Discussion Paper Series

Grogger, Jeff, and Nick Ronan , "The Intergenerational Effects of Fatherlessness on Educational Attainment and Entry-Level Wages." NLS Discussion Paper, Report: NLS 96-30

In 1960, only five percent of all births occurred out of wedlock, and only 13 percent of all children lived in a single-parent family. By 1990, increasing divorce and out-of-wedlock childbearing had raised these figures substantially: more than one-fourth of all births were to unmarried mothers and 27 percent of all children lived with only one parent. Bumpass (1984) predicted that, as a result of rising divorce and non-marital fertility, 50 percent of all children born as early at 1980 would spend part of their childhood in a fatherless family.

Assuming that workers enter the labor market at age 20, and that the workforce turns over every 40 years, these demographic trends imply that, within the next 20 years or so, roughly one-fourth of the labor force will have spent part of its childhood in a single -parent home. Put differently, this means that within the next couple of decades, there will be as many workers who grew up fatherless as workers who hold college degrees. The fraction of the labor force that was raised in a single-parent family will undoubtedly exceed the fraction that is unionized.

These trends in family structure may bode ill for tomorrow's labor force, at least if the predictions of economic theory are correct. In Becker's (1981) model, for example, children raised in families with fewer resources tend to have lower human capital. Thus economic theory would predict that, all else equal, the next generation of workers will enter the labor market with less human capital than the last.

The objective of this study is to estimate the effects of fatherlessness on children's educational attainment and entry-level wages. We consider an important methodological issue not addressed by previous researchers: unobserved heterogeneity across families. One can imagine that families vary greatly in a number of ways that are unobservable to the analyst. Moreover, many of these unobservable family characteristics are likely to be correlated both with the probability of divorce and with the well-being of the children. Thus a cross sectional regression of children's educational attainment on a measure of their childhood family structure fails to identify the effect of living in a fatherless family, because the effects of fatherlessness are confounded with the effects of the family-specific unobservables. We would generally expect such unobserved heterogeneity to lead to exaggerated estimates of the true effect of fatherlessness.

We adjust for family-specific unobservables by making within-family comparisons. Drawing on previous research, we specify a child's human capital to depend on the number of years she spends in a single-parent family. Because children enter and leave the family at different times, the duration of a spell of fatherlessness generally will vary among siblings. To eliminate the effects of family-specific unobservables, we difference the data within families, relating differences in human capital to differences in the duration of the fatherless spell.

If spells were measured accurately, then differencing within families would provide valid estimates of the effects of a year of fatherlessness. Our data on childhood living arrangements are measured retrospectively, however, and there is evidence of a fair amount of measurement error, particularly in the differenced data. Under standard assumptions measurement error causes the estimated regression coefficients to be biased downward.

The usual solution to this problem, instrumental variables estimation, performs poorly in this case. The approach we adopt instead is method-of-moments estimation. We implement this approach by using sibling caparisons to estimate the extent of the measurement error in our retrospective data.

The data are taken from the NLSY. This longitudinal survey has several important features without which our analysis would be impossible. First, roughly half of its members have siblings who also took part in the survey. It therefore offers samples of siblings that are large enough for meaningful analyses. Second, it has detailed information on the childhood living arrangements of its respondents. Finally, its participants have, for the most part, completed their education. Thus we can analyze the effect of fatherlessness on children's ultimate educational attainment, rather than intermediate measures such as high school graduation.

For whites we find that fatherlessness has a negative effect on educational attainment, regardless how we estimate the models. Moreover, although the estimates vary somewhat, all are significant, at least at the then percent level of significance. Thus for whites, the evidence is clear: longer spells in a single-parent family lower educational attainment. The only question concerns the precise magnitude of this negative effect. We find some evidence that is consistent with the unobserved heterogeneity hypothesis, but we also find evidence of measurement error. On the basis of a number of tests, we cannot rule out the explanation that these two countervailing specification errors simply cancel each other out. Our best estimate is that each additional year of fatherlessness reduces the child's educational attainment by six-tenths of a year. Since the typical fatherless spell lasts about nine years, we conclude that the typical white child in a single parent family will acquire about one-half year less education than she would have had her parents remained together.

The picture is similar for Hispanics: additional years of fatherlessness lead to lower educational attainment, and the estimates are largely similar regardless how they are computed. Because our samples of Hispanics are smaller than our samples of whites, we must necessarily be somewhat more cautious about our conclusions. To a great extent, however, the typical Hispanic child who spends time in a simple-parent family faces disadvantages similar to those of his white counterparts. On average, Hispanic children who grow up fatherless acquire one-half year less education than they would have if they had lived with both parents.

For blacks the results are most surprising. Our unadjusted estimated indicate the typical pattern: longer spells of fatherlessness lead to lower educational attainment. When we control for family-specific unobservables, however, the estimate changes sign. Furthermore, it is difficult to attribute this occurrence to chance, since the adjusted estimates are statistically significant. Thus on the surface, our estimates suggest that black children who live in single-parent homes actually acquire more education than they would if they lived with both parents. This result is puzzling, and calls for further study. Indeed, until it is confirmed by future research, it is best to view this finding as tentative.

Because fatherlessness reduces educational attainment, at least for whites and Hispanics, and because education is an important determinant of adult wages, we expect that fatherlessness would contribute adversely to the children's adult earnings as well. When we analyze the effects of fatherlessness on wages directly, however, the evidence is mixed. Although the unadjusted estimates are typical negative, estimates that adjust for family effects are either positive or negative but very small. Adding further to the difficulty in interpreting these results, the adjusted estimates typically are insignificant.

We suspect that they mixed results stem from the nature of the wage data available in the NLSY. By necessity, the NLSY includes only entry-level wages, which for a number of reasons may be rather noisy compared to the wages of prime-age workers. We conclude that it would be best to revisit this issue with data on older workers.

Copies of this and other papers in this series are available from BLS by contacting Rita Jain at Jain.Rita@bls.gov or at (202) 691 - 7405.

Last Modified Date: July 09, 2003