суббота, 3 марта 2012 г.

Premarital birth among young Hispanic women: evidence from semiparametric competing risks analysis.(Statistical table)(Report)

Introduction

The Aid to Families with Dependent Children (AFDC) program [replaced by Temporary Assistance for Needy Families (TANF)] was often blamed for increasing illegitimate births, declining marriage, and rising divorce. The main argument was, by favoring single-parent families, the AFDC welfare system reduced the costs of having children and increased the value of single parenthood as an alternative to marriage.

Between 1990 and 1999, the proportion of minority welfare families increased from 60% to just over 67% and the Hispanic welfare population grew from 17% to 25% (U.S. Department of Health and Human Services 2005). However, the existing literature on welfare's effects on marriage and fertility has largely focused on groups of white and black women. In a review of 112 studies on the subject, Moffitt (1998) refers to Hispanic only twice in relation to the papers by Plotnick (1990) and Lichter et al. (1997). Plotnick (1990) uses National Longitudinal Survey of Youth (NLSY) data to analyze the link between welfare and teenage out-of-wedlock childbearing using a simple logit hazard framework. He does not find a strong link. His analysis does not allow for state fixed effects or unobserved heterogeneity and marriage is treated as random censoring. Schultz (1994) uses data from the 1980 U.S. Census to analyze probit equations on the effect of welfare on the "Probability of a Woman Living with a Spouse," and Tobit and OLS equations to analyze the effects of welfare on the number of children born. He considers blacks and whites and three age groups. He finds that AFDC generosity has a negative, statistically significant, effect on marriage for the 14 to 24 age group. The effect of AFDC on fertility is negative and statistically significant only for whites ages 14 to 24. He does not consider state fixed effects. Hoynes (1997) uses data from the Panel Study of Income Dynamics (PSID) to study the effect of welfare generosity on the probability of female headship. She finds no significant effect of welfare for white women; for black women the effect became insignificant after controlling for individual unobserved heterogeneity.

Lichter et al. (1997) consider the impact of welfare on the percentage of female-headed families. They find welfare only has a small impact. Their analysis blends different sources of female headship (premarital birth, divorce, etc.) and considers race dummies as explanatory variables.

Rosenzweig (1999) uses NLSY data to study the effect of welfare on the probability of a woman having a premarital birth versus only marital births or no births by age 22. After controlling for state fixed effects and cohort fixed effects, he finds a significant and quantitatively large positive effect of welfare on the probability of premarital fertility, especially among low income women. Because his analysis is only concerned with the status quo at a specific age, it cannot distinguish between the hypothesis that the increase in premarital birth is the result of a direct effect of welfare on teen fertility, and the hypothesis that welfare reduces the likelihood of marriage. Any one of these effects can increase premarital births.

In a recent paper, Keane and Wolpin (2007) structurally estimate a dynamic programming model of life-cycle decisions of young black, Hispanic and white women. They account for a number of labor and marriage market outcomes, but do not focus on premarital birth. While TANF was substituted for AFDC, the relative simplicity of the old welfare program provides a unique opportunity to analyze the effects of economic incentives on marriage and fertility behavior. (1)

In this paper, we analyze the effects of the AFDC welfare generosity on a sample of young Hispanic women's premarital fertility and marriage choices from the NLSY data. A bivariate competing risks duration model framework allows us to identify a young women's premarital fertility and marriage choices. To evaluate the sensitivity of the empirical results, we estimate a variety of econometric specifications for different sets of explanatory variables, state fixed effects, and individual specific unobserved heterogeneity. We also provide a comparative analysis using a sample of young white and black women. Overall, our results for the black and white samples are consistent with previous research and we interpret this as validation of our empirical strategy.

The empirical evidence indicates welfare has a significant effect on a woman's decision to marry and a woman's premarital birth decisions. Results from a policy simulation indicate a 10% increase in welfare generosity results in a 10% increase in premarital births and a 7% decrease in marriage by age 24; both effects are significant. Our analysis also suggests a significant proportion of early marriage and premarital births for whites and blacks are due to individual specific characteristics, unobservable to the econometrician or unobservable heterogeneity, while this is not the case for Hispanics.

