« Justin Wolfers to Guest Blog! | Main | Ad-supported cell phones »
The Divorce Myth
I want to start my week guest blogging by talking about divorce. Betsey Stevenson and
I had an
op-ed in yesterday’s New York Times noting a very simple fact: those
married in the 1990s have proved less likely to divorce than those wed in the
1980s, which were less likely to divorce than those wed in the 1970s. The
Divorce Facts are that divorce is falling, and marriages are more stable.
What is surprising, is just how easily and how often the
Divorce Facts lose out to the Divorce Myth. The Divorce Myth is that divorce
is rising. When the latest
divorce numbers came out last week, they once again confirm this
quarter-century long decline in divorce, but the media (including the Times,
Post,
and the Inquirer)
chose instead to write (incorrectly) about rising divorce. (In their defense, the data were presented in
a way that invited misinterpretation, a subject that I shall return to in a
future post.)
Why the persistence of the Divorce Myth?
- Blame the public for underestimating divorce: Tyler has argued that Americans “underestimate the probability of divorce”, and so when the statistics show that divorce is quite common, they infer divorce must have risen.
- Blame the public for overestimating divorce: Greg Mankiw thinks that this “seems be an example of what Bryan Caplan calls ‘the pessimistic bias’, a tendency to overestimate the severity of economic problems.”
- Blame the press: Mankiw may be a bit unfair on Joe Citizen: the average person gets their news from the press, and in this case, the press reported falsehoods as facts.
- Blame the politics: We argued that “Reporting on our families is a lot like reporting on the economy: statistical tales of woe provide the foundation for reform proposals. The only difference is that conservatives use these data to make the case for greater government intervention in the marriage market, while liberals use them to promote deregulation of marriage.”
- Blame the
professors: Academics are meant to provide the facts offsetting the
political hacks. But we don’t. Economists have had too little respect for
simple facts; publication glory lies with grand theories. Ideologically-motivated profs teaching family
sociology or family law would rather reinforce the Myth than offset it.
Personally, I go for #4 causing #3, unchecked by #5, and
would love to see research by Bryan testing #1 v. #2. Your thoughts?
Posted by Justin Wolfers on September 30, 2007 at 10:28 AM in Current Affairs, Data Source, Economics | Permalink
TrackBack
TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c66b253ef00e55097eb608834
Listed below are links to weblogs that reference The Divorce Myth:
Comments
Just a small nit to pick with #2. I think that the people that fall into the pessimistic bias tend to see all of society in a pessimistic way. I doubt that most people consider divorce and economic issue, but a moral/ethical one. I have not read any of Bryan's works, but I was under the impression that he was referring to more than just economic pessimism. As a child of a failed marriage from the 70's, I can't help but ask what you think were the reasons for the spike of divorces in the 70's?
Isaac Crawford
Blogging in Yemen
www.isaharr.com
Posted by: Isaac Crawford at Sep 30, 2007 10:52:17 AM
Hold on a sec. Before we list putative explanations for the data, let's agree on the existence of the phenomenon. Suppose there's a constant p(divorce) annually for any marriage. It follows that a marriage that began in the distant past is more likely to have ended in divorce than a marriage that began, say, yesterday. That explains the 'falling' divorce rate. I know you're too smart to miss this, so -- what am I missing?
Posted by: Don at Sep 30, 2007 11:09:41 AM
1) I think you're measuring the wrong stat to examine the pervasiveness of divorce.
2) I think your observed decline in the divorce rate is entirely attributable to the decline in the rate of marriages.
The narrative of rising divorce is also completely at odds with counts of divorce certificates, which show the divorce rate as having peaked at 22.8 divorces per 1,000 married couples in 1979 and to have fallen by 2005 to 16.7.
That certainly looks impressive. However, the rate of marriages over that same time went from 10.4 in 1979 to 7.5 in 2005 (See this SD report which has the US marriage rate in Table 1)
From that same report, divorce rates/1000 went from 5.2 in 1979 to 3.6 in 2005. So the marriage rate declines by 2.9 and the divorce rate by 1.6.
