Doing the Marriage Thing Again

Melissa Kearney is promoting a book where she dares to speak a truth that nobody else is brave enough to say: marriage and two-parent families are very good. So stigmatized is this view that in the last three days alone, the New York Times, the Atlantic, and the Washington Post (twice) have featured pieces promoting the book.

Given the hype, I was hoping that Kearney had something new to say on the topic, but sadly she does not. It’s the same dumb shit with the same dumb analytical mistakes that I’ve been seeing every few months since I started doing this work in 2011.

Not Controlling Correctly

The naive way to prove that kids in one-parent households (1P-kids) fare worse than kids in two-parent households (2P-kids) is to track the average life outcomes of each type of kid and see who does better.

But sophisticated researchers, like Kearney, know that this kind of comparison is not sufficient because 1P-kids and 2P-kids are different from one another in other ways, including their socioeconomic status, that may also explain their differences in outcome.

The solution to this problem is to control for these other differences by, for example, comparing 1P-kids and 2P-kids who have the same demographic characteristics. This is described as controlling for these other variables in order to isolate the variable of interest, which is the number of parents present.

But this solution is just as naive as the first one because 1P-kids and 2P-kids with the same demographic characteristics also differ in another important respect: the underlying quality of their parents’ relationship. The lack of two parents in the household is not a random occurrence. It generally happens because, for one reason or another, the parents could not get along.

To understand my point here, imagine we have two married families that are demographically and socioeconomically identical. Now imagine that one of those two families experienced a divorce. If we wanted to isolate the effect of that divorce on the kids in that family, would we be able to do so by comparing those kids to the kids in the other family that did not experience a divorce?

Kearney and her ilk say yes. But this is clearly not true. In general, there was some kind of problem present in the family that experienced the divorce that was not present in the family that did not experience the divorce. To actually analyze the effect of the divorce, you would need to compare the divorced family to a hypothetical version of itself that did not divorce despite whatever the problem was that lead to the divorce.

Another way to put this is: you need to compare couples with the same kinds of relationship dysfunctions, but where one couple splits and the other couple stays together. Comparing a split-couple with a certain relationship dysfunction to an intact-couple that does not have that relationship dysfunction, which is what all of this research does, is an obvious mistake.

To put this in less abstract terms, if dad beats mom and they get divorced, it’s not accurate to say that, but for the divorce, the kids would have fared exactly the same as a similar non-divorced family where dad does not beat mom. Yet that is what the research Kearney relies upon assumes.

Thinking at the Margin

A similar problem that plagues this kind of writing is that it casually assumes that the parent that is “missing” from a 1P family would, if he or she was present, be an average parent. In Kearney’s NYT piece, she writes about a hypothetical missing parent who earns $44,000 per year (the median for a high school graduate) and contributes “considerable time and energy to taking care of children.”

The assumption that the missing parents in 1P families are average parents who would contribute an average amount of earnings and an average amount of child care is obviously ridiculous. As with any group, the missing parents are a heterogenous bunch, but that population almost certainly skews towards below-average earnings and below-average domestic contribution, with many actually having a net-negative domestic contribution, whether because they are abusive, demanding, or otherwise.

This is easily the goofiest thing about this discourse. To hear Kearney talk about it, you’d think that the single mothers of the world are turning down $44,000 and hundreds of hours of free child care each year. But why would someone do that? If that option is as incredibly beneficial as Kearney says it is, why don’t people select it? Do people want their lives to be bad rather than good?

The solution to this seeming paradox is that Kearney and her ilk are just obviously wrong that the missing parent is otherwise an average person who would make average monetary and non-monetary contributions. The relationship that is on the margin of staying intact is not the same as the average intact relationship. This is pretty basic stuff and yet every few years, the discourse is blessed by someone who doesn’t get it.

A Better Mental Model

Rather than only make critiques here, let me propose a better framework for thinking through this whole question.

Imagine if the two parents of all of the children in the country lived as couples in the same household. Contrary to the assumptions used in many of the arguments in this discourse, these couples are not all average. Instead, their relationships have varying quality: some extremely low quality, some extremely high quality, and some in the middle.

Let’s say we sorted these couples based on the relationship quality thereby creating a distribution of relationship quality. In this distribution, the 1st percentile couple is in a living hell: substance abuse, violence, criminality, you name it. The 100th percentile couple is in paradise: no conflict ever, super-high-earning, equal contributors, and so on.

I think we can all agree that at the lower end of that distribution, it would be better for the kids if their parents’ cohabitation ended. We can also all agree that at the higher end of the distribution, it would be worse for the kids if their parents’ cohabitation ended.

With these agreements established, the question becomes purely one of where in the distribution does net-negative turn into net-positive. Is the 5th percentile relationship a net-positive? The 10th? The 30th?

This is obviously a very bloodless and abstract way to look at the question, but it is the actual question we should be debating. Kearney surely does not believe that the 1st percentile parenting couple should remain intact and that keeping it intact would be a net positive for kids. So if Kearney wants to actually make an interesting contribution, then she should walk us to the spot in the relationship-quality distribution where she thinks negative turns into positive.

At present around 78 percent of kids live in two-parent households, meaning that 22 percent do not. The interesting question is how many of those 22 percent of kids would be better off if they lived in a two-parent household with their parents at the relationship quality that those parents would have (not at the average relationship quality). We could also ask the opposite question, which is how many of those 78 percent of kids in two-parent households would be better off in one-parent households. Nobody ever asks that question, but the number is definitely not zero.

Hands Off The Welfare State

When I write stuff like this, people sometimes conclude that I have some stake in the “marriage debates.” And really I don’t. I think people should evangelize whatever lifestyles they want to evangelize. Spread the good news about how good your religion, relationship, exercise, and diet are. I am all for it.

Where I draw the line is when these lifestyle battles bleed into arguments about how only this lifestyle, and not the welfare state, can achieve distributive equality in society. That is a lie and it is the best way to actually hurt the most vulnerable people in the society by laying the ground work for opposition to the welfare state.