In a Freakonomics editorial published on the New York Times website, economist Justin Wolfers laments the inaccuracy of some widely quoted divorce statistics, which paint a bleaker than warranted picture of the state of marriage in America. Specifically, Wolfers takes aim at a frequently reported 43% figure, supposedly the likelihood of a woman’s first marriage ending in divorce or separation before the fifteenth wedding anniversary.
Wolfers identifies an obvious flaw in the analysis of data (Table 2) from a 2004 US Bureau of Census survey, leading to the above, significantly overestimated divorce rate. When properly examined, the data yield in fact an estimate of the incidence of divorce during the first fifteen years closer to 33.4% . The same procedural mistake leads to an inaccurate estimate of the rate at which divorce occurs within the first 25 years of marriage, as noted by Wolfers’ collaborator and colleague Betsey Stevenson. In this case, however, the difference (47%, as opposed to the initially assessed 51%) hardly seems of any statistical significance.
Wolfers has argued elsewhere that the divorce rate is actually declining, and that, contrary to popular belief, marriage as a whole is showing in the US clear signs of increasing stability.
I am puzzled by Wolfers’ analysis, and really do not share his enthusiasm, based on the data that are available. Focusing on a specific anniversary (e.g., the 15th) is only meaningful to the extent that the probability of divorce past that mark becomes negligibly small. Otherwise it scarcely makes sense; divorce is divorce, and it is certainly no less painful if it happens after 20 years than after 5 (in fact, I would argue that a late divorce can be much more devastating).
As it is, the 33% 15-year figure seems to mean very little, as the likelihood of a marriage to end prematurely goes up by as many as 14 percentage points, if the first 25 years are considered  (that a downward revision from 51% to 47% should be cause for much excitement, is a matter of opinions — the usual thing with seeing the bottle as half-empty or half-full, I suppose). The fact is, the “one-in-two” figure that I have often heard in conversation, and which I had always thought of as just another “urban legend”, appears instead to be appallingly accurate.
To me, that a marriage should collapse after 15 years is most disconcerting, as I would naively have regarded a marriage that has lasted that long, as fairly solid. Evidently, however, that is not the case. Who knows, maybe, more often than I thought, men and women endure difficult marriages for the sake of providing a stable family environment to their growing children, but decide to call it quit as soon as sons and daughters have left home.
But more importantly, even if the divorce trend is reassuring , as Wolfers (rightfully, I am sure) claims, the fact remains that, if one is to assess the robustness of marriage as an institution, divorce rate is only one aspect to consider. For example, data from the Statistical Abstract of the United States show that, while the number of divorces per 1,000 population has been steadily declining since 1981, the ratio of divorces to marriages in a given year has held remarkably steady at around (slightly below) one divorce for every two marriages, over the same time period.
I am no expert but this seems to suggest that not only divorce, but marriage as well is in decline, i.e., more people opt out of it as a life choice. But, a significant fraction of couples choosing co-habitation over formal marriage  hardly lends support to the notion that marriage is gaining traction. Furthermore, it may conceivably lead to the same societal outcomes often associated to marriage crisis (e.g., children raised by a single parent), even with falling divorce rates.
Again, I am no economist, and would be hardly qualified to interpret data anyway; I have found articles in the popular press suggesting that the above hypothetical scenario of marriage rates falling concomitantly with those of divorce, may indeed be taking place, and may well have a significant effect1,2,3. Be that as it may, I think that “stability of marriage” is a complex issue to assess; especially in light of societal changes that have occurred over the past two decades, any single indicator, such as divorce rate, is likely to offer an incomplete and misleading picture. Even if those who enter nowadays into a marriage do so (as data suggest) more responsibly, and with a stronger commitment to making it work than before, this may be of little significance if at the same time considerably fewer people marry in the first place.
 The problem with the 43% figure is the following: A survey conducted in the Summer of 2004, found that 43.1% of women who had married between 1985 and 1989 had not reached their fifteenth wedding anniversary. Obviously, for approximately 10% of those women this would have been impossible, as their marriage was celebrated in the second half of 1989, i.e., less than 15 years prior to the survey itself; yet, apparently these women have been counted in several studies as divorced or separated. A straightforward correction, excluding 10% of the sample from the count, brings the number down to 36.8%. The lower figure quoted by Wolfers comes from a more detailed analysis of the same sample (ostensibly, people marry more frequently in second half of the year…)
 Wolfers makes the point that there is further ambiguity with the interpretation of the data from the survey, for one should also revise the figure to exclude those cases in which marriage ends with the death of one of the two spouses, again often unduly counted in as “failed marriages”. However (as Wolfers himself admits), such a correction is likely to be small, as the median age of marriage is around 25, and passing before 50 is relatively uncommon. And actually, the “death” argument can be also turned around. For, because life expectancy has been steadily increasing, these days death of a spouse ought to end fewer marriages before the 25th anniversary, than it did, say, 50 years ago. The fact that the rate at which marriages end before the 25th anniversary does not show the same convincing downward trend than that for the 15th one, is actually disheartening.
 I am not convinced that the above-mentioned Bureau of Census data unambiguously point to an overall decline; they could also be interpreted in terms of an increased tendency for divorces to occur, on average, later in the marriage than in the past. In that sense, data for the 25-, 30- and 35-year marks may be more telling, and it is far from clear whether they show any decline, especially if the increased life expectancy is factored in.
 Obviously, this makes it all the more difficult to assess the duration of a relationship, as often these living arrangements, and their terminations, are not legally recorded.