In Post-marital spinsterhood part two: the data, I showed how stats from a NCHS/CDC study in 2002 using data of women who divorced in the mid 80s was causing women to have an unrealistic expectation of their chances of remarriage following divorce today. The other day I was looking for more (and hopefully newer and better) data on the question, and found another paper using the same 1995 NSFG data set: First Marriage Dissolution, Divorce, and Remarriage: United States. Advance Data 323.
This paper has the same results but shares more detail on the data used, methodology, and what they actually found. For example, this table on page 9 shows the values for the data points included in the charts I referenced in the other post:
If you click on the chart to see the full size version, you will notice the final row includes the percent of divorced women who’s data was censored in the study because they had not yet remarried at the time of the interview. Also note the age range for the data set. They only interviewed women aged 15 to 44 for this study.
The limited age range really made this a poor data set to try to determine remarriage rates. Of the women they interviewed who were over 25 at the time of divorce, 48% had not remarried! Yet this is the very data set referenced when we hear that a woman has a 75% chance of remarriage within 10 years of divorce.
The researchers used a mechanism called Life tables to try to make the best use of their limited data (page 3). This allows them to use data points which have partial value instead of throwing them out entirely:
So a woman who divorced 5 years ago and never remarried can be used to determine the remarriage probability for the first five years. However, her data can’t be used to calculate remarriage probability for years 6 and beyond because her future status can’t be known. A woman who divorced 5 years ago and remarried after 3 years would be calculated as unmarried for the first 3 years and calculated as married for the remaining 15 years since they know remarriage is an event which can’t be “undone” (divorce doesn’t change the fact that she remarried).
At first glance this sounds much better than throwing out all of the surveys for women who hadn’t remarried. Not only would they be shrinking their sample size, but by studying only those women who had remarried they would bias their results towards remarriage. However, it seems the Life table approach would also create a bias towards remarriage, especially for longer time durations. For example, both women in the examples above had divorced 5 years ago. However, only the one who remarried will be included in the data used to calculate 10 and 15 year remarriage probabilities. Given that nearly half of the data they used had this problem, this would seem to be a very significant bias which isn’t accounted for in the methodological explanation of the paper.
This one bias alone seems quite significant to me, but in addition I noticed the following other problems:
- Instead of reporting on likelihood of remarriage following divorces which occurred in 1985 as I originally thought, the study actually is going farther back and examining remarriage rates for divorces which occurred as far back as 1965. As I showed in the first post, remarriage rates are declining over time.
- The results are being reported as for women 15-44, but because of how they structured their sample they ensured that they wouldn’t have any data on 10 year or longer remarriage rates for women who divorced over age 35. The five year data only includes women who divorced at age 40 or younger and is skewed strongly towards younger because only the oldest of those surveyed could possibly have divorced at this older age. Even worse, the data is being represented as showing remarriage rates for women under 25 and those 25 and over. The over 25 results most closely represent remarriage rates of women who divorce between 25 and 35. Even worse, the statistic most often cited (the stat for all women) is skewed even younger, primarily representing women who divorced between 15 and 35.
All three of the biases I found would cause an overestimation of remarriage rates. If you add all three effects together, this study must be massively overestimating remarriage rates, and this is before taking into account the evidence that the remarriage rate itself has continued to fall in the 15 years since the data was gathered (see my original post).
Why confusion about age is such a disaster when women see these stats:
As I have shown the study is strongly biased toward divorces which occur earlier in life when a woman’s chance of remarriage is highest. While the population of divorced women is almost entirely over age 35, the study focuses almost entirely on women 35 and under. The age skew of the divorcée population occurs partly because you have to marry first to divorce (and marriage is occurring later in life), and partly because of the greater likelihood of remarrying quickly if still young. According to the 2010 Census Data, there are 13.8 million divorced (and not remarried) women in the US. Over 80% of divorced women are over age 40, and 89% of them are over age 35. We know from the AARP study and others that past age 40 divorcées are much less likely to remarry, and much more likely to be terribly alone if they don’t. As I have shared before:
Almost 9 in 10 men (87%) dated after their divorce, compared to 8 in 10 women (79%)… Among those who dated after the divorce, more than half of men (54%) but fewer women remarried (39%). (Page 39)
Many women, especially those who have not remarried (69%), do not touch or hug at all sexually. An even larger majority of women who have not remarried do not engage in sexual intercourse (77% saying not at all), in comparison with about half of men (49%) who have not remarried. (Page 6)
Keep in mind that the AARP study itself is reporting outcomes for women who divorced roughly 20 years ago, so the results for current divorcées are likely to be even worse based on the declining remarriage trend.
I know this is all very grim, and if you are like me you are pretty well exhausted with all of the data. But knowing the reality of the data is important, and I’m not aware of anyone else doing this work. Additionally no new studies have been performed for over 15 years, and I’m not aware of any on the horizon. Ignoring reality and actively selling late life divorce has gotten us over 11 million divorcées over age 40 in the US alone. Telling the painful truth can only help women not yet divorced make the best choice possible for their specific situation. On the whole the direction this moves their choices as a group will also be more moral (keeping their most solemn promises), in addition to being better for their children and hopefully themselves and their husbands.
Note: It is likely that one of my readers has much more experience with this sort of analysis than I do. From my own analysis I’m very confident that I am right. However, feel free to correct me if I have misunderstood or misrepresented the data or methodology of the report. I don’t mean Dalrock is a meanie misogynist so it must be wrong. But please share an actual logical explanation of where I have misinterpreted the report if you believe I have made an error.
Dec 21 2012 Update: We now have a fresh snapshot of remarriage rates using what appears to be a more solid data set.