Commenter Fenton raised an interesting point in response to my post Forfeiting The Patriarchal Dividend:
A hypothesis: men who lose their jobs (and sink to little or no income) are then often divorced, to free up the woman to find someone with more resources. The two statuses are linked not by motivation to work harder, but by female filtering: men with earnings are acceptable to marry, men without earnings are then divorced. I have no data with which to test this, unfortunately.
This question may not be of interest to many of my readers, but I wanted to dig into the data and see how much of this possible effect I could rule in or out. There are two ways this could be causing the group of married men to have a smaller percentage of zero earnings than the group of unmarried men:
- Instantaneous: In any given year a (mostly) random group of married men find themselves without earnings, and during that same basic time period their wife divorces them, moving them from the married to the unmarried category. Assuming a higher percentage of the men going through a divorce in any given year have zero earnings than the population of married men, that year’s divorces will cause the married category to look better and the unmarried category to look worse than they otherwise would.
- Cumulative: Over time wives are divorcing the least motivated men, and since these men are unmotivated they are remaining unmarried. This would cause the concentration of non earners to build up in the divorced category (especially in older age brackets), and the high concentration of non earning divorced men would increase the percentage of unmarried men with zero earnings.
How big is the instantaneous (same year) impact of divorce on the percentage of married vs unmarried men with zero earnings?
The instantaneous impact will be largest in the age bracket with the highest divorce rates, so divorce rates by age is the first piece of data we need. While I don’t have divorce rate data on White men in 2012, the Census put out a report for 2009 with a breakdown of the number of men (all races) who went through a divorce that year by age bracket. I combined this data with the number of married men in the same year to create the following table:
These figures are for all races in 2009, but they at least give an estimate I can plug back into my calculations for Whites in 2012. As you can see, only a small percentage of married men are experiencing divorce in any given year, but there is significant variance by age. Since very few 15-24 year old men are married, I’ll focus on the 25-34 year age bracket as it has the next highest divorce rate.
The next piece of data we need is the percentage of men going through a divorce in 2009 who had zero earnings. The same US Census paper includes this information as well, although unfortunately it doesn’t break it down by age:
Adding the percentage of men divorcing in 2009 who were on unemployment (9.5%) to the percent not in the labor force (17.2%) gives us a total of 26.7% with no earnings. This number almost certainly overstates the percent of White men age 25-34 divorcing with no earnings because:
- It includes men who are at or nearing retirement age, both of which have higher percentages of zero earnings than 25-34 year old men do.
- Zero earnings rates are lower for White men than men of all races.
From here we can estimate:
A) The number of 25-34 year old White men who went through a divorce in 2012. We get this by multiplying the number of married men in that demographic by the percent of 25-34 year old married men (all races) who went through a divorce in 2009 (2.9%).
B) How many of those divorces involved men with zero earnings (26.7% of the answer from “A”).
Once we have these two estimates, we can add the estimated number of men who went through a divorce in 2012 back into the White married category and subtract them from the unmarried group. This allows us to quantify the (estimated) impact of that years divorces on the results. Here is what the revised calculations look like for married men:
We do the same thing with the 2012 White unmarried category, but in this case we need to subtract the estimated divorces:
This tells us what the figures would have been had no divorces occurred in 2012 (again assuming the estimates are accurate). The married category would have had 7.2% of men with zero earnings (an extra .56%). The unmarried category would have had 15% of men with zero earnings (a difference of .28%).
Since the original total difference between married and unmarried White men was 8% for the 25-29 bracket and 8.5% for the 30-34 bracket the estimated combined instantaneous effect of (at most) .84% (.56% + .28%) would only account for a small part of the difference we are seeing. Since the later age brackets have lower divorce rates, the instantaneous impact would be even smaller for older men.
What about a possible cumulative impact of divorce?
The best way to measure this is by breaking down the unmarried category to its different sub groups (divorced, widowed, never married, etc) and calculate zero earnings percentages for each sub group. Fortunately this data is available by age bracket for White men in 2012. Since the biggest difference between married and unmarried men was in the 45-54 year old age bracket (and the cumulative effect would be greatest for the older age categories) I calculated the percent of each sub group of unmarried White men age 45-54 in 2012:
As I did with the charts in the original post, I’ve included “Separated” in the unmarried category because for Whites separation fairly quickly leads to divorce. See fig 32 of this study for more details. Interestingly the percent of divorced men with zero earnings was identical to the same figure for separated men.
Note also that divorced and separated men have the lowest percentage of men with zero earnings for the entire unmarried category. Divorced men are actually pulling the average down. This rules out a cumulative impact of divorce as increasing the percentage unmarried men with zero earnings. The other possible cumulative impact would be on married men. However, this is less clear cut because while it is clear that there is some selection going on regarding the men experiencing divorce in any given year, the new marital status also means less incentive to earn.
What is especially surprising is the high percentage of widowed men with zero earnings. At first glance it would seem that there shouldn’t be a correlation between a man’s earnings and his wife’s probability of passing away. However, since this is a 10 year age bracket what we might be seeing is older men being more likely to both have 0 earnings (either due to retirement or disability) and to have lost their wife. However, when I broke it out for just 45-49 year old white men the result for widowed men was nearly identical. It might also be that there is a correlation with lifestyle choices of the married couple that we are seeing, such as drug or alcohol abuse leading to higher than average levels of both early death and unemployment/disability. But it is also possible (I would argue likely) that this difference represents the formerly married man’s reduced incentive to work now that he no longer has a wife. My guess is both are in play here, but I won’t speculate on the relative weight of the two factors.
Some of the differences we see between married and unmarried men does seem to be explained by wives divorcing non earning men as Fenton hypothesized, but this impact is small compared to the overall difference observed and has the biggest impact on the younger age brackets where divorce rates are highest. The biggest difference we find in zero earnings rates between married and unmarried men is in the older age brackets, and for these brackets the cumulative impact of divorce is actually reducing the observed difference. All of this confirms that marriage does appear to motivate men to earn more than they would otherwise earn, and that sorting due to divorce isn’t what is driving the differences we see. The lower earnings rates for widowed men appears to further corroborate this, but it is possible that part of what we are observing for widowed men could be explained by joint lifestyle choices or some other factor.