Google searches foretold Census report of divorce increase?

But my confidence in this whole enterprise is limited.

Trying to figure out what’s happening with divorce — an extended activity for me lately — might be better left alone for a year or two, till we get some more data and do some serious analysis. But, the bad combination of a trickle of numbers, the ideological ambitions of some people with PhDs, and writers pitching trend stories to their editors, keeps dragging me back in.

But last week we actually got some real new numbers, and some new reporting, so here’s a quick update.

Reporting

The unfortunate new reporting is from this USA Today article, speculating that the recession is giving couples who can’t afford a divorce a chance to reconcile before it’s too late. There are real stories of people who considered divorce but then didn’t, like this couple:

As usual, though, the personal stories are interesting but not worth the damage caused by the causal leaps:

Adding to the confusion is the financial reality that a split is expensive. Census data released last week suggest that the economy has indeed caused a dip in divorce. Some [unnamed] experts predict a divorce explosion when the economy improves, but others [unnamed] say the recession may keep some together long enough to work it out. … The Census bureau counted 65,000 fewer divorces in 2010 than in 2008, a 7% drop.

The total number of divorces between two time points — 2008 and 2010 — is not evidence that “the economy has indeed caused a dip in divorce.” What’s really going on is hard to discover because of at least three things happening at once:

1. Long term decline in divorce rates, but also declines in marriage rates, changes in marriage timing, and changes in pre-marriage cohabitation.
2. Muddled data collection and reporting on divorce, so that some vital statistics reports exclude important states (such as California), while the great new data from the American Community Survey (ACS) is based on self reports that have as-yet unknown biases.
3. The recession which is probably influencing overall trends in all different directions at once.

New numbers

But there are new numbers, so we should at least report them correctly.

The ACS is a giant, national sample survey, which since 2008 has asked people about “marital events” in the previous year. These include marriage, divorce and widowhood. This is definitely not the same as counting up divorce decrees, since it’s an uncorroborated self-report. And, because the rates appear higher than those from other sources, and separation is not a “marital event” they ask about, we suspect some people say they were divorced before the legal deed is done.

Here’s what the ACS has for the past three years. I calculated “refined marriage rates,” or the number of divorces reported per 1,000 married people in the same group in the same period.

The first thing you might notice is that the divorce rates are different for men than for women, which should send up a cautionary flare on the data. Some day, when more homogamous couples have the right to legal divorces, we could have real gender differences in divorce rates, but right now they just mean differences in reporting of unknown origin. But the trends are parallel, so maybe the mismeasurement is at least consistent and the trends are legit.

Anyway, real legal divorces take a while to complete — in fact there are mandated waiting periods in many states. If the recession started in 2008 or so, did it “cause” the dip in divorce from 2008 to 2009? Or, was that just a routine annual decline, counteracted by an unusual upward blip in 2010?

We’re not going to know till we have detailed studies that look at who divorced, when and where (timing, states, unemployment rates, that sort of thing). But, I can’t resist one more trip into the Google-back machine.

Three months ago I reported Google trends “that might support the image I have of divorce pressure building up behind a levee of real estate and unemployment barriers that limit people’s options for breaking up and moving out.”

I used the same method with three more months data, and it still fits. Again, I took the top 100 searches most correlated temporally with “divorce” from Google Correlate (get it here), and broke out the ones that seem clearly divorce related:

• filed for divorce
• file for divorce
• filing for divorce
• file divorce
• after divorce
• attorney search
• dissolution of marriage
• after marriage
• getting divorced

This graph shows the trend for just “divorce” (dots) and the average of all the terms (red line):

If I were the type to gloat prematurely, I would claim credit for predicting the increase in divorce rates Census just reported for 2010. Instead, I put a question mark in the title of the post, and let you draw your own conclusions.