Also FYI, by my calculation that's a decline in annual divorces/marriage from 50% to 48% which seems like an inconsequential drop to me particularly as in 2002 it was 4.0/7.8 = 51% which would represent an actual increase in the divorce/marriage ratio.
Personally, I think that marriage as an institution is declining, specifically, what we're observing is an across-the-board decrease in the desirability of marriage.
Posted by: Jody at Sep 30, 2007 11:12:18 AM
The General Social Survey gives us a crude way to test #1/#2.
From 1972-1982 (the only time the samples seem comparable), the GSS asked for two pieces of information:
1. Whether the respondent had been divorced in the past year
2. How many of the respondent's relatives had been divorced in the past year
Out of 2,960 respondents (40 of the total sample of 3,000 did not answer the question), 85 had been divorced in the past year. Thus, there was a 2.87% chance of having been divorced over the past year (assuming nonrespondents are similar to respondents).
Out of the full sample of 3,000 respondents, 426 of their relatives had been divorced in the past year. This leaves us with a 14.2% chance that a relative had been divorced.
Assuming that the respondents knew whether they had been divorced themselves, but were less certain about relatives, we can use the chance that a relative has been divorced as an extremely rough proxy for public perception of the divorce rate.
The ratio of relatives divorces to personal divorces is 5 to 1. If the average individual has more than five relatives, then it's likely that the average person is underestimating the prevalence of divorce. The opposite would of course be true if the average individual has fewer than five relatives. I don't have any solid data on that, but my intuition says that the average person has more than five relatives (lending credence to Tyler's hypothesis in #1).
Disclaimer: I don't know enough about the GSS to know who was asked these questions. If the question was only asked of married people or people that had been married, that would skew the results. The data are also 25-35 years old.
Data: http://www.icpsr.umich.edu/GSS/rnd1998/merged/indx-sub/maritals.htm
I used the variables DIVORCE5 and DIVREL1 for 1972-1982.
Posted by: Steven Bass at Sep 30, 2007 11:57:50 AM
Jody --
The statistic Wolfers cites is per married couple, not per population.
Posted by: Isaac at Sep 30, 2007 12:14:24 PM
Issac - I understand that. But I'm arguing Justin is looking at the wrong statistic.
If the rate of marriage is declining (which it is, dramatically) and divorces are more likely to occur in new marriages than long standing marriages (which is the case), then a declining marriage rate will give the appearance of marriage becoming a more stable institution even if the probability of any particular marriage ending in divorce remains constant which is what the annual divorce/marriage ratio shows (staying right at 50%).
Posted by: Jody at Sep 30, 2007 12:26:27 PM
1) Re Jody's point re declining rate of marriage: If many couples that in the past would have married in haste and then divorced now don't marry, then it is possible that we are getting a falling divorce rate simply from a higher proportion of marriages today being among people who are older at time of marriage and who have thought it through a bit more.
2) Re who is to blame for the misperception: I would favor #3 and #5. I would be surprised if 1 reporter in 100 had any stats training. I do not believe (and am too lazy to check) that any even Masters in Journalism program has a required Stats course, let alone 2 or 3. The reason the profs (#5) share the blame is because they are not providing the training for the journalists, nor timely, succint commentary. We may be seeing the start of a process of rectification of this with blogging, but that is still a haphazard process with a limited audience.
Personally, I would love to see the major papers appoint a statistical editor who would review all stats stories to help the reporters understand what they are reporting on, and who would add an editor's comment box pointing out the limitations of the reported study.
Posted by: Acad Ronin at Sep 30, 2007 12:35:50 PM
I've always thought that the first cause of divorce is marriage.
Posted by: Chaps at Sep 30, 2007 12:38:47 PM
A simple graph would be very helpful. Has the recent decline taken to levels of divorce circa 1965?
Posted by: eric at Sep 30, 2007 2:28:30 PM
The fraction of children not living in two parent households has doubled since 1970. Since this is the social downside of both unmarried mothers giving birth and divorce, it is not surprising that people think both are increasing. Whether couples live together and them separate or marry and then divorce has little social consequences if there are no children. The measure of the overall divorce rate is not only ambiguous but not really that important.
