Racial earnings disparities and family structure.

• Author: Darity, William A., Jr.

1. Introduction

   A heated debate of the past decade centered on explanations for the putative reductions in racial earnings gaps.(1) Recent evidence shows, however, that the trend toward convergence in black and white earnings came to a halt by the end of the 1970s and reversed during the Reagan-Bush years (see Cotton 1989; Juhn, Murphy, and Pierce 1991; Bound and Freeman 1992; Darity and Myers 1992a, b; Myers 1993). The earnings position of black family heads, moreover, shows a clear deterioration when comparisons are made between the mid-1970s and the mid-1980s. In 1976, the ratio of black to white earnings among family heads was 0.63. In 1985, it was 0.59. This means, of course, that black family heads earned $63 for every $100 that white family heads earned in 1976; but, in 1985, they earned only $59 for every $100 earned by a white family head. Thus, in the span of a decade, black family heads' earnings fell behind that of white family heads by four additional dollars for every $100 whites earned.

   During this same period, from the mid-1970s until the mid-1980s, the proportion of black families headed by females soared (see Ellwood and Crane 1990). In 1976, 36% of black families were headed by females. By 1985, 45% were headed by females. It is tempting to blame deterioration in the earnings of black family heads on this trend since females have lower labor force participation rates than males and lower labor force participation yields lower earnings.

   Can the drop in blacks' relative earnings be attributed to the dramatic shift in family structure? Does the answer depend on whether or not labor force participation and family structure are interdependent? This paper seeks answers to these two questions.

   First, we examine evidence on rising trends of female-headed families and on the problem of declining labor force participation rates among family heads. We then adopt a decomposition methodology to assess the importance of changing family structure in understanding the rise in earnings inequality among family heads. Two different versions of the decomposition model are presented: One considers family structure to be exogenous and unrelated to work behavior; a second explicitly models the impact of family structure on labor force participation.

   The period chosen for study is unique in that it represents a major transformation in income distributions in the U.S. As Bishop, Formby, and Smith (1991) confirm, this era was a highly volatile period. Moreover, it was a period when the apparent convergence in black and white individual incomes was not matched by similar movements in black and white family incomes. As Figure 1 reveals, while the ratio of black to white incomes among males increased during the mid-1970s to the mid-1980s, the ratio of black-to-white incomes among females declined, as did the ratio of black to white family incomes. In 1970, the ratios of black to white money incomes among males, females, and families were 0.5973, 0.8612, and 0.6474, respectively. Thereafter, into the 1980s, the ratio among males fluctuated little along a slight upward path; among females, it rose and then fell; and among families as a whole, it fell. Thus, by 1985, the ratios were 0.6214, 0.8724, and 0.6214, respectively. Even though there is some reversal in the black-white family income widening by the mid-1990s, the widening of the gaps during the 1980s poses a heretofore unresolved puzzle.

   Our findings in this paper shed new light on what happened during this critical period of changing family structure and changing income distributions. The results uncover only limited support for the view that racial differences in family structures alone contributed to widening earnings gaps. One version of our decomposition model does reveal widening earnings gaps resulting at least partly from increases in female-family headship and/or racial differences in family structure. But this version of the model demonstrates that gender and race discrimination also explain substantial portions of the racial gap in earnings. A second version of the model reveals that endowment impacts are modest in comparison to the impacts of racial discrimination. In other words, although family structure matters, racial discrimination is a more strongly influential factor in the determination of the widening racial gaps in earnings among family heads during the shift begun toward the end of the Carter administration and lasting into the Reagan and Bush administrations.

2. Trends in Family Structure and Labor Force Participation

   The percent of black families headed by females climbed from 20 to 25% in the 1950s to more than 40% in the 1980s (see Darity and Myers 1984). By the 1990s, slightly less than half of all black families were headed by females.

