts about its validity. With a 70/30 distribution, the global population of poor would comprise 910 million women and girls and 390 million men and boys, a 2.33 female/male ratio, and an 'excess' of 520 million female members. But although this is tall enough, there actually is more to it because children are hardly affected by the two phenomena that may contribute to the excess of females in poor households, namely male out-migration and excess male mortality.[2] The excess number of females is therefore concentrated in older age groups. Now, taking the age structure of the 1300 million poor to be similar to that of the whole population of the low-income countries, they comprise slightly more than 60 percent - about 800 million persons - aged over 15 (henceforth 'adults'). A 520 million sex imbalance will then translate into 660 million female and 140 million male adult poor, with a female/male ratio of about 4.7, twice the ratio based on the total population. It does not seem that an imbalance of close to five females for each male among the adult poor has ever been observed on any significant scale.

Good observations are scarce indeed, and solid statistical information on the reality of the gender bias in poverty is lacking. Large-scale assessments of poverty, for a start, only occasionally rely on household surveys, the correct instrument for assessing levels of living and hence poverty. As a result, "much of the analysis of poverty and gender rests on assumptions and inference from very limited data and case studies" (United Nations 1995a: 129), a situation which fostered the circulation of 'guesstimates' of uneven quality.

Yet the information analysed during the 1970s did not point to large sex differentials in the incidence of poverty.[3] Could the feminization of poverty have progressed so quickly as to produce a 70/30 sex distribution in two decades? This idea is by no means widely accepted among scholars, and some recent assessments argue that women are not generally over-represented in poor households (Lipton and Ravallion 1995). It still is widely assumed "that women are disproportionately represented among the poor", but there is little "robust evidence" to support that assumption (Quisumbing, Haddad and Peña 1995: 1). Let us now see what some recent field observations do tell us.

Survey data

In a study commissioned by the UN Statistical Division (United Nations, 1995a: 129-130), data from household surveys were compiled and analysed by the International Food Policy Research Institute (IFPRI). The sex distribution of the population was thus assessed in the poorer households of 14 developing countries (Bangladesh, Botswana, Côte d'Ivoire, Ethiopia, Ghana, Guatemala, Honduras, Indonesia, Madagascar, Nepal, Niger, Pakistan, the Philippines and Rwanda) and eight developed countries (Australia, Canada, Germany, Italy, the Netherlands, Sweden, the United Kingdom and the United States). "Poorer households" for this purpose meant those in the lowest quintile for income; this definition is appropriate for us, as we refer here to an estimated population of poor of 1300 million, or currently 22 percent of the global population. The observations regarding the female/male ratio were the following: In brief, where adequate data are available the average proportion of women among the poor is lower than 55 percent. It could be higher in places and countries not included in the study (it could be lower too), but scientific correction dictates that in the absence of data no claim can be made on the situation in those areas.

On the other hand, the numbers of women poor seem to grow faster than those of men poor. Could this make possible a 70/30 ratio today? Jazairy, Alamgir and Panuccio (op.cit.) estimated the female/male differential growth rate at 1.2 percent, i.e.. 3.9 minus 2.7 percent among the rural poor of developing countries between 1965-70 and 1988 (at the latter date the proportion of women among those poor would have been 60 percent [5]). A continuation of the said differential (at lower rates, to be consistent with the estimated 1300 million total number of poor world-wide) would have brought the proportion of women in that same population to no more than 62 percent in 1997. This could have raised the global proportion of women poor by at most 1.5 percentage points, far from the level of 70 percent.

Let us nevertheless examine whether a 70/30 distribution could conceivably have been true, for this exercise will teach us a lesson about using illustrative figures for advocacy.

Considerations on demographic consistency

An explanation commonly given for the supposed excess of female poverty is the occurrence of poor, women-headed households (henceforth PWHH), inasmuch as those households comprise more female than male members. Let us examine to what extent this can suggest a 70/30 sex distribution of poverty at the global level.

