The coming divorce decline

Unless something changes outside the demogosphere, the divorce rate is going to go down in the coming years.

Divorce represents a number of problems from a social science perspective.

• • Most people seem to assume “the divorce rate” is always going up, compared with the good old days, which are supposed to be the whole past but are actually represented by the anomalous 1950s.
  • On other hand, social scientists have known for a few decades that “the divorce rate” has actually been declining since the 1980s. That shows up in the official statistics, with the simple calculation — known as the refined divorce rate — of the number of divorces per 1,000 married women.
  • On the third hand, the official statistics are very flawed. The federal system, which relies on states voluntarily coughing up their divorce records, broke down in the 1990s and no one fixed it (hello, California doesn’t participate). In the debate over different ways of getting good answers, a key 2014 paper from Sheela Kennedy and Stephen Ruggles showed that the decline in divorce after 1980 was mostly because the whole married population was getting older, and older people get divorced less. That refined divorce rate doesn’t account for age patterns. When you remove the age patterns from the data, you see a continuously increasing divorce rate. Yikes!
  • On the fourth hand, Kennedy and Ruggles stopped in about 2010. Since then, the very divorce-prone, multi-marrying, multi-divorcing Baby Boomers have moved further out of their peak action years, and it’s increasingly clear that divorce rates really are falling for younger people.

In my new analysis, which I wrote up as a short paper for submission to the Population Association of America 2019 meetings, I argue that all signs point to a divorce decline in the coming years. Here is the paper on SocArXiv, where you will also find the data and code. And here is the story, in figures (click to enlarge).

1. The proportion of married women who divorce each year has fallen 18% in the decade after 2008. (There are reasons to do this for women — some neutral, some good, some bad — but one good thing nowadays is at least this includes women divorcing women.) And when you control for age, number of times married, years married, education, race/ethnicity, and nativity, it has still fallen 8%.

2. The pattern of increasing divorce at older ages, described by Susan Brown and I-Fen Lin as gray divorce, is no longer apparent. In the decade after 2008, the only apparent change in age effects is the decline at younger ages, holding other variables constant.

3. The longer term trends, identified by Kennedy and Ruggles, which I extend to 2016, show that the upward trajectory is all about older people. These are prevalences (divorced people in the population), not divorce rates, but they are good for illustrating this trend.

4. In fact, when you look just at the last decade, all of the decline in age-specific divorce rates is among people under age 45. This implies there will be more older people who have been married a long time, which means low divorce rates. Also, their kids won’t be as likely to have divorced parents, although more kids will have parents who aren’t married, which might work in the other direction. (You can ignore then under-20s, who are 0.2% of the total.)

5. Finally, to get a glimpse of the future, I looked at women who report getting married in the year before the survey, and how they have changed between 2008 and 2016 on traits associated with the risk of divorce. They clearly show a lower divorce-risk profile. They are more likely to be in their first marriage, to have college degrees, to be older, and to have no children in their households (race/ethnicity appears to be a wash, with fewer Whites but more Latinas).

6. Finally finally, I also looked at the spouses of the newly-married women, and made an arbitrary divorce-protection scale, with one point to each couple for each spouse who was: age 30 or more, White or Hispanic, BA or higher education, first marriage, and no own children. Since 2008 the high scale scores have become more common and the low scores have become rarer.

7. It’s interesting that the decline in divorce goes against the (non-expert) conventional wisdom. And it is happening at a time when public acceptance of divorce has reached record levels (which might be part of why people think it’s growing more common — less stigma). Here are the trends in attitudes from Pew and Gallup:

That’s my story — thanks for listening!

Who are you gonna marry? That one big assumption marriage promotion gets totally wrong

First preamble, then new analysis.

One critique of the marriage promotion movement is that it ignores the problem of available spouses, especially for Black women. Joanna Pepin and I addressed this with an analysis of marriage markets in this paper. White women ages 20-45, who are more than twice as likely to marry as Black women, live in metro areas with an average of 118 unmarried White men per 100 unmarried White women. Black women, on the other hand, face markets with only 78 single men per 100 single women. This is one reason for the difference in marriage rates; given very low rates of intermarriage, especially for Black women, some women essentially can’t marry.

