Misreporting on Divorce

Today is apparently D-Day here at Freakonomics — the “D” stands for divorce. Along with Hamermesh’s earlier post and this post by Wolfers, there’s one more on the way.

One of the most frustrating things about doing research on families is seeing how often even the simple facts are misreported in the press. And Sue Shellenbarger, writing in this morning’s Wall Street Journal “Work and Family” column, provides yet another example.

First is a clear mistake: “On average, 43 percent of first marriages end in separation or divorce within 15 years, a federal study shows.” I can only assume that she is citing Census Bureau data, showing that 56.9 percent of women married from 1985 to 1989 had reached their 15th anniversary, and from that, inferring that 43.1 percent had divorced or separated within 15 years of marriage.

But as Betsey Stevenson and I noted in this New York Times op-ed, these federal statistics were widely mis-cited. The problem is simply that about one in ten of those who were married from 1985 to 1989 had not reached their 15th anniversary by the time of the 2004 survey, simply because their marriage occurred less than 15 years ago!

In this paper, we re-calculate the equivalent probabilities analyzing only those who had actually been married at least 15 years ago, and find that 33.4 percent of these first marriages ended before their 15th anniversary. Or for those who don’t like our attempted “fix,” I would suggest referring instead to an earlier census analysis (where this problem didn’t occur), which found that 34.8 percent of first marriages ended before their 15th anniversary.

There are two other minor problems with Shellenbarger’s numbers. First, the 43 percent number she cites is based only on the reports of women, while the same federal study notes that the equivalent number for men is a less dramatic 39.4 percent. Applying our corrections lowers these numbers to 35.7 percent for women and 31.2 percent for men, respectively.

And second, some of these couples failed to make it to their 15th anniversary because of death, rather than divorce. Whichever way you look at it, the relevant statistic is surely closer to 33 percent than 43 percent.

The article begins with this motivating fact: “As the national divorce rate plateaus at historically high levels … .”

“Historically high levels”: Yes. “Plateaus”? No way: the divorce rate in the United States is currently at its lowest level in twenty-five years, and has fallen nearly every year since 1979.

The number of divorces per thousand marriages has now fallen by 27 percent since the peak in 1979. The latest data suggest that the divorce rate for 2007 will be even lower still. And our own analysis of the stability of marriages suggests that those married in the 1990’s appear to be less likely to divorce than those married in the 1980’s, who in turn are less likely to divorce than those married in the 1970’s. As such, the divorce rate seems likely to continue to decline for some time yet.

Those interested in a broader assessment of the facts and what may be driving changes in our families may find our Journal of Economic Perspectives paper to be useful. (Also, see my earlier posts at Marginal Revolution: here and here.)


Laurie, this data set answers just that since it only deals with first marriages. If out of 100 first marriages, 33 end in divorce, that indicates that an individual's odds of divorcing *at least once* are 33/100 = 33%.

Granted, the number may be higher than 33 (some people will divorce after being married for 15 years) or lower than 33 (spouses may die), but the math still works.

I don't think one could ever calculate the probability for a particular individual, but you ought to be able to refine the data set by controlling for age, race, gender, religion, income, area, etc.


I wonder if someone will ever calculate the likelihood of an individual getting a divorce. What I mean is, my mother-in-law has been divorced three times, my father-in-law has been divorced once, and neither of my own parents have ever been divorced. That means that if you look at that tiny sample-set, 50% of those people have gotten divorced, and 75% of those marriages ended in divorce. It seems that if you factor in the people who have multiple divorces, you're giving a skewed picture of how likely it is that a person's marriage will succeed. What's the percentage of people who will go through a divorce (or more than one) in their lifetime? Isn't that a more relevant way of looking at it?


For the answer to how men and women can have different stats just think of it this way. The stats are only looking at people that are EVER married and thus exclude people that are NEVER married. So since this isn't a closed system both men and women will (possibly) have different numerators. There are a whole host of reasons. I would speculate a guess that it is a lot easier for women to end up in the never married category because they outlive men and typically out number men as a proportion of any cohort. I'd also guess that along the lines of the original post, that more men "end" a marriage with a death (and thus get subtracted from the numerator again) than women.


which is worse, death or divorce?


