Do Kids Get an 'Intact-Family Premium'?
Today, Robert I. Lerman and W. Bradford Wilcox have a new paper out through the American Enterprise Institute and the Institute for Family Studies. They argue that marriage is economically beneficial for adults and children alike and offer some policy ideas for shoring it up.
They are no doubt correct that family structure is important, but this is an exceedingly difficult topic to study, and I suspect their estimates of what they call the "intact-family premium" are high. Here I'll focus on their chapter "The Role of Childhood Family Structure in Future Economic Success."
Before getting to the nitty-gritty of the new study, let's start at the beginning: the breakdown of the American family. Out-of-wedlock childbearing and divorce were quite rare a half-century ago, and it's not hard to understand why. Women had few economic opportunities, and the government did little to help women who had children they couldn't support. If a woman wanted to have children, the only real option was to secure the economic support of a man first. Cultural norms reinforced this reality: Society looked down on unwed mothers and men who left their families.
None of that is true anymore. About 40 percent of American children are born out of wedlock, and divorce is common. And what makes this difficult to study is that the process didn't unfold at random -- it targeted specific parts of the population. For example, people who have children out of wedlock are disproportionately poor and less-educated: precisely the demographic of parents whose children have worse outcomes anyway, for reasons ranging from genetics to parenting. And even among the poor and less educated, the decision to have children out of wedlock is just that -- a decision -- meaning that it depends on a host of personal traits.
If we want to figure out what effect non-intact families have on kids, we need to account for all of these differences without also washing away differences that actually do stem from family structure. That's a tall order.
Lerman and Wilcox compare children using a model that takes into account "age, race/ethnicity, mother's age at birth, mother's education, and respondents' scores on the Armed Forces Qualifying Test (AFQT), which measures intelligence and knowledge of a range of subjects." Some of their calculations also control for the child's own education, household size, and whether he lives with his parents. The authors find that, with these variables removed from the equation, kids from intact families still have much higher personal incomes (to the tune of around five grand a year, depending on gender), better high-school graduation rates, and so on and so forth.
Two variables in particular highlight the pitfalls of this type of analysis. First, the AFQT: Intelligence is highly genetic, and (as we'll see in a second) it varies between married and unmarried parents, so controlling it is justifiable. But no one says academic ability is completely genetic, and it's possible that kids score somewhat better on tests if they have two parents to support their intellectual development. So by controlling away the kids' AFQT scores, the authors might have nibbled at the effect they were trying to find.
On the flipside is parental income, which the authors deliberately exclude from their analysis: "The higher income that children enjoy in a two-parent home is part of the benefit of being raised in an intact family." Indeed, two incomes are better than one (assuming the second parent works instead of staying at home), and in a different chapter the authors give some good evidence that marriage itself increases income. But it's also likely that people who form non-intact families have lower earning potential, on average, than those who get married and stay married -- and that this would affect their children to some degree regardless of family structure.
In other words, including a variable can wipe out the actual effects of family structure, but leaving it out can blame family structure for problems that kids would have had to deal with no matter what. There's no telling for sure, but my best guess is that the authors excluded far more of these kinds of variables than they included, and overestimated the impact of family structure. In particular, their results are hard to square with those from the field of behavioral genetics, which suggest that children who share genes tend to be similar in countless ways, even if they're not raised together, while unrelated children raised in the same home do not.
Importantly, I'm not so much disagreeing with the authors as emphasizing problems they admit are there: In their introduction, they concede that "both marriage and economic well-being may be the result of some third factor, such as unobserved differences in personality or character, like the capacity to delay gratification."
I'm a journalist, not a social scientist, so take the following with a grain of salt. But I have spent some time with the 1997 National Longitudinal Survey of Youth, one of the data sets Lerman and Wilcox use. To show what I'm getting at, I used it to look at two groups of non-black, non-Hispanic parents: those who got married the same year they started having kids or earlier, and those who did not. What's great about the NLSY is that, because it interviews the same people over and over again (the most recent data being from 2011), it allows us to time-travel and see how these groups compared when their members were as young as 12-16. With the caveat that my analysis was crude -- no fancy weights or controls -- it turns out that they were quite different in a lot of ways.
To start with, I looked at some numbers collected when these parents were in their school years. On one part of the AFQT, the math and verbal sections of the ASVAB, those who'd become married parents averaged around the 58th percentile nationally, while those who'd become unmarried parents averaged around the 40th percentile. Similarly, on the NLSY's GPA measure (designed to standardize scores across different schools), married parents averaged about a 3.0 while the unmarried were around a 2.6.
The members of these two groups grew up in different types of homes as well. Married parents came from families with a median income of about $50,000 in 1997, while the typical unmarried parent came from a family earning a little under $40,000. And unsurprisingly, those in the married-parent group were more likely to come from intact families themselves -- nearly 60 percent lived with both biological parents at age 12, compared with a little under 40 percent for those in the unmarried-parent group -- raising yet another slew of questions about nature and nurture. (A different chapter in the new study addresses people who grew up in intact families and went on to form intact families themselves.)
Moving beyond high school -- and into territory Lerman and Wilcox don't cover -- probably the most striking thing I found was a serious criminality gap, to the point that women who had children out of wedlock were roughly comparable to men who had children after getting married. Those in the unmarried-parent group were much more likely to report being arrested at least once: 69 vs. 31 percent for males and 40 percent to 16 percent for females. (Click here if these numbers seem far too high to you.) Ditto for reports of incarceration: 22 to 6 percent for men and 8 to 2 percent for women.
And here's a more salacious time-traveling comparison: Among those who had not married or had a child yet when they were asked in 2003, men in the unmarried-parent group were much more likely to say they'd had sex with a stranger since the date of the last interview, 18 vs. 9 percent. (The difference for women was similar in magnitude, 7 vs. 4 percent, but it was not statistically significant.) Morality aside, this is certainly indicative of personality traits that can affect other aspects of a person's life too.
Perhaps all we're seeing here is the fact that the unmarried-parent group grew up poorer -- after all, there's evidence that poverty can sometimes swamp the effect of genes in shaping outcomes. But most of these results held up when I excluded all respondents whose parents earned less than $30,000, or whose parents reported the respondent had experienced "hard times" as a child -- and even when, on top of that, I restricted the sample to those who'd lived with both biological parents at age 12. (None of the differences disappeared completely, but some fell out of statistical significance as the sample sizes shrank.)
You can see my spreadsheet and R Code for yourself, and download additional variables here. The NLSY is a treasure trove of data on all sorts of traits, and there's much I didn't look at. In fact, the differences -- including those not even measured in data sets like the NLSY -- may be so varied and complex that the tools of social science don't equip us to sort them out. Anything that differs between married and unmarried parents could affect children's outcomes through genetic transmission or parental influence, regardless of family structure.
It's obvious enough that family matters -- the simple, stable presence of another adult in the household makes everything easier, an extra income can bring a child out of poverty, and science is constantly uncovering ways that humans evolved to fill the roles of parent and child. And the benefits of a two-parent household can stay with children as they become adults. It's just difficult to quantify the precise extent of this "intact-family premium."
Robert VerBruggen is editor of RealClearPolicy. Twitter: @RAVerBruggen