Inequality and the Marriage Crisis

International New York Times | 26 March 2015 

This is coming later than I’d hoped and the world has moved on, but as promised I want to take up the other frequent rebuttal to the much-debated claim that economic changes can’t explain the two-parent family’s half century of decline — the argument that it’s relative income, not absolute income, that matters most to people’s lives and hopes and aspirations, and therefore the post-1960s increase in income inequality suffices to explain the social crisis among the working class even if working-class wages themselves haven’t really fallen.

As far as potential evidence for this hypothesis goes, here are a couple of the best cases I’ve seen. (No doubt there are others.) First, there’s this paper from Melissa Kearney and Phillip Levine, which correlates higher rates of unwed teenage births to higher inequality across both the United States and Europe, and suggests that stratification creates a “culture of despair” among the young and poor that encourages poorer women to seek validation and even honor through early motherhood rather than work or advancement. Second, there’s the chart attached to this Andrew Cherlin op-ed (and featured in his —recommended — new book), which shows not only that the overall marriage rate was lower in the Gilded Age and in our own high-Gini-coeffiecient era than in the more income-compressed 1940s and 1950s, but also that the class gap in the marriage rate shrank as incomes compressed and then widened as the Gini coefficient rose.

Now here are the various reasons why I’m skeptical. First, note that Kearney and Levine are looking at early nonmarital childbearing — births to mothers 19 years old and younger. But the American teen birth rate, with the exception of a 1980s surge, has actually been in steep decline from the 1960s to the present, the period of rising inequality that we’re discussing here, for reasons that are debatable and partially mysterious. (Interestingly, Kearney and Levine are also the authors of the paper arguing that the influence of “16 and Pregnant” may have played a recent role in this decline, which is one for the “what’s on television matters” file.) Now: Some of that trend is driven by later marriage and delayed childbearing; a lot of those teenage parents in the 1950s and 1960s were young marrieds (and sometimes shotgun marrieds) having their first kids, which means they wouldn’t be part of the Kearney-Levine sample or affected by their proposed explanation. But even if you take the nonmarital teen birth rate, what you’ll see (look at Table 16) is a rise and then a fall between 1970 and the present (even as the U.S. Gini coefficient kept rising through the 1990s and 2000s) such that the rate today is back where it was in 1975; meanwhile it was the nonmarital birth rate for everybody else roughly doubled, with larger increases for older demographics. So whatever the link between inequality and unwed teen childbearing, it hasn’t manifested itself across the last four decades the way the specific correlation in their paper would lead to one to expect.

This doesn’t mean the paper is wrong about the basic idea that, ceteris paribus, people with weaker economics prospects and fewer hopes for advancement tend to make different choices around sex and family than do the middle class and affluent. (As the authors note, today the choice that matters most is often whether or not to have an abortion.) But for the sweeping social change we’ve seen since the 1950s — the spread of single-parent families, serial cohabitation and multi-partner fertility out of the underclass and into the working and middle class — to be explained by an increase in inequality-driven socioeconomic despair, you would need evidence that rising inequality is effectively trapping, not just the poorest Americans, but many millions of people making closer to the median income in the kind of no-hope environment that we associate with the deep poverty of the underclass. And since, again, unwed teen births aren’t soaring with the Gini coefficient, this sense of despair would have to be kicking in somewhat later in the life cycle … so you would need to make the case that more and more late twentysomethings and early thirtysomethings are hitting a wall of crushing hopelessness about their future that previous generations never ran into or had to face.

And I just don’t think the evidence is there. It isn’t visible in trends in socioeconomic mobility, which is weaker than it should be but which doesn’t show much of a change over the last half-century. And it isn’t really visible in income inequality trends themselves, which show the upper class pulling away from the 80 percent and (especially) the 1 percent pulling away from the 99 percent, but which don’t show a vast, impossible-to-bridge chasm opening between, say, the fourth quintile and the second. If despair over opportunity were really driving post-1960s family trends in the lower middle class, you would expect to see a visible decrease in mobility over that period, and a wide enough gap between incomes in the lower middle and the upper middle that someone born to lower middle class parents or starting out at a low-wage job would have reason to give up hope of ever bringing home $50,000 a year. But that isn’t what see: That hypothetical worker’s chances of material advancement may not be what we would want them to be, but on the best evidence we have they have not declined as income inequality has climbed.

