Table of Contents | Previous | Next |
II. Approaches to Examining the Literature
Covering the vast literature dealing with marriage, labor-market outcomes, and family functioning requires setting limits on what to include and thus what to exclude. This appendix examines a considerable number of relevant studies but does not offer a full, comprehensive review. Figure 1 depicts the three-way link between marital status, job-market outcomes, and family functioning. The triangular pattern displayed in the figure highlights the potential links running in both directions between any two of the three poles. One goal of this review is to highlight which arrows represent the strongest relationships. Now, in doing so, we recognize that each pole generally represents several possible variables or statuses. The labor-force status pole can represent employment, hours worked, wage rates, job stability or occupational status. The marital status pole can involve comparing marriage to never-married, separated, divorced, or remarried status, and, within each of the nonmarried statuses, to cohabiting or not cohabiting. Family functioning incorporates an extensive array of variables, from the health of relationships between partners and between parents and children to the way in which children are acculturated, mentored, and educated. Thus, in highlighting which are the key relationships, we must recognize, for example, that marriage (relative to being never-married or divorced and noncohabiting) may exert large and significant effects on some family-functioning outcomes but not on others.
The timing of relationships adds another level of complexity to defining the marriage, employment, and family-functioning poles. Current employment may affect the timing of marriage, divorce, or remarriage, but may have little impact on whether someone ultimately marries. Current unemployment may exert little effect on family functioning today, but frequent unemployment may lead to long-term family problems. Job availability in a single year may have little impact on marital status in the following year, but a sustained period of abundant employment opportunities may increase marriage rates. The relationships may vary substantially by age. Marrying too early or too late might weaken long-term job-market outcomes. Good jobs may do little to encourage marriage at some ages but may affect marriage greatly at other ages.
[D] |
The interactions between employment and marital status are likely to be strongly mediated by decisions about childbearing. Although no studies are able to capture the joint and simultaneous decisions of men and women concerning employment, marriage, and fertility, the trends suggest diverging patterns. College-educated women have been delaying marriage and childbearing by investing early in their careers. Meanwhile, less-educated women are continuing to have children at young ages and increasingly prior to marriage, perhaps because they see fewer work-related reasons to delay childbearing and because economic and social gains from marrying low-wage potential spouses are modest or nonexistent (Ellwood and Jencks 2004).
A second element of our approach is the distinction between associations (or correlations) and causation. Ideally, we would like to know about both association and causation. But, causation is typically difficult to determine, despite extensive efforts by researchers. Certainly, much of the literature aims (usually imperfectly) at establishing causation. Association, however, might be important for targeting purposes. If certain types of marital situations are correlated with the poor functioning of families, then focusing on families in these situations may be desirable even if the relationship is not causal. Often, studies yield evidence of correlations that control for a variety of observable differences between individuals. A good example is the positive correlation between marriage and wage rates, even among those with the same levels of education, age, race, and family background. Such results do not prove causation but are nonetheless informative, showing that the observed association is not an artifact of association with those factors.
The potential limitations on causal inference come mainly from biases associated with simultaneity and unmeasured heterogeneity. For example, simultaneity can arise in a statistical association between marriage and earnings, since higher earnings might be leading to marriage, marriage might be leading to higher earnings, or both. The idea of unmeasured heterogeneity (sometimes linked with selection) is that a third unmeasured factor (say, good looks, a good personality, or a strong responsibility ethic) is causing both marriage and higher earnings. The simultaneity and unmeasured heterogeneity problems pose difficulties in estimating causal links among marriage, employment, and family functioning, especially with nonexperimental analyses.
Third, in the review we recognize that the studies cover individuals in a variety of contexts and time periods. Findings about relationships that take place in some contexts may not be enduring and carry over to other contexts. For example, the interactions between work and marriage may operate in a different way for the generation born in the 1940s than for the generation born in the 1970s or 1980s. It is important to report the context (whether the study deals with various subgroups, from all U.S. residents to subgroups classified by age, sex, race, educational level, initial marital or parental status, economic status, and region of residence), the time period, and the economic conditions.
Although the primary interest in this review is how the various relationships operate among individuals with low to moderate expected incomes, the observational studies often cover a wider spectrum of the population. The review looks most closely at results for relevant target groups but does not exclude broader studies. As seen in section IV, the demonstrations and program evaluations typically focus on those from the low-income or lower middle-income groups. In reviewing these findings, we include results from projects that emphasized one outcome (such as higher employment and earnings), but where there are data on its link with the outcomes displayed in figure 1. For example, the primary goals of Job Corps and the focus of the Job Corps evaluation are to raise the education and skill levels of at-risk, out-of-school youth and to improve their career outcomes (Child Trends 2003). To the extent the project did so, it would have exerted an exogenous effect on earnings and career outcomes. We can then examine whether demonstrations also yielded positive effects on marriage or family functioning, possibly as an indirect effect of improved labor-market outcomes or possibly as a direct, though unintended, effect of the services provided.
| Table of Contents | Previous | Next |