The next section describes the data. This is followed by a brief description of the analytical framework and the econometric model. After presenting the main estimation results for the Hispanic sample, including simulation results that illustrate the quantitative impact of welfare, we discuss the racial differences that emerge from the comparative analysis. Finally, concluding remarks are provided.

The Data

Teenagers and older women may have different reasons to conceive out of wedlock. In addition, we focus our attention on the problem of premarital birth among teenage women. The evidence suggests this group of women bears the highest burden from a premarital birth (Hotz et al. 1997; Ribar 1996; Haveman et al. 1997). We consider a sample of 886 Hispanic women from the main sample and the supplemental sample of Hispanic women in the NLSY. Individuals in the sample have been interviewed every year since 1979 and extensive information on family background, fertility and marital history has been collected. In addition, retrospective fertility and marital histories were collected in 1979. We follow women until the time of their first transition to marriage, premarital birth or until age 24, whichever comes first.

The premarital birth duration is defined as the time the young woman becomes at risk of having a child, age thirteen, (2) until the time of the first premarital birth. Similarly, the first marriage duration begins at age thirteen and continues until the time of the first marriage. In the empirical analysis, we classify births from cohabitating relationships as premarital births. Unfortunately, the information available in the NLSY about cohabitation is not as detailed as the information on marriage and fertility. At the time of each interview, the woman is asked if she is cohabitating with a partner of the opposite sex. There is no information about the duration of the cohabitation and no retrospective information on cohabitation for the years prior to 1979. We have examined data on the youngest cohorts in the NLSY to study the extent of cohabitation. Overall, cohabitation appears to be a long-term alternative to marriage for only a small percentage of the women in our sample who experience a premarital birth. (3) Thus, given data limitations, we feel not accounting for cohabitation in our empirical work is a reasonable compromise.

For each individual, the NLSY provides the state of residence at age 14 and each survey year. However, the state of residence between age 14 and the age at the time of their interview is unknown. Following Rosenzweig (1999), we substitute the missing information with the state of residence at age 14 or at the age of the first interview, depending on which is closest in time. We will use this information later to give state dummies to each individual-year observation. Another shortcoming of the NLSY data set is its failure to provide information on parental income. As in Rosenzweig (1999), a measure of parental income is constructed by combining information from the characteristics of the woman's parents in the sample when they were 14 and information on median wages by occupation, education level, gender, race and marital status, obtained from the census. Finally, the NLSY data is supplemented with data on welfare generosity, the availability of abortion services, and wages at the state level.

Table 1 reports descriptive statistics for the relevant variables in our Hispanic sample. Some of the variables describe the household characteristics where each woman resided at 14. About 17% of Hispanic women were living with a single mother and about 68% were living with both parents at 14. Table 1 also provides descriptive statistics for other important variables that will be used in the estimation of the model, like the standardized Armed Forces Qualification Test (AFQT), (4) religiosity and urban residence. There are no important observable differences between women that experience a premarital birth and women that experience marriage first. The second group of women live in households with slightly lower average incomes and education and are less likely to live in an urban area. These women also have a lower standardized AFQT, are more likely to live with both parents and are less likely to live with a single mother.

AFDC payments have been declining consistently over the time period analyzed. Older cohorts in the NLSY have been exposed to a more generous welfare program but there are a higher proportion of premarital births among the younger cohorts (Rosenzweig 1999). We also examined a composite measure of welfare generosity defined and used by Moffitt (1994). Because participants on the AFDC program usually participate in other welfare programs, Moffitt's measure of welfare generosity attempts to capture the overall value of welfare by including the AFDC, food stamps and Medicaid benefits. Moffitt's measure of welfare generosity for a family of two has remained relatively constant over the period. We use Moffitt's measure of welfare generosity in our analysis. (5)

Table 2 reports the descriptive statistics of the marriage and premarital birth processes for three ethnic groups: Hispanic, black and white. The average ages of premarital birth and first marriage are very similar for Hispanic and black women and somewhat higher for white women. However, there are important differences in the percentage of women in the marriage and premarital-birth states for different age groups. The marriage process among Hispanic and white women seem to be very similar until age 22; by age 24, we observe a 7% higher proportion of marriage among white women; black women are less likely to marry. In contrast, the number of Hispanic women that have experienced a premarital birth by a certain age is significantly lower than the …

Комментариев нет:

Отправить комментарий