Similarly, if inequality were driving stratification which in turn mainly drives family breakdown, you would expect the three would be very tightly associated with one another. But in the big mobility study, by Harvard’s Raj Chetty and others, that found that U.S. mobility was stagnant rather than falling, there was a tight correlation between socioeconomic mobility and intact families … but a much weaker correlation between mobility and income equality itself. So it’s hard to see exactly how the more weakly-correlated trend could be the main driver of two more tightly-correlated trends.

Then I have similar arrow-of-causation questions about Cherlin’s (very interesting) charts showing marriage rates converging and diverging by occupation as inequality diminishes and then increases. My bigger issue with the chart is that it doesn’t tell us much about family stability, across classes or in general, prior to the 1950s: It’s true that mid-20th century America was distinctively marriage-oriented and fecund compared to some preceding eras, but while those eras had lower marriage rates they did not have the rates of divorce and single parenthood associated with the post-sexual revolution landscape. (Also, marriage rates hit a unique high in the 1950s, but the decline since is now taking us into uncharted territory rather than just reverting to a mean.) But even if we just focus on the post-1950s period, Brad Wilcox has pointed out that using Cherlin’s own classifications it looks like both the decline in marriage rates and the class divergence begin in the ’60s, before the widening of inequality in the 1970s. So interpreted crudely, his chart could very easily be used to just make the case that the sexual revolution weakens marriage, and that weakening in turn causes inequality – rather than inequality itself being the root cause.

This view, that income inequality really is mostly the result of family instability, is not without its champions, since the most common measurements of inequality tend to use household income, which is vulnerable to changes in family size and composition and stability. (The same holds for measurements of income volatility, a subject I’ve touched on earlier.) Indeed, you can find a sweeping version of this analysis that argues that controlling for household size basically eliminates the entire Gini-coefficient change since the 1950s … but people I trust don’t trust the results there. For more restrained takes, here are two papers — one from the mid-2000s, one from around 2010 — that tackle the question; both suggest that family fragmentation had a meaningful effect on household inequality; at the same time, both point to research suggesting that female workforce participation has reduced household inequality, possibly offsetting the fragmentation effect. So some of the discussion turns on whether you think of the decline of marriage and rising female earnings as a package deal, which is implicit (at least) in some left-wing and some right-wing analysis, or whether you think they didn’t have to happen together and don’t need to be permanently conjoined.

That’s a question for another time, but in closing I want to reach for a little common ground. The fact of female earnings growth also points to another issue — the impact of increasing assortative mating (the well-educated marrying the well-educated) in an era of rising female earnings and attainment, which probably tends to magnify inequality itself and which is bound up in a larger trend that I’ve written about before: Namely, the growing geographic segregation of the well-educated, the way that college graduates cluster more and more in specific regions and cities and neighborhoods than they did a few generations back.

Of all the mechanisms through which economic inequality could be shaping working class family structure and communal life for the worse, this strikes me as the most plausible. The problem, that is, isn’t that inequality imposes outright deprivation on the poor or prevents them from ever becoming middle class; it’s that when the returns to the talented are significantly higher than for everyone else, it’s much easier and more natural for well-educated Americans to self-segregate socially, to live and go to school and (sometimes) go to church in cities and communities that are effectively gated against anyone whose earnings miss the cut-off for rent, a down payment, private school, or whatever else you need to flourish. And that segregation leads, in a hard-to-quantify but somewhat commonsensical way, to a kind of hoarding of social capital, in which the pool of potential friends and mentors and significant others and civic leaders is shallower and less diverse in working class communities than it would have been in the past, and certain kinds of social examples and everyday encounters and support structures that kids and families need are less immediately available than they would be in a society less efficient in meritocratic sorting than our own.

This problem, as I’ve argued before in the context of the Piketty debate, isn’t really a 1 percent issue, since the richest 1 percent are always going to find ways to self-segregate. (The Harvard of the 1950s was not somehow a middle-class friendly place just because American incomes were more compressed overall.) Rather, it’s more a problem of what Piketty calls the “petits rentiers,” the mass upper class, whose interests are opposed to sound public policy in various ways, but whose increasing self-segregation, in particular, is effectively underwritten by everything from the tax code to urban zoning rules to immigration policy. And since that underwriting has critics on both sides of the partisan divide, notwithstanding all the right-left divisions on inequality it’s possible to imagine a certain kind of egalitarian agenda taking shape around anti-rentier reforms — and it’s further possible to hope that this kind of egalitarianism could, by encouraging a little more dispersal of talent and social capital, gradually work to counteract the social and familial crisis further down the education/income ladder.

But to return to my major theme here: This hope, or the hope involved in any public policy intervention, only makes sense if it’s tempered by the acknowledgment that Something Else is also going on.