Philip N. Cohen, a sociologist at the University of Maryland, has written a critique of my and W. Bradford Wilcox’s research brief on red states, blue states, and family stability for children. In it, he claims to undercut our conclusions based on his own analysis of data from the American Community Survey. He argues that state-level factors account for very little variation in family stability after accounting for parental race and education. Taken at face value, Cohen’s critique calls into question not only our argument about the role of red state culture, but also Naomi Cahn and June Carbone’s claim that blue states have an advantage when it comes to fostering family stability.

However, the Cohen critique has the following flaws:

  • Using a different outcome measure;
  • Using a different level of analysis;
  • Using a less than ideal method of analysis;
  • Breaking down the Hispanic group into subgroups that are confounded with state of residence;
  • Ignoring an important part of our findings.

Different Outcome Measure

Rather than looking at the proportion of teens (aged 15-17) in a state who have grown up with married birth parents throughout their childhood, he looks at the likelihood that an individual young child (zero to five years old) is currently living with married parents (including birth parents, stepparents, adoptive parents, and possibly grandparents). This greatly reduces the impact of divorces, many of which occur after children are five years old.  It also ignores the greater developmental risks that children in stepfamilies and other non-biological, two-parent families have been found to encounter. In short, Cohen’s analyses are performed using a different and less valuable outcome measure.

Different Level of Analysis

Rather than using a state level of analysis, Cohen analyzes ACS data at the individual child level, with state fixed effects.  One can argue about which level is appropriate for policy-relevant conclusions. Many economists and sociologists believe that state-level analyses are suitable for drawing inferences about the potential impact of state-level policies or practices.

Less Than Ideal Method of Analysis

Given his use of individual child data, I believe that Cohen would have done better to perform a multilevel analysis that included state-level factors, such as the Red State Index Wilcox and I created or the level of spending on education, which Cohen deems to be critically important.

Ironically, Cohen criticizes “red state family culture” as being “amorphous,” then goes on to invoke the importance of state “legacies of racial inequality and educational under­investment.” Talk about amorphous and value-laden terminology!

Breaking Down Hispanic Group into Subgroups

By breaking down “Hispanics” as he does, into Mexicans, Puerto Ricans, Cubans, etc., Cohen draws away some of the policy-related, red state-blue state variance into race and ethnicity. Detailed Hispanic origin and state are highly confounded. To illustrate: 74 percent of Puerto Ricans in the US (excluding Puerto Rico itself) live in seven states (New York, Florida, New Jersey, Pennsylvania, Massachusetts, Connecticut, Illinois).  Except for Florida, these are all blue states.  Likewise, 76 percent of all Cubans in the US live in 3 states (Florida, New Jersey, New York), and 71 percent of all Mexicans in the US live in 4 states (California, Texas, Illinois, Arizona). The concentration differences would be even more extreme with children aged zero to five, which are Cohen’s subjects of analysis.

Ignoring an Important Finding

Cohen ignores the positive relationship we found between the red state index and the proportion of the state adult population that is married. This is important not only in helping to make the concept of state family culture more concrete, but also because of the increasing trend of college-educated young adults postponing marriage and childbearing into their thirties and forties, or winding up not having children at all. The red state of Utah differs dramatically from the blue state of Massachusetts in average age of women and men at first marriage and first childbirth, as well as in the proportions who have children and the numbers of children they have.

Finally, in no way did Wilcox and I discount the importance of education and race and ethnicity and their relationships to family behavior and child wellbeing.  We believe that, in general, cross-state policy and cultural differences have modest impacts compared to these potent socioeconomic and demographic factors. But differences in educational attainment are not the same as differences in state educational investment. Indeed, we know that educational attainment is also shaped by family structure: children from intact, married parents are more likely to graduate from college than are children from non-intact families, even controlling for their parents’ socioeconomic status. Nor are disparities across racial and ethnic groups prima facie evidence of differences in racial discrimination. It is noteworthy that we found significant positive relationships with the predominant political orientation of a state even after controlling for powerful socioeconomic and demographic factors.