In a “letter from the editor” last week, The New York Times' David Leonhardt claimed that liberals overlook evidence that changing household structure, meaning the relative decline of households headed by a married couple, increases inequality. He wrote, “I’d say that family structure is an area where many liberals are putting more weight on their preconceptions (inequality is bad for society) than on the evidence (changes in family structure are both an effect and a cause of inequality).” But there is substantial evidence that household structure is what’s known as an endogenous variable: People make marriage, divorce, and child-rearing decisions due to their economic circumstances. Furthermore, the causes of rising inequality are the same reasons why household structure has changed: stagnant wages, labor market detachment, and job lock.
When questioned on Twitter about the evidence for his claim that changes in family structure cause inequality to rise, Leonhardt cited a paper by Professor Molly Martin of Penn State. But Martin argues nothing of the sort.
Her paper, "Family structure and income inequality in families with children, 1976 to 2000," essentially asks1 this question: "To what extent is rising inequality due to growing inequality between types of households (e.g., the gap between married couples and single mothers has grown, and a larger share of households belong to the more disadvantaged group), and to what extent is it attributable to growth of inequality within types of households?” Martin finds that 41 percent of the increase in household inequality from 1976-2000 is attributable to changing family structure.
But in a section entitled “Limitations,” Martin writes, “First, the results do not document causal relationships. I cannot determine the degree to which family structure changes caused the observed changes in inequality.” And Martin concludes an appendix further addressing these issues with “family formation probably reacts to prevailing economic conditions and, in that response, sets the conditions for perpetuating broader inequality patterns.” A single clause from the abstract of her paper, "family structure shifts explain 41% of the increase in inequality,” has been incorrectly cited as evidence that the decline in married-parent households caused inequality to increase, when the author herself disavowed that interpretation both in the paper and thereafter. She notes that, “the relationship between family formation behavior and inequality appears to be declining over time” and even during the period where it was most influential, it accounted for very little of the change. See her chart:
There are further problems with using Martin’s paper as the basis for the argument that changing family structure caused inequality to rise. She uses survey data that top-code the highest income households. That means that households earning more than some threshold are reported as earning at the threshold in order to guard their confidentiality. The Current Population Survey is considered to be informative about the bottom 98 percent of the income distribution (more or less), but the top is where all the action has happened, so to speak. Thomas Piketty and his coauthors have shown the rise of top income shares in numerous papers, and it is why they have developed the World Top Incomes Database to track the incomes of the richest.
In a recent summary of the literature, Bruce Western writes, “Most of the increase in family income inequality was due to increasing within group inequality that was widely shared across family types and levels of schooling.” He finds that, “Though family structure, more than the educational inequality in earnings, is closely associated with the rise in inequality from 1975 to 1995, both effects were small after 1995.”
There is a large literature that seeks to treat changing family structure as an outcome in need of explanation, exactly the sort of analysis that Martin stipulates her paper does not provide. Daniel Schneider and Adam Reich investigate whether young men who join unions are more likely to get married thereafter, and find strong evidence that they do in high-quality longitudinal data (meaning that the same individuals are followed over time). Importantly, the effect of joining a union on first marriage disappears when the authors also consider the increase in wages and job security that come with union membership. The implication is that the decline in overall marriage rates and the decline in unionization rates are linked. The reason why is that young men increasingly lack access to secure, well-paying jobs.
In a similar vein, Jennifer Lundquist finds that the military fosters high rates of marriage and low divorce, and notably those outcomes are indistinguishable for whites and blacks in the military, whereas in the larger society black households are significantly more likely to be headed by a single parent. Over a series of papers, she reports on the reasons why household structure is so much more traditional in the military: that employment is stable and that a safety net exists for raising children.
Consistent with the idea that economic wellbeing enables a stable marriage, Yu Xie, James Raymo, Kimberly Goyette, and Arland Thornton construct an individual-specific measure of “economic potential” designed to capture the outlook for young people that may not be discernible from current income or a crude predictor of future income, such as education credentials. They find that economic potential increases the marriage rate for young men, but not for women. On the other hand, potential does not affect the probability of cohabitation for either gender.
In a way, the idea that household formation decisions are an economic outcome should not be surprising. One of the foundational publications of modern economics, An Essay on the Principle of Population by Thomas Malthus, argues that the reason why the birthrate increases in prosperous times is that people have the wherewithal to marry earlier. Malthus argued that this effect eventually drives the standard of living back to equilibrium, which thankfully did not turn out to be true for the most part after the 18th century. But the bare point, that stable marriage is an aspiration that falls victim when economic opportunities worsen has been a notable feature of social science scholarship since, including the books When Work Disappears by William Julius Wilson and the recent, and excellent, Labor’s Love Lost by Andrew Cherlin.
Cherlin brings a historical perspective to bear: the last time marriage rates for working Americans were low and age at first marriage was high was during the Gilded Age in the late 19th century, when inequality was also high. He attributes the rise and fall of the working class family to the rise and fall of the manufacturing sector and stable employment opportunities for young men. He also points out that the working class family was a gendered institution, that it was premised on a male breadwinner earning enough to support a female home-maker, and that its decline is in parallel to the relative expansion of women’s opportunities for employment outside the home. Thus, we can see the prevalence of married households as reflecting both the absolute economic opportunity of men and women and the relative economic opportunity of women. On the latter point, Judith Hellerstein and Melinda Morrill find that divorce rates fall when unemployment is high, suggesting that when economic outcomes outside marriage are worse, spouses remain in marriages they would otherwise have left.
It is facile to divide rising inequality into “between” and “within” effects with respect to household types, and to argue that since inequality between types has grown and more households are now in worse-off types, changing family structure has caused inequality to increase. The evidence shows that family structure has changed because economic opportunities for most people have worsened. Why has that happened? There are some suggestive answers, but much more research is necessary. Leonhardt's claim that changing family structure causes rising inequality simply doesn't hold up.
The paper decomposes the change in a measure of inequality called the Mean Logarithmic Deviation as calculated from the March Current Population Survey into “between” and “within” effects.