Posted by: joan at Sep 30, 2007 3:22:49 PM
Let's perhaps clarify the question that we are asking here. To say the divorce rate is declining seems to me to be saying something like:
Of those people who get married, it is less likely today that after X years of marriage a marriage will end in divorce than it was for a marriage of X years to have ended in divorce in the past.
Put more concretely: In, say, 1990, how likely was it that couples married in 1980 were divorced vs. in, say, 2000, how likely was it that couples married in 1990 were divorced? If the latter number is lower, then the divorce rate can be said to have fallen.
This would seem to take into account the lower rate of marriage concern.
And it's also important to note that a "falling divorce rate" (understood as the percentage of marriages that end in divorce) is compatible with a decline in marriage as an institution if the rate of marriage is falling, and especially if it's falling faster than the divorce rate. It is possible that we are seeing fewer but "better" marriages.
That would be consistent with the general decline in the benefits of marriage as the gains from specialization have dramatically decreased with the increase in the female labor force participation rate and the closing of the gender wage gap. The reasons to marry have much more to do with personal happiness than broadly economic considerations. Thus, we may well be seeing fewer but better marriages.
Posted by: Steven Horwitz at Sep 30, 2007 3:24:07 PM
I find this thread encouraging. There have been several possible statistical explanations given for the change in divorce rate. I don't think we have, as yet, seen any response which provides clarifying statistics.
If this data is forthcoming it will complete the circle that an interactive forum provides. This is a big improvement over the prior model - the pundit speaks, we read it in the media. Maybe someone gets a letter to the editor published which comments (150 words or less, please).
If writers are going to become aware that their statements are going to be examined in detail, one can hope that there will be less polemicizing and more attention to backing their claims. This should improve the quality of the discourse for everyone.
It is not even necessary for the writer to participate online directly, the online community is perfectly willing to go after the traditional media when they feel something has been misrepresented.
The open questions in this discussion seem to have to do with the relationship of the divorce "rate" to the marriage rate and to the longevity of the marriages. A politician in Germany has just proposed that all marriages should only be for seven years.
As to the rise of single mothers. That is an interesting area of its own. Some of this rise seems related to the disappearance of social stigma to being a single mother and the ability of women to have a career. Another factor is that many women want a family, but don't like the men that they have to chose as husbands. So they have a baby, but not a husband.
There is a famous case of a single mother from the 1940's which shows how much attitudes have changed. Laura Hobson (who wrote "Gentleman's Agreement") had a child out of wedlock. She had to vanish for a few months while pregnant and then had to "adopt" her own child in order to make the event socially acceptable. She was a highly paid writer and advertising executive so she had the financial resources (and friendly lawyers and doctors) to help her pull this off. Even the idea of an unmarried woman adopting a child was unusual. This, despite the fact that she already had another child (really adopted).
Posted by: robertdfeinman at Sep 30, 2007 4:10:02 PM
Actually, Jody, you're missing the other statistic Wolfers and Stevenson mention in the op-ed:
"For instance, marriages that began in the 1990s were more likely to celebrate a 10th anniversary than those that started in the 1980s, which, in turn, were also more likely to last than marriages that began back in the 1970s."
Since Justin is fixing the endpoint, then that automatically takes care of your worry that he's overcounting older, more stable marriages.
The total sample of marriages that started in 1995 and we examine today at 2007 are doing better than the total sample of marriages that began in 1975, when we evaluate their status in 1987.
Based on the comments so far, I'm really voting for the pessimistic bias.
Posted by: Keith at Sep 30, 2007 5:16:23 PM
For instance, marriages that began in the 1990s were more likely to celebrate a 10th anniversary than those that started in the 1980s, which, in turn, were also more likely to last than marriages that began back in the 1970s.
I want to see the actual numbers. Based on the stats I have handy the drop may be trivial.
Posted by: Jody at Sep 30, 2007 5:19:54 PM
I generally interpret "divorce is rising" as equivalent to "women have more opportunity", e.g. as good news.
Posted by: michael vassar at Sep 30, 2007 7:34:32 PM
I find this "Divorce Myth" very hard to believe. Now, these facts that came from the New York Post, who were the people that were targeted? I think that maybe the groups of people who were looked at were a certain population; something had to have been biased in this. There is no way divorce has decreased over the years.