An estimate of the number of women-headed households by region is given in Table 1 (all estimates will refer to the year 1995, when the 70/30 slogan was already widely publicized). Average household sizes and proportions of women-headed households are estimated as of 1990. It is sensible to assume that values have not changed enough between 1990 and 1995 to affect the calculation significantly. Proportions of women-headed households are on the rise, but so are average household sizes in several developing areas (MacKellar et al. 1995: 851); these two factors tend to compensate each other where they apply. On the other hand, there often is some under-registration of women-headed households, especially in developing countries. Someone seeking a maximum estimate of the number of women-headed households world-wide, may want to take it to be 25 percent (instead of 19 percent) of the estimated total number of households in numbers, 355 million.

Table 1. Estimation of the numbers of women-headed households by major region, 1995
Region (a) Population 1995 (millions) (b) Average household size (c) Millions of households = (a) : (b) (d) Proportion women- headed Millions of women-headed households = (c) x (d)
North Africa 158 5.7 27.7 0.13 3.6
Sub-Saharan Africa 561 5.1 110.0 0.20 22.0
Eastern Asia 1421 3.7 384.1 0.21 80.7
Southeast Asia 482 4.9 98.4 0.13 12.8
South/Central Asia 1367 5.7 239.8 0.10 24.0
Western Asia 168 5.1 32.9 0.12 3.9
Latin America 441 4.7 93.8 0.21 19.7
Caribbean 36 4.1 8.8 0.35 3.1
Oceania 28 4.9 5.7 0.17 1.0
Developed countries 1171 2.8 418.2 0.24 100.4
World 5687 (4.0) 1419.4 (0.19) 271.2
Sources: (a) United Nations 1996c; (b) United Nations 1995a, and MacKellar et al. 1995; (d) United Nations 1995a (sub-regional figures are either as published in the document or calculated from country data presented therein).

One should now estimate the proportion of PWHH among the women-headed households. First of all, what is the total number of poor households? One must divide the total number of poor by average household sizes separately for developing and developed countries, in view of the very large differences in household sizes and in the incidence of poverty. With regard to the latter factor Table 2 reflects the common notion that, given the far greater population size and proportion of poor, absolute numbers of poor are overwhelmingly concentrated in developing regions; the figures correspond to a level of poverty incidence in these regions about treble that in developed ones (27 percent to 9 percent).

Average household size according to Table 1 is 4.7 in developing countries and 2.8 in developed countries. The average size of poor households, however, must be different: "poverty risk is almost always ... much greater among members of big households. Conversely, single-member households ... are heavily under-represented among the poor" (Lipton 1988: 39). Accordingly, and in line with empirical findings (e.g. Lopez, Pollack and Villarreal 1992: 89, 126) Table 2 assumes that poor households of developing countries have 0.5 more units on average than total households. For developed countries one may use a uniform household size as "both the predominance of big households ... and the scarcity of small ones ... among the poor are much less clearly established" (Lipton, ibid.).

Table 2. Estimation of the numbers of poor households in developing and developed countries, 1995
Number of poor (millions) Average household size Poor households (millions)
Developing countries 1200 5.2 231
Developed countries 100 2.8 36
World 1300 (4.9) 267

Finally one must distribute all households into the four categories resulting from the combination of the two criteria: {men-headed/women-headed} x {poor/non-poor}. Women-headed households do seem more vulnerable to poverty than men-headed ones, but where comparable data are available the actual difference in poverty incidence is not very great (see Annex).[6] Although the evidence points to a more modest bias, in order to supply a maximum estimate of female poverty one can assume, like for Table 3, that the incidence of poverty in women-headed households is 50 percent greater than in men-headed ones (over 25 percent to less than 17 percent).[7]

Table 3. Estimation of the numbers of poor and non-poor, men-headed and women-headed households, 1995 (in millions)
Men-headed households Women-headed households Total
Poor 177 90 267
Non-poor 888 265 1153
Total 1065 355 1420

The problems inherent in explaining excess female poverty at the level of a 70/30 ratio on grounds of the PWHH phenomenon become rapidly apparent:

Can other hypotheses account for the 'remainder' of the gap? The question seems futile in view of the magnitude of the unexplained difference: well over 400 million persons. At any rate the only hypothesis seems to be the occurrence of poor women in non-poor households, e.g. living-in maids. As no statistical information is available on such patterns of occurrence of poverty, one can only ask: could those women possibly be so numerous? The answer is clearly 'no': women working in the "Community, social and personal services" branch world-wide appear to be less than 120 million [9] (and part of those only would fall into the category of relevance to us). These figures might underestimate the phenomenon, in particular as regards young girls; for the assessment of total female poverty, however, this would be irrelevant: the higher the number of girl poor in non-poor households, the lower that number in the households of origin, and the lower the female/male ratio in the latter.

