Demographic Intelligence, low bar edition

U.S. marriage rates are falling generally, so that’s the real news. And it’s important. In fact, one classic projection has the rate hitting zero at 2042. But the news shenanigans are in the prediction business.

I first wrote here about Demographic Intelligence, a profit-making venture founded by Brad Wilcox (full file). They prey on companies’ ignorance about demography and the news media’s desire to stay ahead of the story, making ridiculous claims like “99% accuracy” in their forecasts. Here’s an update.

In a Washington Post Wonkblog entry meaninglessly titled, “Why parents should stop hoping their kids will get married,” we read:

“Millennials are such a big generation, we’re going to have more people of prime marriage age in the next five years than we’ve had at any time in U.S. history. For that alone, we’d expect an uptick in marriage rates,” said Sam Sturgeon, president of Demographic Intelligence.

Setting aside the knowledge-sucking obsession with generational marketing terms, let’s just hold the president of a company with “demographic” in the title to a slightly-higher-than-complete-nonsense standard of demographic intelligence. The median age at marriage is now 28 (combining men and women). At that age there were 4.3 million people in 2013. That is actually the highest number of people ever at the median age at marriage. For example, in 1900, when the median age at marriage was 24, there were only 1.5 million people that age. Wow!

However, intelligent demographer Sturgeon said “for that alone, we’d expect an uptick in marriage rates.” And marriage rates are based on population size. that 4.3 million people at age 28 in 2013 was only 1.35% of the population, while the 1.5 million people at age 24 in 1900 was 1.96% of the population. In fact, the pattern is the opposite of what Sturgeon said: we have never had fewer people — proportionately — in the prime marrying age. Double wow!

Here is the population distribution by age from 1900 to 2013, from the online table maker (try it, it’s easy!). The color coding helpfully shows where the number is above average (red) versus below average (blue). I’ve highlighted the five-year age interval that contains the median marriage age for each decade:

Microsoft PowerPoint - uspop-age-dist-marriage-age.pptx

That the marriage rate is falling — Sturgeon’s expert prediction (see below) is that it will reach an all-time low in 2016 (as it has in 16 of the last 33 years) — is in large part driven by this age composition trend.

How accurate is that forecast?

Demographic Intelligence boasts “99% accuracy” in its wedding forecasts. And these forecasts, they say, are very useful:

This unique forecast is especially valuable as the federal marriage statistics are usually released 12 to 24 months after the date to which they apply, making official data of limited usefulness to the wedding industry. Our forecast is available 24 months before weddings happen, thereby offering a tremendous value to companies that focus on weddings and ancillary businesses.

Now, I’m all in favor of wasting the wedding industry’s money, but I don’t like deceiving the public. So I have to tell you: for every year from 2001 to 2012, if you had simply used last year’s marriage rate to predict this year’s, you would have averaged 98.3% accuracy. That is the deer-in-headlights method of forecasting. In fact, the deer-in-the-headlights forecast for 2012 — that is, assuming no change from 2011 — yields an astonishing accuracy of 99.87% (see below). Not bad! I’ll sell that to you for just 98% of what Demographic Intelligence is charging (except you’re already paying for my services, so you’re welcome).

Of course, demographers like projections, and I’m no exception. It is frustrating that official marriage statistics lag “real time” so much more than other important statistics, such as the unemployment rate or the number of named storms per season. That’s why in 2013 I announced a marriage forecast contest to predict the 2012 marriage rate, and provided some trends in key variables for you to experiment with (in a spreadsheet here): Google searches for wedding invitations, bridal showers, and wedding gifts; the unemployment rate, the Index of Consumer Sentiment, and the number of women ages 20-39:

There was so little interest in my contest (go figure), that I never got around to updating the results. So here goes. We now know from official statistics that there were 2,131,000 marriages in 2012, which, for a population of 313,914,040, yields a marriage-per-1000 rate of 6.788, down from 6.797 in 2011. Using different combinations of these variables, I generated projections using linear regressions. As I noted, the no-change performed very well, at 99.87% accuracy. But the winning model was actually the one that used the Google search trends only, which predicted 2,133,647 weddings, an astonishing 99.88% accurate. If Google is not using their data to get filthy rich — oh wait.

Anyway, in this exercise I’m just predicting the next year in the series — it gets a little trickier if you want to go four years out. And demographic projections are a serious science. But this prediction business is just wasting money and confusing people.

4 thoughts on “Demographic Intelligence, low bar edition

  1. Phil,
    Thanks for the write up. I had a very cordial conversation with the reporter for this article (Brigid Shulte is a very polite). I am familiar with the how rates are applied to populations. We talked for some time about how even though we are forecasting the marriage rate to go down in the near term, we expect total marriages to go up simply because of the size of the age cohorts involved. If you had seen the whole report, you would see that we never do predict an increase in the marriage rate. I believe I was accurately quoted in the article–though I would have preferred that in both instances total marriages would have been used instead of the term rates. Admittedly, this could have been my error. What is missing from the quotes above is a little context. Assuming my memory is correct, at this time in the interview we are talking about how one might expect the marriage rate to go up given the relative size of millennial cohorts. However, as I mentioned, we do not predict an increase in the marriage rate.

    You must admit–just as a thought experiment–if current age-specific marriage rates were held constant, we could see a slight uptick in the actual marriage rate (albeit somewhat small) as the largest group of millennials (currently age 24, I believe) reaches the prime marriage age of 28 that you highlight. However, we do not expect this to happen, and the anticipated further delay in marriage would suggest that age-specific marriage rates will not remain constant.

    You are correct to note that in absolute terms, we currently have more people in the prime age for marriage than at any time previously. However, as you also note, as a proportion of the population, those of prime marrying age are relatively small, at least by historical standards. I also spoke with the author about about how part of what is driving the marriage rate down is the age structure of the population. As you point out, the crude marriage rate is a very rough measure of marriages, especially in an aging population where aging cohorts add significantly to the denominator of the calculation without contributing as much to the numerator. You are correct in noting that there is much more complexity to this than the two quotes that are provided.

    Overall, I thought this was a fair summary of some of the information included in our report, though admittedly the use of the term “rates” was unfortunate (and again, most likely my error). Thanks for holding my feet to the fire.

    Sam Sturgeon


  2. Perhaps we should interpret trends in marriage rates together with trends in cohabitation. So as not to include those who are just shacking up, one might condition on a one-year minimum cohabitation duration.


Comments welcome (may be moderated)

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s