Although family poverty in the United States has fluctuated over the last twenty or so years, one thing has remained unchanged: the poverty rate among female-headed families consistently has been about three times that of male-headed families(1) and about six times that of married-couple families (United States Department of Commerce, Bureau of the Census, 1991, Table 4). The "conventional wisdom" is that female-headed families have high poverty rates because, in general, they consist of one adult and many children and the women who head them are not in the workforce (Schiller, 1984, p. 92; Peterson, 1987, p. 330). This explains society's concern about the high degree of poverty among female-headed families for it implies (1) that female-headed families are likely to be dependent on welfare for long periods of time and (2) that many children are being raised in poverty and, consequently, are not getting access to the resources needed to enable them to become productive adults. There is also a fear that poverty will be passed from one generation to another and that in the process it will grow. This paper attempts a systematic investigation of why poverty rates for female-headed families are so much higher than those of male-headed families and married-couple families.(2)
Section 2 describes the data used in the study. Three reduced-form probit models -- one for each family type (married-couple, male-headed and female-headed), all with the same set of exogenous variables -- are presented in Section 3, and estimated in Section 4. In Section 5 the models are used to measure the extent to which the observed differences in poverty rates between female-headed families and the other two family types are due to (1) differences in the levels of the exogenous variables and (2) to differences in the impacts of the exogenous variables on the probability of being poor. The marginal effects on poverty of various exogenous variables are investigated for each family type. The conclusions and policy implications of the analysis are discussed in Section 6.
2. The Data
The data used throughout this paper are the Public Use Microdata Sample (C Sample) collected by the U.S. Department of Commerce, Bureau of the Census.(3) This is a one percent random sample of households from the 1980 United States Census of Population and Housing, and it records income during the 1979 calendar year.(4) Unlike more recent data sets, it contains a large enough number of male-headed families for a meaningful analysis of poverty rate differentials across family types.
Not all households are included in the analysis. Unrelated individuals, families without children, and families with children headed by elderly adults who are not in the workforce are excluded from the analysis because their incomes are likely to be affected by factors different from those that determine the incomes of families with children. Many unrelated individuals and married couples are elderly, have retired from the workforce, and get most of their income from social security. Others are young, have not yet entered the workforce full time, and receive financial support from their parents. Most families with children, on the other hand, are headed by adults of working age and the family's poverty status is closely related to the earning capacities of its adult members.
The study attempts to explain pre-transfer poverty of families with children so the definition of family income employed in the paper is the pre-tax aggregate of wages and salaries; self-employment income; interest, dividends and net rental income (sources 1 through 4 in Table 1). Government transfers in the form of public assistance and social security income are excluded as is all "other income" (see United States Department of Commerce, Bureau of the Census, 1983, pp. 90-91)(5). It is recognized that observed pre-transfer poverty rates are not necessarily those that would be observed in the absence of government transfers. …