But surely some people are still passing up potential marriages, or so the marriage promoters would have us believe, and in so doing they undermine their own futures and those of their children. Even if you can get past the sex ratio problem, you still have the issue of the benefits of marriage. Of course married people, and their kids, are better off on average. (There are great methodological lessons to be learned from their big lie use of this fact.) But who gets those benefits? The intellectual water-carriers of the movement, principally Brad Wilcox and his co-authors, always describe the benefits of increasing marriage as if the next marriage to occur will provide the same benefits as the average existing marriage. I wrote about how this wrong in Enduring Bonds:

The idea that the “benefits” of marriage—that is, the observed association between marriage and nonpoverty—would accrue to single mothers if they “simply” married their current partners is bonkers. The notion of a “marriage market” is not perfect, but there is something like a marriage queue that arranges people from most likely to least likely to marry. When you say, “Married people are better off than single people,” a big part of what you’re observing is that, on average, the richer, healthier, better-at-relationships people are at the front of that queue, more likely to marry and then to display what look like the benefits of marriage. Those at the back of the queue, who are more (if not totally) “unmarriageable,” clearly aren’t going to have those highly beneficial marriages if they “simply” marry the closest person.

In fact, I assume this problem has gotten worse as marriage has become more selective, as “it’s increasingly the most well off who are getting and staying married,” and those who aren’t marrying “may not have the assets that lead to marriage benefits: skills, wealth, social networks, and so on.”

Note on race

People who promote marriage don’t like to talk about race, but if it weren’t for race — and racism — they would never have gotten as far as they have in selling their agenda. They use supposedly race-neutral language to talk about fatherhood and a “culture of marriage” and “sustainably escaping poverty,” in ways that are all highly relevant to Black families and racial disparities. If you think the problem of marriage is that poor people are not marrying enough, you should not avoid the fact that you’re talking about race. Black women, especially mothers, are much less likely to be married than most other groups of women, even at the same level of income or education (last I checked Black college graduates were 5-times more likely than White college graduates to be single when they had a baby). So, don’t avoid that this is about race, own it  — the demographic facts and political machinations in this area are all highly interwoven with race. I do this analysis, like the paper Joanna and I did, separately for Black and White women, because that’s the main faultline in this area. The code I share below is adaptable to use with other groups as well.

Data illustration

In this data exercise I try to operationalize something like that marriage market queue, to show that women who are least likely to marry are also least likely to enter an economically beneficial marriage if they did marry. See how you like this, and let me know what you think. Or take the data and code and come up with a different way of doing it.

The logic is to take a sample of never-married women, and women who just got married in the last year, and predict membership in the latter group. This generates a predicted probability of marrying for each woman, and it means I can look at the never-married women and see which among them are more or less likely to marry in a given year. For example, based on the models below, I would estimate that a Black woman under age 25, with less than a BA degree, who had a job with less-than-average earnings, has a 0.4% probability of marrying in one year. On the other hand, if she were age 25+, with a BA degree and above-average earnings, her chance of marrying rises to 3.5% per year. (Round numbers.)*

Next, I look at the husbands of women who married men in the year prior to the survey, and I assign them economic scores on an 11-point scale (this is totally arbitrary): up to four points for education, up to four points for earnings, and up to three points for employment level (weeks and hours worked in the previous year). So, a woman whose husband has a high school education, earned $30,000 last year, and worked full-time, year-round, would have 7 points.

Finally, I show the relationship between the odds of marriage for women who didn’t get married and the economic score of the men they would have married if they did.

There are two descriptive conclusions, which I assumed I would find: (1) women who get married marry men with better economic scores than the women who don’t get married would if they did get married; and, (2) the greater the odds of marriage, the better the economic prospects of the man they would marry. The substantive conclusion from this is that marriage promotion, if it could get more people to marry, would pull from the women on the lower rungs of marriage probability, so those new marriages would be less economically beneficial than the average marriage, and the use of married people’s characteristics to project the benefits of marriage for unmarried people is wrong. Like I said, I already believed this, so this is a way of confirming it or showing the extent to which it fits my expectations. (Or, I could be wrong.)

Here are the details.

I use the 2012-2016 five-year American Community Survey data from IPUMS.org (for larger sample). The sample is women ages 18-44, not living in group quarters, single-race Black or White, non-Hispanic, and US-born. I further limited the sample to those who never married, and those who are married for the first time in the previous 12 months. That condition — just married — is the dependent variable in a model predicting odds of first marriage. (Women with female spouses or partners are excluded, too.) The variables used to predict marriage are age (and its square), education, earnings in the previous year (logged), and having no earnings in the previous year (these women are most likely to marry), disability status, metro area residence, and state dummy variables. It’s a simple model, not trying for statistical efficiency but rather the best prediction of marriage odds. Then I use the same set of variables, limiting the analysis to just-married women, to predict their husbands’ economic scores. The regression models are in a table at the end.**

Figure 1 shows how the prediction models assign marriage probabilities. White women have much higher odds of marrying, and those who married have higher odds than those who didn’t, which is reassuring. In particular, a large proportion of never-married Black women are predicted to have very low odds of marrying (click to enlarge).