Posted by: Chelsey at Sep 30, 2007 8:26:05 PM
"I want to see the actual numbers. Based on the stats I have handy the drop may be trivial."
Well, you may not be able to claim that divorce is falling based on "trivial" drops. But you sure as heck can't claim that divorce is rising. At best, you're down to arguing that the probability of divorce stayed the same over 20 years.
You appear to concede now that the 10-year measure adjusts for your earlier overcounting concerns, and that the measure shows at least "trivial" drops in divorce over time.
Posted by: Keith at Sep 30, 2007 8:38:12 PM
I don't know about reason #5. Academic demographers have been documenting the changing divorce rate for some time now. They've been providing these facts and discussing them actively. They don't go out of their way to contact the media about it, but the info. is easy enough to find.
Posted by: at Sep 30, 2007 11:35:27 PM
You appear to concede now that the 10-year measure adjusts for your earlier overcounting concerns, and that the measure shows at least "trivial" drops in divorce over time.
Actually, I'm saying that the other stats may address the concern, but it would contradict my cited numbers, so I want to see what the actual values are as there was also a trivial drop from 79-05 in my numbers. It may be a difference in methodologies, but that would interesting in and of itself as mine says flat and the unenumerated trend says decline. However, if unenumerated trend's decline is also trivial, then there's no tension between the methods.
Posted by: Jody at Oct 1, 2007 12:01:34 AM
I'm confused. Let's say that most failing marriages fail after twenty or thirty years. It's only been 17 years since 1990, so none of those marriages have failed. It's been 27 years since 1980, so most of the failing marriages have already failed. It's been 37 years since 1970, so all of those failing marriages have already failed. Nothing has changed, and yet there are more divorces among the 70's marriages, somewhat less among the 80's marriages, and just a few of the 90's marriages.
Is that the evidence that's being used to show that the divorce rate is dropping?
Posted by: Russell Nelson at Oct 1, 2007 1:42:26 AM
Russ, no. From the New York Times article:
"For instance, marriages that began in the 1990s were more likely to celebrate a 10th anniversary than those that started in the 1980s, which, in turn, were also more likely to last than marriages that began back in the 1970s."
So they see how couples that got married in 1970 were doing in 1980, compare that to couples that got married in 1980 and how those couples were doing in 1990, and compare those to couples that got married in 1990 and how those couples were doing in 2000.
So it's apples-to-apples.
Posted by: Keith at Oct 1, 2007 7:58:35 AM
Actually, Justin reported the stats this morning in the new post, and he's just interpreting the statistics wrong as it went from 76% down to 70% for 10 year survival with both cohorts affected by the same undercounting phenomenon.
Posted by: Jody at Oct 1, 2007 8:40:22 AM
All Tyler said in that post was that people underestimate the probability of their own divorce, not that they underestimate divorce rates in general.
Posted by: mobile at Oct 1, 2007 9:14:12 AM
Thanks Keith, I missed that.
Posted by: Russ Nelson at Oct 1, 2007 9:22:36 AM
In response to http://www.marginalrevolution.com/ blog byJustin Wolfers on Divorce Myth:
In result to Mr. wolfers blog I choose #2 Blame the public for overestimating divorce: and # 3 Blame the press in my opinion neither one of the sources really understand the meaning of devorce and why they occur. Some are personal issues, some are related to cheating and alot are realated to finatual problems. It is proven in the numbers that each ten years the divorce rates go down, from the previous ten years before that. I agree with #'s 2 and 3 beacsue in my opinion it relates back to exactly what Mr. Wolfers said, In their defense, the data were presented in a way that invited misinterpretation so there for the press blows it up and then the public overestimates divorce rates. Mostly beacuse the public eye believes every thing they read and see. So in result that makes the public and the press make it seem as though divorce rate has gone threw the charts. When in the real life sinario we should be looking at pre marital kids and those rates, there is way more of them than divorce rates in the first place.
Posted by: Tiz at Oct 1, 2007 12:31:40 PM
