“Fifty percent of marriages end in divorce.” You’ve probably heard that claim several times—just as you may also have heard from other sources that it’s inaccurate. As I’ll explain below, the real number is likely lower, but perhaps not by a lot. One thing is for sure. Arguments over what the divorce rate is and whether it’s dropping are ongoing and unlikely to end anytime soon.

Just last month, Claire Cain Miller argued in the New York Times that the divorce rate has been coming down for a long time even while the odds of divorce remain greatly exaggerated in the minds of many. She highlighted the conclusions of economist Justin Wolfers, who told her that “If current trends continue, nearly two-thirds of marriages will never involve a divorce.” In a follow-up piece, Wolfers explained more about the complexity of the issue and defended his claims.

Some go even further than Cain Miller, arguing that the likelihood of divorcing has never been anything like 50 percent. For example, Shaunti Feldhahn, the author (with Tally Whitehead) of a recent book on the subject, argues that it was never true that half of newly married couples would end up divorced, and that 30 percent is closer to the mark. While not a social scientist, Feldhahn has studied the history of the divorce rate and believes people are too pessimistic about the odds of success in marriage. Although I’m not persuaded that the risk of divorce is that low, I agree with her that many people avoid marriage for fear of divorce even when their own risks are quite low.

In contrast to those who argue that the divorce rate has been coming down, or that it was never that high, demographers Sheela Kennedy and Steven Ruggles argued in an article last year that divorce did not level off or decline in recent decades but actually continued to rise from 1980 to 2010. In fact, Ruggles commented on Cain Miller’s and Wolfers’ New York Times pieces, here and here, arguing that conclusions in both are likely incorrect and that most professional demographers have not accepted the notion that the overall risk of divorce declined during the period in question.

While these researchers may not agree about what has happened in past decades, they all seem to suggest that the risk of divorce has become lower, or is likely to be dropping, among those who are younger and marrying now. Kennedy and Ruggles examined an “age-standardized refined divorce rate” and found no support for an overall decline in divorce, but noted that this is largely due to the fact that divorce rates have continued to climb over the years among baby boomers in comparison to other cohorts (see also Susan Brown and I-Fen Lin).
Arguments over the risk of divorce are not new, which raises the question as to why there is so much room for disagreement.

At Any Rate, It’s Confusing

Kennedy and Ruggles titled their paper “Breaking Up Is Hard to Count: The Rise of Divorce in the United States, 1980–2010,” and with good reason. They lay out the history of challenges in tracking divorce, detailing issues about public records, differing data sets, and various cohort issues. Wolfers’ New York Times article and the comments by Ruggles further illuminate the vast complexity facing researchers who try to come up with definitive statements about the risk of divorcing.

I will not attempt to capture all that complexity here, but I can focus on a couple of the reasons why this subject is so confusing to most people. Let’s start with the fact that there is no single metric on divorce. There are many. One of the simplest is the number of people who divorce, per year, per 1000 people in the U.S. (the so-called “crude” divorce rate). By this measure, the divorce rate peaked at 5.3 divorces per 1000 people in 1981 (CDC) and has come down steadily since to 3.6 in 2011 (CDC). A variation on this type of metric is the number of people who divorce per year, per 1000 married women—such as used in this National Marriage Project report.

Another simple metric is the percent of people, at any given time, who have already been divorced. For example, based on their 2007-08 national survey, the Barna Group found that 33 percent of ever-married adults, and 25 percent of adults overall, have experienced divorce. Even these seemingly straightforward numbers can be difficult to interpret because of societal changes in the number and nature of people who marry today compared to the past.

Predicting the Lifetime Divorce Rate

All three of these metrics are different from the likelihood of divorce for a couple marrying for the first time, which is what the oft-repeated “50 percent chance of divorce” is about. Coming up with a number for lifetime divorce risk is crazy complicated because it’s a projection about what will happen in the future based on what has happened in the past. As I understand it, those demographers who have constructed such projections do so based on careful analyses of the likelihood of divorcing in various years of marriage based on the history of divorce within existing samples.

