Prediction is cheap.
After about five years of increasing fertility rates, by this time last year I was reviewing evidence that the recession had turned that around, leading to a steep decline in births. This update shows that trend was very real – but a review of future events suggests fertility is ready to rebound.
By this spring we had more evidence that supported a role for the recession specifically in this turnaround. Based on the timing, on the steeper drop among couples that already have some children, and the pattern across states, I joined in the Pew Research Center’s conclusion that the decline was linked to the recession.
Where do we go from here?
But this data story is slow to evolve. Even now, August 2011, the CDC has released only provisional numbers from June 2010. A year ago I used some Google search terms to try to read closer to the present. Now that we have Google Correlate I think I can do a better job.
Unfortunately, the time series for fertility rates is monthly, while Google Correlate likes weekly numbers. So I decided to use the state variation to identify fertility-related searches, and then feed that back into the time trends using Google Trends. I took the state fertility rates from 2009 and asked Google for the 100 search terms most closely correlated with those rates across states. The result had a convincing level of face validity — that is, lots of the searches were about pregnancy and births. The #1 correlated term was “pregnancy workout,” with a correlation of .88 out of 1.0 (and #2 was “baby diarrhea”). In fact, 12 of the top 100 terms included references to pregnancy.
To develop the time series, I had to use those that were common enough to appear in Google Trends consistently back to 2007, so I settled on pregnancy growth, pregnancy tips, and pregnancy contractions (see the results here). Here’s the pattern for state fertility rates (left) and searches for pregnancy growth:
Averaging those three search term trends, back to 2007, gives a trend like this:
It’s not definite, but that could be a bottom in the second half of 2010 — after the end of the currently-available fertility rates from CDC. Does it line up with the fertility trend for the period when the two series overlap? It’s not bad. Here are the searchers (blue line) and the fertility trend (red line), which are correlated with each other at .32:
So, in conclusion, I predict we will see a fertility rebound at least in the first half of 2011. I don’t know why else we would be seeing an increase in people doing these searches (though if the Google numbers are not reliable, all bets are off).
Unbelievably, this post also involves Brad Wilcox. I had no intention of even mentioning him today, but I feel ethically obligated to report that – after doing this analysis – I Googled “fertility predictions demography” just to make sure someone else didn’t already do this (remember, kids, after doing your research it’s always good to check and see if it was necessary in the first place). What I came up with was something called “Demographic Intelligence, LLC,” which has as its web home a WordPress page, the “About” page of which says (I’m not making this up):
Professor [W. Bradford] Wilcox founded Demographic Intelligence to provide companies, governments, and NGOs with the ability to anticipate shifts in U.S. fertility, to learn what types of American adults are most likely to be having children, and to understand the fundamental economic, social, and cultural drivers of fertility in the U.S.
This limited liability company, which according to its LinkedIn profile has an employee, has issued at least one press release claiming “about 98 percent” accuracy with their prediction model. And according to this model, which carries a TM after its name, the fertility drop has bottomed out (which is, obviously, a “good sign for the juvenile products industry”).
You have my word that I made my prediction before reading this press release. But different minds apparently think alike, at least when it comes to seeing the bottom of the fertility trough.