   How is the shift in family structure related to changes in earnings disparities? Superficially, the answer is that female-headed families have lower incomes on average than male-headed families.(2) On average, women earn lower wages and salaries than men. Thus, a potential explanation for the decline in the relative incomes of black families is the phenomenal increase in recent decades in the percent of black families headed by females. This rise is also evident among whites. In 1970, 9% of white families were headed by females with no spouse present; in 1991, the proportion was 13%. [TABULAR DATA FOR TABLE 1 OMITTED] Among black families in 1970, 28% were headed by females; by 1991, the proportion was 46%.

   Evidence on labor force participation rates suggests another possible explanation for widening racial earnings gaps. The overall trend for the 21-year period from 1970 though 1991 is one of decline in the labor force participation rates among black family heads. In 1970, about 73% of black family heads were in the labor force; by 1991, only 65% were working or looking for work. Labor force participation rates also declined for white family heads, although the ratio of black-to-white labor force participation rates dropped, particularly for female-headed families. For example, in 1976, the ratio was 0.94 for male-headed families, 0.93 for female-headed families, and 0.85 for all families. In 1985, the ratio dropped to 0.93 for male-headed families, 0.85 for female-headed families, and 0.82 overall (see Darity and Myers 1984).

   Table 1 reproduces elasticities obtained by estimating the effects of age, expected welfare, and family structure on labor force participation rates among family heads.(3) The data relate to family heads in 1976 and 1985 found in the Current Population Survey, March Supplement tapes. Whether the family head was white or black in years 1976 or 1985, female family heads were less likely to be looking for work or to be working than were male family heads. While the coefficients are nearly identical for blacks and whites in 1976, the marginal effects (i.e., the derivatives) are greater for blacks than whites. The same is true for 1985, even though the differential is smaller. In other words, being a female reduces labor force participation more for black family heads than it does for white family heads.(4)

   The computed elasticities of labor force participation with respect to age, family structure, and expected welfare revealed in Table 1 confirm that family structure is not entirely exogenous to work decisions.(5) Females heading families have a considerably more pronounced effect for blacks than whites. A 1% increase in the probability that a family is female-headed lowers labor force participation rates five to five and one-half times more for blacks than whites; but then, it only lowers the black labor force participation rate by less than one quarter of 1%. The inelastic impact of family structure on work behavior is simply much larger for black family heads.

   Age of family head, however, has a very elastic impact. Quite clearly, increases in the age of the head of family have larger percentage impacts on labor force participation than either welfare income or the family head being female. A 1% decrease in the age of the family head, computed at the average age, reduces labor force participation rates by more than 2% to nearly 4%. The impacts are larger for blacks than whites, with the gap widening from 1976 to 1985.(6)

   Thus, a plausible explanation for the deterioration in the earnings position of black families is the relative decline in labor force participation rates among black female-headed families. This explanation, along with the female-headship explanation, must be examined within the context of a structural model that admits the possibility that there are differing impacts of family structure and labor force participation rates on black and white family earnings.

3. A Model of Racial Earnings Disparities and Family Structure

   In order to gauge the effects of family structure and labor force participation on earnings disparities, we adopt a decomposition methodology. The methodology permits us to ask how much of the change in the racial gap in earnings can be explained by changes in family structure when family structure and labor force participation are independent of one another. The same methodology permits us to ask how much of the changing gap can be explained by changes in endowments or changes in treatment. We examine Current Population Survey data for two periods, 1976 and 1985. By comparing the mid-1970s experience to the mid-1980s experience, we are able to answer many of the questions concerning the causes of the widening racial gap in earnings of family heads.

   Consider the following specification of the earnings equation:

   ln(Y) = [[Xi].sub.0] + [Sigma] [[Xi].sub.i][X.sub.i] + [[Xi].sub.n+1]LFPR + [Lambda], (1)

   where X is a vector of independent variables and LFPR denotes labor force participation, measured by the percent of the year spent working or looking for work.(7) LFPR in this specification is not a dichotomous variable usually employed to...