The use of different parameters, within reasonable ranges, for the preceding calculations invariably leads to the same conclusion: given the huge numerical inconsistencies encountered, there is not a feasible scenario to support the 70/30 slogan. It would have been advisable to test the likelihood of the latter in order to arrive at a more credible figure, which would not have unnecessarily cast shadows on the seriousness of the issue at hand, for if a cause for scepticism is offered, one may end up doubting the very existence of the phenomenon under examination.

Information and data needs

That the gender bias in poverty does not reach the very high levels sometimes attributed to it does not mean that the bias is not real or not growing. Indeed it seems to be both, although very unequally across countries and places. The first need appears to be to continue to document its magnitude and trends in a larger number of settings than has been done so far, to be able to address it where it exists. In so doing, attention to methodological issues will be warranted. It has been ascertained that female household headship is a heterogeneous phenomenon, and that looking into the causes for female headship is extremely relevant in studying poverty (Quisumbing, Haddad and Peña 1995: 25-26). Also, in a policy perspective it is necessary to use additional methods of assessment besides those based on income, for gender biases and their causes may emerge more clearly through approaches that favour social indicators (mortality, health and nutrition, time allocation) and aim to assess individual capability factors (access to resources, level of education etc.).

It might be appropriate to give special attention to the poorest segments of population, since the bias against women sometimes appears to increase along with the degree of poverty (Lipton 1983: 48). Inasmuch as priority action within poverty-alleviation policies ought to address the poorer of the poor, any systematic bias in the composition of the population concerned may suggest targeted interventions and possibly help define them. Hence the value of assessing not only the differential incidence of poverty, but also its gender-specific causes. The rural-urban dichotomy, inter alia, will be of obvious utility here.

More empirical research is also needed in respect of another aspect of gender biases in poverty issues, namely intra-household inequalities in terms of welfare and control of resources. With reference to the discussion on sex imbalances within the poor population, however, let us note that intra-household imbalances can hardly affect the sex distribution of poverty: they would have an impact at this level only if there were households where female members are below the poverty line while male members are above it. There is no known measurement of such patterns.

Of course, the above detracts nothing from the policy relevance of this issue. Some of the more publicized problems, such as gender biases in food consumption, or in-house health care, seem to have been overstated (United Nations 1996b: 13-14; Haddad et al. 1996: 5-22). On the other hand, education (even though gaps are narrowing), or the control of productive assets, remain real issues; and they are critical for strategies aiming at accelerating development as well as rendering it more equitable (Lipton 1988: 44-45; Quibria 1993: 7, 13-19). Better data coverage is needed, to assess those biases and their changes over time.

An indirect but significant benefit of more rigorous fact-finding will be to provide more relevant and convincing materials to be used in making decision makers, as well as the general public, more aware of the magnitude and exact nature of these important issues.


1. Not all UN bodies use these figures; according to UNIFEM (1995: 7), "women constitute at least 60 percent of the world's one billion poor".

2. For developing countries as a whole, the probability of dying by age five is, on average, only two percent higher among males than among females. Male mortality is higher during the first year of life, but thereafter the reverse is true (United Nations 1996b).

3. See the evidence reviewed in Lipton (1983: 48-53).

4. Source for estimates of the rural poor population: Jazairy, Alamgir and Panuccio (1992: 404-405). The result is virtually the same if country indices are weighted by total population.

5. A 60 percent level for rural areas could be roughly consistent with the UN-IFPRI results. But Jazairy, Alamgir and Panuccio certainly overestimated female rural poverty, for they assumed: (a) that all women-headed rural households are poor (op.cit.: 470); (b) that no male absent from those households lives in other poor rural households, since the female/male ratio in these is taken to be 1 (ibid.); and (c) that women constitute 75 percent of the population of women-headed households. The latter assumption (which is not made explicit) is the most misleading: children represent about 40 percent of the population of poor households, and an even higher proportion in poor women-headed households because of absent adults; but even with 20 percent of boys and 20 percent of girls only, reaching an overall ratio of 75/25 would imply 55 females for 5 males among the adults. That the female/male ratio should reach at least 11, entails a rate of out-migration of male adults of the order of 90 percent, which seems preposterous on a scale of several tens of million households.

