In 1979, Martha Hill, a researcher at the University of Michigan, observed a strange fact about married men: they make a lot more money, at least compared to their unmarried peers. (According to Hill’s numbers, marriage led to a roughly 25 percent boost in pay.) What’s more, the effect remained even after Hill controlled for most of the relevant variables, including work experience and training.

In the decades since, this correlation has been repeatedly confirmed, with dozens of studies showing that married men earn between 10 and 50 percent more than their unmarried peers. (Because the world is so unfair, women get hit with a marriage penalty, as married females earn roughly 10 percent less than unmarried women.) What’s more, these income differences among men don’t seem to depend on any of the obvious demographic differences, including age, education, race and IQ scores. For whatever strange reason, companies seem to find married men more valuable. Their human capital is worth more capital.

But why? What does marriage have to do with work performance? A number of competing explanations for the male marriage premium have been proposed. There is the discrimination hypothesis – employers are biased against bachelors – and the selection explanation, which posits that men who are more likely to get married are also more likely to have the character traits required for career success. (If you can get along with your spouse, then you can probably get along with your colleagues.) Another possibility is that married men benefit from the specialization of their labor: because they don’t have to be worry about the dishes or other household chores – that’s the job of their partners - they’re more productive at work. Lastly, there is the marriage-as-education hypothesis, which suggests that married men might learn valuable skills from their marriage. In the midst of a long-term relationship, men might get trained in things like commitment and self-control, which are also useful at the office. If true, this means that the male marriage premium is rooted in something real, and that it has a causal effect on productivity. Companies are right to pay extra for men with wedding rings.

There have been a number of clever attempts to test these different possibilities. Economists have looked at the effect of shotgun weddings and the impact of a working wife on a husband’s earnings. They’ve looked at whether the gains of the marriage premium effect occur all at once or over time – most conclude the wage gains accrue over time – and that the premium dissipates as couples approach divorce.

And yet, despite this bevy of research, the literature remains full of uncertainty. In a new paper, published last month in Labour Economics, the economists Megan de Linde Leonard and T.D. Stanley summarize the current confusion. “Researchers report estimates of the marriage-wage premium that range from 100% to a negative 39% of average wage,” they write. “In fact, over the past four years, we found 258 estimates, nine percent of which are statistically negative, 40% are significantly positive” and many of which were indistinguishable from zero. So the effect is either positive, negative, or non-existent. So much for consensus.

To help parse the disagreement, de Linde Leonard and Stanley use a technique called meta-regression analysis. The statistical equations are way over my head, but de Linde Leonard was kind enough to describe the basic methodology by email:

"To do a meta-analysis, we search for all papers, published and unpublished, that have estimates of the phenomenon that we are interested in. We then record those estimates into a spreadsheet along with other important factors about the study. (Was it published? What data set was used? Was the data from the US?, etc.) Once that is complete, we use statistical analysis to draw out the signal from the noise. We don't only use studies that we consider to be best practice; we use all the studies we can find and let the data tell us what is true. That is the beauty of the technique. We don't have to rely on our (almost always biased) professional judgment to decide what is real and important. We let the body of research do the talking."

So what did the body of research say? After looking at more than 661 estimates of the male marriage premium, de Linde Leonard and Stanley settled on a 9.4 percent premium among male workers in the United States. (The effect seems to be less potent in other countries.) Interestingly, the male marriage premium seems to getting more powerful in the 21st century, as a review of the most recent studies finds an average premium of 12.8 percent.

What’s more, the meta-regression technique allowed the economists to assess the likelihood of various explanations for the marriage premium. Since the effect is increasing among men, even as the percentage of women in the workforce continues to increase, it seems unlikely that labor specialization plays a large role. (In other words, not doing the dishes doesn’t make you more productive at the office.*) De Linde Leonard and Stanley are also skeptical of the selection hypothesis, which suggests that married men only make more money because the men who tend to marry already possess the traits associated with high salaries. While the selection effect is real, the economists conclude that it’s not the main driver of the marriage premium, and probably accounts for just a 2 percent bump in wages, or less than 20 percent of the total marriage premium.

We’re left, then, with the marriage-as-education explanation. According to this theory, matrimony is a kind of college for the emotions, instilling partners with a very valuable set of non-cognitive traits. As De Linde Leonard and Stanley point out, marriage might cause men to “‘settle down,’ be more stable, and focused on work and career.” While we often draw a sharp distinction between the worlds of work and love, and assume that the traits and skills that are essential in one domain are irrelevant in the other, the marriage premium is a reminder that such distinctions are blurry at best. In fact, the talents that married men learn from marriage are roughly equivalent, at least in monetary value, to the income boost the average worker gets from attending college, but not graduating. (A bachelor’s degree gives people a much bigger salary boost.) Of course, women also probably pick up useful mental skills from matrimony, which makes the existence of the female marriage penalty – even if it’s just a penalty against having kids - that much more unjust.

And yet, despite the plausibility of the marriage-as-education theory, we know remarkably little about what’s learned from our closest romantic relationships. There’s some scattered evidence: men who score higher in grit are also more likely to stay married, and those with secure romantic attachments are also happier employees. But these are just glimpses and glances of what remains a mostly mysterious schooling. Besides, the greatest “skills” we learn from marriage (or really any committed relationship) might not be measurable, at least not in the psychometric sense. This is rampant speculation, rooted in my own n of 1 experience, but it seems that marriage can provide us with a valuable sense of perspective, stretching out the timescale of our expectations. We learn that moods pass, fights get forgotten, forgiveness is possible. We realize that everything worthwhile requires years of struggle (even love!), and that success is mixed up with the salty residue of sweat and tears. I have no idea how much that wisdom is worth at the office, but I damned sure know it helps with the rest of life.

*I’m actually partial to what might be called the non-specialization-of-labor hypothesis, which is that spouses often add tremendous value to one’s work. Call it the Vera effect.

de Linde Leonard, Megan, and T. D. Stanley. "Married with children: What remains when observable biases are removed from the reported male marriage wage premium." Labour Economics 33 (2015): 72-80.