If we are talking about first marriages, divorce rates can differ between men and women because some marriages are the first for one spouse and the nth for the other spouse.

If you accept these statistics as valid, you should be able to determine whether the most stable marriage is between two first timers or a first time bride and veteran groom or vice versa.


One Mormon divorce will be counted against one man and several women, skewing the data.

(Yes, I know this is wrong in at least three different ways. Don't bother to point it out.)


These sorts of studies just reek of stupidity. No offense, but divorces are just as much a matter of public record as marriages. Why not simply track how many marriages end in divorce? Why bang your head against the wall trying to massage the numbers to tease divorce out of them, when you can just look at the original data? It's not like we're trying to discern something hidden--like, say, the number of times per week married couples have sex--from public records of marriage and divorce. Don't invent statistical methods to guess at something that can be measured directly.

Adam S

Not debating your statistics. Just trying to understand how there can be a different divorce statistic between men and women (assuming we are only talking about hetrosexual marriage here since you are citing census data). I thought you had to have both to have a marriage.

Dread Pirate Robert

How can the number of devorces be different for men then women?
Is the difference the number of spousel deaths?

Charles D

Excellent whistle blowing on bad stats.


Matt at 12.28, you can only use the data if you can afford to collect it. If you use Census data, it's free but it may not be asking exactly the questions you want--thus you fiddle with the data.

Divorce stats always lead me to wonder about the percentage of people who get divorced, as opposed to the number of marriages which end in divorce. That is, the situation "Alice divorces twice, Brenda never divorces" is different from "Alice and Brenda each divorce once".


Re:"Don't invent statistical methods to guess at something that can be measured directly."
I'm trying to understand your post, but I can't. What are you talking about? Have you considered that divorce doesn't occur at an even rate along time? I think the article does a good job of demonstrating for itself why you can't just look at the original data.


Is spousal murder both death and divorce?


Perhaps the decrease in divorce rate is due to the fact that a large number of couples today first live together without marrying. Presumably a large number eventually decide they don't want to marry, thus lowering the divorce statistics (and saving a lot of lawyer fees.)


If it is true that divorce rates are decreasing over time, I wonder how much of that is due to the increasing age of first-time marriages?


Age at first marriage is the best single predictor of marital stability. The increasing age at first marriage for Americans is undoubtedly a major factor in the declining divorce rate.

I have one question for the authors, though: where are your data on refined divorce rates (divorces per 1,000 marriages) coming from? The Census Bureau and Statistical Abstract sources you cite do not report refined divorce rates -- they report the number of divorces, and the crude divorce rate (divorces per 1,000 population).


#7 Asks "which is worse, death or divorce? "

I think that's an interesting question.

Why doesn't the government tax divorce?
It would certainly curb divorce.
They already tax death.

I guess the divorcee's can still complain.

Karen Travis

Why do we care about divorce in our society? I submit that we see it as a problem or somehow detrimental to family life which is the bed-rock of our society. So why aren't we more concerned about prevention of failure in marriages, and look for ways to support, sustain, and protect couples from the pain and expense of divorce? We have access to many premarital preparation programs, such as Engaged Encounter which allow couples to work through issues they will have to deal with after they marry. We also have enrichment programs such as Marriage Encounter that help couples communicate better on issues such as money, in-laws, children, work/leisure time use, etc. A Pew Foundation survey some years ago found that the divorce rate among couples who had completed a Marriage Encounter weekend was only 2%, significantly less than for those who had never been. So if you know of a remedy that is 98% effective, with no known bad side effects, doesn't cost too much, and is fun besides, wouldn't you use it and recommend it to others?



Re: Maurica,

Divorce rates are higher for couples who live together before getting married. A quick google search gives this webpage http://marriage.about.com/od/cohabitation/qt/cohabfacts.htm, and that is just one of many out there.


What % of relationships end in marriage? Is this number increasing?

Even IF more marriages end in divorce, might this just be because more people are getting married when in the past they wouldn't have taken the plunge? Is this necessarily a bad thing? Sure, it devalues marriage, but I don't see an obvious problem with this. It depends on your feelings about marriage and what it should be. It seems to me that if we accept ANY divorce rate, then a higher divorce rate is, by itself, meaningless.