6. Also see Lipton (1988: 45); United Nations (1995a: 129); and Quisumbing, Haddad and Peña (1995).

7. Our deliberate overstatement of 90 million PWHH for the whole world in 1995 compares with Jazairy, Alamgir and Panuccio's figure (op.cit.: 423) of 75.5 million for the rural segment of developing countries in 1988 (see footnote 5 above).

8. Among countries where data are available, the median difference between men-headed and women-headed household sizes is 1.5 for developed countries and 0.9 for developing ones (United Nations 1995b: 12).

9. The sum of numbers reported, for countries totalling 92 percent of the world population, is under 108 million (United Nations 1996d: 242-251).


Table. Distribution of urban households by poverty stratum, by sex of head of household: 12 Latin American countries, 1994
Country Sex of head of household Distribution of households by poverty stratum
Indigent Other poor Non-poor
Argentina Male
Bolivia Male
Brazil (1993) Male
Chile Male
Colombia Male
Costa Rica Male
Honduras Male
Mexico Male
Panama Male
Paraguay Male
Uruguay Male
Venezuela Male
Source: ad hoc tabulations from national household surveys. Female indices: CEPAL 1996: 202. Male indices: unpublished tables, CEPAL.


CEPAL, 1996: "Panorama social de América Latina". Santiago: Naciones Unidas.

HADDAD, Lawrence, Christine PENA, Chiruzu NISHIDA, Agnes QUISUMBING and Alison SLACK, 1996: "Food security and nutrition implications of intrahousehold bias: a review of literature". FCND Discussion Paper No. 19. Washington: IFPRI.

ICQL (Independent Commission on the Quality of Life), 1996: "Caring for the future". Oxford: Oxford University Press.

JAZAIRY, Idriss, Mohiuddin ALAMGIR and Theresa PANUCCIO, 1992: "The state of rural poverty. An inquiry into its causes and consequences". New York: New York University Press/IFAD.

LIPTON, Michael, 1983: "Demography and poverty". World Bank Staff Working Papers, 623. Washington.

LIPTON, Michael, 1988: "The poor and the poorest. Some interim findings". World Bank Discussion Papers, 25. Washington.

LIPTON, Michael and M. RAVALLION, 1995: "Poverty and policy". In "Handbook of development economics" (J. Behrman and T.N. Srinivasan, ed.), vol. 3. Amsterdam: North Holland.

LOPEZ, Cecilia M., Molly POLLACK and Marcela VILLARREAL, 1992: "Género y mercado de trabajo en América Latina". Santiago: OIT-PREALC.

MACKELLAR, F. Landis, Wolfgang LUTZ, Christopher PRINZ and Anne GOUJON, 1995: "Population, households, and CO2 emissions". Population and Development Review, Vol. 21 No. 4: 849-865.

QUIBRIA, M.G., 1993: "The gender and poverty nexus: issues and policies". Economics Staff Paper No. 51. Manila: Asian Development Bank.

QUISUMBING, Agnes R., Lawrence HADDAD and Christine PENA, 1995: "Gender and poverty: new evidence from 10 developing countries". FCND Discussion Paper No. 9. Washington: IFPRI.

UNDP, 1995: "Human development report". New York.

UNDP, 1996: "Human development report". New York.

UNIFEM, 1995: "The human cost of women's poverty: Perspectives from Latin America and the Caribbean". Mexico.

United Nations, 1995a: "The world's women 1995. Trends and statistics". New York.

United Nations, 1995b: "Living arrangements of women and their children in developing countries. A demographic profile". New York.

United Nations, 1996a: "Food security for all, food security for rural women". International Steering Committee on the Economic Advancement of Rural Women. Geneva.

United Nations, 1996b: "Too young to die: genes or gender?". Population Newsletter, No. 62: 12-15.

United Nations, 1996c: "World population prospects: the 1996 revision". New York.

Back to Top FAO Homepage