When it comes to divorce rates in the US conventional wisdom is that it peaked around 1980 and has been declining ever since. However, there is new data which suggests that the rate of divorce in the US has remained nearly the same since 1990.
The biggest problem with measuring divorce rates in the US over the last two decades is obtaining a complete data set. Official data isn’t available for all years for all states. Here is a portion of Table 133 from the 2012 Statistical Abstract of the United States. You can get the original in PDF or spreadsheet form here.
As you can see, the number of divorces hasn’t been available from the state of California for over twenty years*. California isn’t the only state with missing data; recent data is also missing for Georgia, Hawaii, Indiana, Louisiana, and Minnesota. Click here to see the full table with the missing data highlighted.
There is another problem with the table from the Statistical Abstract; it is only showing the raw number of divorces as well as the number per 1,000 population. The problem with the first number is obvious; the US population is increasing, so you can’t make a meaningful comparison over time using the raw number of divorces. The problem with the other way they report divorce rates is more subtle. While showing it as a value per 1,000 population adjusts for population growth, it doesn’t take into account another important trend. Only people who are married are at risk for divorce, but the percentage of the population which is married at any given time has declined steadily:
To adjust for this groups like the National Marriage Project divide the raw number of divorces by the number of people who are married in the same period and convert this into divorces per 1,000 married couples (or married women):
This metric makes the most sense when looking at the long term trend, but the data source is problematic. The chart above suffers from the same missing data as the table from the 2012 Statistical Abstract. Out of six 5 year periods of decline since 1980, only the first two include data from California, the largest state in the country by population. Starting with 1995 data from California isn’t included, and as I pointed out above five other states are missing for one or more recent periods as well. Louisiana and Indiana don’t report data going back to at least 1990, perhaps further. This is a very significant gap, and until recently the choices were to use the partial data or not measure national divorce rates at all.
But now there is another option. The American Community Survey (ACS) performed by the US Census has recently added questions about divorce. The ACS uses a nationally representative sample, which solves the problem of the states which aren’t reporting divorces. As I shared here, the National Center for Family & Marriage Research (NCFMR) at Bowling Green State University used ACS data to calculate divorce rates by age in the US for 2010. In their report they make a surprising statement:
The overall U.S. divorce rate has remained essentially unchanged over the past 20 years. In 1990, 19 people divorced for every 1,000 marrieds versus 18 per 1,000 in 2010.
I’m not sure why the NCFMR is calculating the 1990 rate as 19 while the National Marriage Project calculates this as 20.9. While the two are calculating slightly different metrics, I’m not sure this would account for a nearly two point difference. NCFMR is reporting the number of divorces per 1,000 married couples, while the Marriage Project figure is per 1,000 married women.
At any rate, NCFMR is reporting a drop of only 1 divorce annually per 1,000 married couples between 1990 and 2010. This is very different than conventional wisdom, and it will be interesting to see how the researchers in this field sort this out. The NCFMR document which makes this statement is dated 2012 and since it references a working paper from March of 2012 it appears to be very recent. Interestingly, a slightly older document from the same group makes a very different statement about recent divorce rates. Here is what they wrote in First Divorce Rate, 2010, published in 2011 (emphasis mine):
Over the last four decades, there has been more than a three-fold percentage increase in the proportion of Americans who are currently divorced, rising from 2.9% in 1970 to 10.7% in 2008 (FP-10-01). Despite this growing proportion, research suggests that among married couples, the divorce rate—which peaked in the late 70s at about 23 divorced per 1,000—has declined over the past 25 years (Stevenson & Wolfers, 2007).
The full title of the Stevenson and Wolfers paper they cite is Marriage and divorce: Changes and their driving forces, and you can view the paper here (see Figure 1 on P3). As with the data from the Marriage Project, Stevenson and Wolfers show a steady decline in divorces per 1,000 married, although they show what looks like some yearly variation. They don’t give specific numbers in the chart, but their value for 2005 (the final year they show) appears to be the same as the Marriage Project is reporting for the same year (16.4).
I was curious about the different data sets being used by the National Marriage Project and the recent NCFMR report, so I emailed the National Marriage Project. Brad Wilcox, the Director of the National Marriage Project was kind enough to reply quickly to my question. Here is his reply:
Thanks for your note. Because the ACS data provides a more geographically comprehensive portrait of current divorce trends than does the data from the National Center for Health Statistics, the National Marriage Project is considering switching to ACS data in the 2012 edition of the State of Our Unions.
Given how recent the ACS data is I think this is a very fair response. I’m looking forward to seeing what they do in their upcoming report. Their last two reports have been released in December, so I assume that is when the 2012 version will come out as well.
*Table 133 from the 2012 Statistical Abstract doesn’t show divorce numbers between 1990 and 2000, but the final table from the 1995 Statistical Abstract shows California data missing for 1993.
Update: The National Marriage Project is using the ACS data in their latest report.