It’s hard to track down the first source of the 50-percent statistic, but it seems to originate from projections of this sort made by scholars in the early 1980s, around the time when the crude divorce rate was peaking. For example, in 1984, Paul Glick published a study saying, among other things, “About one-half of the first marriages of young adults today are likely to end in divorce.” Subsequent projections, like this 1992 projection by the Census Bureau, came up with similar estimates—but each projection only applies to couples marrying at the time the projection is made.

Such era-bound estimates are as good as researchers can do, because no one can know the precise lifetime risk of divorce for those marrying right now. Here’s one illustration showing why that’s the case. Suppose we undertook a study following a representative sample of 20,000 people from birth to death, while gathering complete marital histories along the way. We will not know exactly how likely our subjects are to divorce until all of them are dead (or, technically, until all are dead, divorced, or widowed—that would work, too).  When we get there, the number for the lifetime divorce risk will be rock solid.

What’s wrong with this mythical study? A lot. First, it would be extraordinarily expensive and difficult to follow such a sample without losing track of people. Two, the original researchers will be dead by the time the answer comes in. (This dampens enthusiasm to start the study.) Three, once you get this robust answer about the likelihood of divorcing, it’s old news. The answer will apply to a generation that has almost entirely died out, not to those who are young when the study ends. People want to know the future, not the past.

Moreover, all projections of this type are affected by societal trends that can change—and a lot has changed in recent decades when it comes to marriage and divorce. For example, those at lower levels of income and education are less likely to marry than in the past while those with college degrees are the most likely to succeed at marriage. Glick noted this in 1984, and, in my favorite paper about the risk of divorce, R. Kelly Raley and Larry Bumpass showed in 2003 that this differential increased from the early  to mid-1980s to the mid-1990s. It may be increasing still.

The Lifetime Divorce Risk for Newlyweds Today

Even if projections about divorce are always tentative and subject to change, many will want to know: If the 50-percent statistic dates to the 1980s and there is some evidence that divorce rates have declined for those getting married now, what’s the right number for today?

I periodically ask sociologist Paul Amato what he believes a solid prediction would be for couples getting married now for the first time, and I did so again last week. He noted that it is, indeed, “difficult to know what’s going on with the divorce rate.” But taking everything he knows into account—including the most recent elements of the debate noted here—he believes that the lifetime risk of divorce today is 42 to 45 percent. “And if you throw in permanent separations that don’t end in divorce,” he added, “then the overall likelihood of marital disruption is pushing 50 percent.”

Amato relies a good deal on the calculations of Robert Schoen and Vladimir Canudas-Romo (2006), and their conclusion that “it is premature to believe that the probability of divorce has begun to decline.” But he hastened to add that it is very difficult to predict the future divorce rate. Nevertheless, he noted that young married adults are not divorcing at the same rate as their parents did at similar ages, so it is likely that the divorce rate will decline in the future, once the Baby Boomers (who were and continue to be highly divorce prone) leave the population. Thus, as others have suggested, the future may well be brighter than the 42 to 45 percent risk estimate suggests, but we do not yet know this will be the case. And there are factors that work in both directions; for example, as Wolfers noted, health gains mean people are living longer, which also means added years for the possibility of divorce.

Whatever the future holds for the risk of divorce, divorce isn’t the only family stability metric that matters today (a fact that Raley and Bumpass, and others, have emphasized). While the divorce rate for young couples starting out in marriage may be coming down, I believe that the percentage of children impacted by family instability keeps going up due to the combination of divorce and never-married parents breaking up (as the majority of them do by the time their child turns five). This is why I have written that we may be approaching a perfect storm with regard to children and attachment insecurity, and that the timing of marriage relative to childbearing remains a big deal. As sociologist Andrew Cherlin has argued, American families have become marked by turbulence and churning, and this is not without consequence.

Naturally, young people worry less about societal trends than about their own likelihood of divorcing, a worry that leads some to avoid marriage altogether. Of course, that clearly does not mean avoiding the pain of breaking up. Many others who are already married wonder if they will make it. There is, however, some good news in all this. For example, there are things people can do to lower their own risks of divorce and to increase their chances of having a lasting, loving marriage. And there are many people who are at a substantially lower risk of divorce than they think—a key point argued by people such as Feldhahn. Projections do not have to be destiny. I’ll take up that subject the next time.

Update, 1/23/15: The third-to-last paragraph was added to this piece following further correspondence with Paul Amato.