Wonk Wong

By Henry Potrykus

The Washington Post recently posed the question of how we should react to the “unbelievable” finding that marriage is dissolving and single parenthood is rising in America.  The Post makes “clear:” “none of the findings [on which it relies] mean that children would necessarily be better off if their biological parents married.”

From that point of clarity, the Post (and the researchers on whom it relies) goes on to advocate that for the sake of postponing motherhood, more educational and career opportunities must be created for lower-educated women.  This, says the Post, should go along with the rather broad policy of improving the economic prospects of “suitable partners” those women are “searching for.”

First off, it’s downright un-American to be against more education.  This we must all stand for, even if a degree does not confer human capital as we expect (human capital: the skills, capacities, and know-how of value in the labor market), and even if the newly minted grads simply end up part-time baristas with a fat bill to pay.  So, for this rebuttal, let me put the policy aside.  Soundly critiquing such policies requires a more thorough examination of economic sociology.

Here, let me rebut the Post’s logic.  The Post is wrong when it says two sociologists’ work - the statistics of McLanahan and Jenks - mean the problem of non-marriage demands a solution “far beyond marriage.”

Now, the Post relies on work by McLanahan and Jenks, who caveated their findings:  Recall the laden term “necessarily” quoted above.

The real issue is what do statistics say for policy, and, specifically, what does research on marriage in America say for policy? 

Here’s the easy part: statistics are about the general - usually average - case.  So of course they don’t speak to something necessarily affecting any given individual.  They speak to what generally happens.  Governance is also best construed around the general, so statistics do have a use there.

Here’s the harder part, which is about the methodologies employed by researchers nowadays.  That is, it’s about science:  Researchers have more or less three classes of tools to study a phenomenon like marriage.  One is descriptive statistics, the second is called regression, and the third are - sometime occult - analyses of “natural” experiments.

Let’s put aside descriptive statistics (even though the work of McLanahan and Jenks has many nice figures); work by McLanahan and Jenks and other sociologists should be considered at par when they involve so-called multiple regression analysis.  These analyses show the simultaneous influence of different factors on some outcome.  Informally, if you say a group’s workforce participation level (a pro-social activity of interest to the Post and McLanahan and Jenks) is to be found irrespective of that group’s proclivity to be married you contradict this second type of analysis.  These analyses say that marriage influences partners’ workforce participation.

Of course, we are not interested in whether marriage and working are merely correlated.  Maybe only the guys already with jobs get the girls (in marriage of course).

This is where the hard part comes to a head:  Sophisticated analyses of the third type can show that the dissolution of marriage actually does affect behaviors and prospects, and does affect the outcomes of the children the Post wants helped.  Sophisticated analyses do uncover a positive effect of marriage on social outcomes.  These analyses confirm the more basic, second kind of analyses.

Let me reiterate that: The workhorse, more basic analyses of sociology tend not to be wrong-headed.  In fact, good analysis of the second type does point to causal relations.  It is just not in itself completely conclusive.  It nevertheless tends to align correctly - and even quantitatively - with the much more difficultly arrived-at causal analyses.  (There are reasons for this.)

We are at the end of the critique of the Post’s logic:  Children in general would be better off if their biological parents married.  Just because sociology’s baseline method (regression) is not totally conclusive, one cannot infer that that method shows things that are not there.  Saying “this fact is uncertainly arrived at, so it is false” is a bad inference!  (Marriage is important to social outcomes even if all that science we fund through NSF grants shows it!)

Even without delving deeper into the science, we can nonetheless conclude that the Post cannot make the inference it wants: none of this body of evidence means the problem of non-marriage demands a solution “far beyond” marriage itself. 

Quite the contrary: Solutions are found within marriage.  Sophisticated studies indicate marriage causes positive behavior changes that are [very] difficult to affect otherwise. 

The good analyses of the second type which line up with the causal studies show the same.  If the Post wants the economic prospects of “suitable partners” to improve, I suggest it stop looking “far beyond” an empirically proven means of doing so.

To close, I want to confess my own befuddlement in the Post’s choosing to call the last few decades’ flight from marriage “unbelievable.”  The calamity follows on the heels of the sexual revolution.  According to the Post, one behavior doesn’t beget another?  No, we should be about as bewildered by this - yes, seismic  - shift away from marriage as we would be in observing people buying more bananas once the price of bananas falls. 

And that elementary point should be what we pivot on to get back to good policy.