High marks for Census

What a difference a bureaucracy makes.

The Census Bureau has for the first time released a count of homogamous unmarried and married couples from the Decennial Census. This was occasioned by an Obama administration decision to reinterpret the Defense of Marriage Act – which had previously been used to prevent such a count. Although the annual American Community Survey had already started collecting figures on married homogamous couples, this is the first time for the symbolically and demographically important Decennial Census.

But because Obama only got elected in 2008, it was too late to make a crucial correction to (some of) the Decennial Census forms before they went out in 2010. As a result, millions of people completed a form which had a layout that just vaguely encouraged an error in checking the “sex” boxes. And if one out of every 1,000 of the 130 million heterogamous married couples makes that mistake, the number of homogamous married couples — about 130,000 — doubles. And that’s what happened, more or less. Hence the corrected numbers released today.

Culture shift

The culture shift at Census is not just in counting gay and lesbian married couples. They have also done this with a high degree of transparency (even allowing me, as an unpaid consultant, to publish with the experimental data they collected).

Here is the video they released today, explaining the situation. I like a few things about it:

  • It’s truly scientifically informative, and even admits that, “as scientists, demographers, and statisticians, we are obligated…”
  • They explain that there were errors, and that they tried to correct them. (They don’t get into the details — but that’s in a thorough technical paper.)
  • They show actual man-man and woman-woman couples filling out their forms, including a couple with children.
  • I sort of like the little man and woman icons they use in the chart (a recurring issue). The bodies are the same, but the clothes are different. I can get behind that, what with the social construction of gender difference. On the other hand, it’s too bad the man is naked and the woman is wearing a skirt, because of the man=normal / woman=different connotation.

All in all, this is good progress, in my opinion. I credit to Census Bureau Director Robert Groves, who coincidentally is a real sociologist, as well as Martin O’Connell and the others who do the work on family stuff at the Bureau. In the press release, Groves is quoted as saying,

We understand how important it is for all groups to have accurate statistics that reflect who we are as a nation. As scientists, we noticed the inconsistency and developed the revised estimates to provide a more accurate portrait of the number of same-sex couples.  We’re providing all three – the revised, original and ACS estimates – together to provide users with the full, transparent picture of our current measurement of same-sex couples.

What it’s all worth, in work-life cash

A Census Bureau research report estimates lifetime earnings by education, race/ethnicity and gender.

The report, by the Bureau’s Tiffany Julian and Robert Kominski, uses national data from the American Community Survey to create “synthetic work-life estimates” of earnings.

The method takes earnings information from one time period — in this case the years 2006-2008 combined (before the recession) — and calculates how much money people would make if they lived through their whole work lives (40 years, from age 25 to 64) during that period. Demographers use the same method to estimate life expectancy. It’s a way of using the most current period to project an image of the future in today’s shape. It’s a better look at the future, for most purposes, than looking back at the lives of people who are wrapping it up today.

Here is a figure they made, using earnings from people working full-time and year round:

That is for people working full-time and year-round at their jobs. That is not reasonable, of course, if people take time out of the labor force, or out of full-time work. So this understates the earnings gaps, especially by gender, since women take more time out of the labor force than men, on average.

They also reported the projected lifetime earnings for all workers — including those working only part-time or part of the year. The figure above showed a ratio of 4.7-to-1 from top to bottom, whereas the all-worker data has a ratio of 5.6-to-1 from White male professional-degree holders to Latina high school graduates.

I turned their all-worker table into this graph with men and women color coded:

This is not a real prediction, just a projection of the present into the future. But the scale is good for the imagination — the gap from top to bottom is 3.65 million dollars in 2008 terms.

Note that in addition to employer discrimination, these gaps reflects the full range of influences on people’s earnings, including sorting into occupations, part-time work, lost tenure and experience from time out of the labor force, and regional variation (which is one reason Asian workers show up high – many live in expensive cities like San Francisco and Honolulu).

Birth of a headline

One day this summer I got an email message from a reporter. We had an exchange over email. I was on vacation. Ten days later a story appeared. Here’s my end of the back-story. Pleasantries and personal details omitted.


I’m a reporter with the News & Observer, and I’m hoping you’ll be able to speak with me for a few minutes today for a story I’m working on. My story is about the latest census data from NC, specifically the male/female ratio throughout age, along with the ways it has changed over time.  I have some specific questions about why the ratio flip-flops in middle age, and why it used to — but no longer does — flip-flop again in late middle age/early old age.


I’d be happy to try to help you. I’m not familiar with the pattern you’re seeing, however. Can you send me or refer me to the numbers you have?


I’m … interested in why boys outnumber girls, but then later in life women outnumber men… [info on locating a spreadsheet file with the data, then a list of “driving questions”]:

  • Boys outnumber girls – why?
  • Women outnumber men – why?
  • Why have both trends flattened out in the last few decades?
  • What drives variation over time and place?

I looked over the numbers, then wrote back.

This looks quite normal to me. There are about 104 boys born for every 100 girls, naturally. That’s apparently evolution’s answer to the fact that, as the weaker sex, males die more often at all ages. Some of that is from social causes — like violence, war, and drunk driving — but it is also natural biology for humans. So you always have more boys born, and then at some point in the age distribution it crosses over and you have more females living.

So it’s an interesting question what makes that change over time, or vary from place to place. Change over time has to do with mortality. For example, men smoke(d) much more than women, so that’s contributes a few years to the age difference in life expectancy, although that is decreasing now. In North Carolina, I would guess that what is going on has to do with our large military presence, and our immigration flows (though I don’t know which way those skew here).

The general movement of the crossover age from younger ages to older ages reflects declining overall mortality, so everyone lives longer. I don’t know why it would have gone back down from 35 to 25 in the last decade. That’s an interesting question — I wonder if the national data show the same thing? I’m sorry I don’t know off hand.

That was off the top of my head. Hoping to have more time to look into it, I tried to set up a time to call, but she was out of time on the story. 10 days later, on page 1 of the News & Observer, was this:

And the quote from me:

Women giving birth to more boys than girls is “apparently evolution’s answer to the fact that, as the weaker sex, males die more often at all ages,” said Philip Cohen, a professor of sociology at UNC-Chapel Hill.

From, “This looks quite normal to me,” to “What’s happening to all of North Carolina’s men?” That’s something.

Anyway, it doesn’t justify the headline, but there is an interesting pattern here. North Carolina is not unique, but it’s among the states with fewer men from about age 25 to 60. The differences aren’t great, but there has to be a reason, and it’s worth looking at.

Here is the pattern, with North Carolina, California, and New York calculated from the 2010 Census (which isn’t available for every state yet), and the national average from the 2009 American Community Survey:

The News & Observer had a graph, too, but it didn’t show how North Carolina differed from other states:

Nice touch, though: “After age 24, the ratio swings in favor of the ladies.” I know it’s sometimes hard to make demography an attractive subject, but still.