Figure 2 shows the distribution of husbands’ economic scores for Black and White women who married and those who didn’t. The women who didn’t marry have lower predicted husband scores, with the model giving them husbands with a mode of about 7.0 for Whites and 6.5 for Blacks (click to enlarge).

Finally, the last figure includes only never-married women. It shows the relationship between predicted marriage probability and predicted husband score, using median splines. So, for example, the average unmarried Black woman has a marriage probability of about 1.7%. Figure 3 shows that her predicted husband would have a median score of about 6.4. So he could be a full-time, full-year worker with a high school education, earning $19,000 per year, which would not be enough to lift her and one child out of poverty. The average never-married White woman has a predicted marriage probability of 5.1%, and her imaginary husband has a score of about 7.4 (e.g., a similar husband, but earning $25,000 per year).

Figure 3 implies  what I thought was obvious at the beginning: the further down the marriage market queue you go, the worse the economic prospects of the men they would marry, if there were men for them to marry (whom they wanted to marry, and who wanted to marry them).

I will now be holding my breath while marriage promotion activists develop a more sensible set of assumptions for their assessment of the benefits of the promoted marriages they assure us they will be able to conjure if only we give them a few billion more dollars.

I’m posting the data and code used on the Open Science Framework, here. Please feel free to work with it and let me know what you come up with!

---

* This looks pretty similar to what Dohoon Lee did in this paper, including his figures, and since I was on his dissertation committee, and read his paper, which has similar figures, I credit him with this idea — I should have remembered earlier.

** Here are the regression models used to (1) predict marriage, and then (2) predict husband’s economic scores.

 

Theology majors marry each other a lot, but business majors don’t (and other tales of BAs and marriage)

The American Community Survey collects data on the college majors of people who’ve graduated college. This excellent data has lots of untapped potential for family research, because it tells us something about people’s character and experience that we don’t have from any other variables in this massive annual dataset. (It even asks about a second major, but I’m not getting into that.)

To illustrate this, I did two data exercises that combine college major with marital events, in this case marriage. Looking at people who just married in the previous year, and college major, I ask: Which majors are most and least likely to marry each other, and which majors are most likely to marry people who aren’t college graduates?

I combined eight years of the ACS (2009-2016), which gave me a sample of 27,806 college graduates who got married in the year before they were surveyed (to someone of the other sex). Then I cross-tabbed the major of wife and major of husband, and produced a table of frequencies. To see how majors marry each other, I calculated a ratio of observed to expected frequencies in each cell on the table.

Example: With weights (rounding here), there were a total of 2,737,000 BA-BA marriages. I got 168,00 business majors marrying each other, out of 614,000 male and 462,000 female business majors marrying altogether. So I figured the expected number of business-business pairs was the proportion of all marrying men that were business majors (.22) times the number of women that were business majors (461,904), for an expected number of 103,677 pairs. Because there were 168,163 business-business pairs, the ratio is 1.6.  (When I got the same answer flipping the genders, I figured it was probably right, but if you’ve got a different or better way of doing it, I wouldn’t be surprised!)

It turns out business majors, which are the most numerous of all majors (sigh), have the lowest tendency to marry each other of any major pair. The most homophilous major is theology, where the ratio is a whopping 31. (You have to watch out for the very small cells though; I didn’t calculate confidence intervals.) You can compare them with the rest of the pairs along the diagonal in this heat map (generated with conditional formatting in Excel):

Of course, not all people with college degrees marry others with college degrees. In the old days it was more common for a man with higher education to marry a woman without than the reverse. Now that more women have BAs, I find in this sample that 35% of the women with BAs married men without BAs, compared to just 22% of BA-wielding men who married “down.” But the rates of down-marriage vary a lot depending on what kind of BA people have. So I made the next figure, which shows the proportion of male and female BAs, by major, marrying people without BAs (with markers scaled to the size of each major). At the extreme, almost 60% of the female criminal justice majors who married ended up with a man without a BA (quite a bit higher than the proportion of male crim majors who did the same). On the other hand, engineering had the lowest overall rate of down-marriage. Is that a good thing about engineering? Something people should look at!

We could do a lot with this, right? If you’re interested in this data, and the code I used, I put up data and Stata code zips for each of these analyses (including the spreadsheet): BA matching, BA’s down-marrying. Free to use!

Unequal marriage markets for Black and White women

Joanna Pepin and I have posted a new paper titled, “Unequal marriage markets: Sex ratios and first marriage among Black and White women.” In the paper, we find that the marriage markets of Black and White women are very different, with Black women living in metropolitan areas that have many fewer single men than White women do. And, in a regression model with other important predictors of marriage, this unmarried sex ratio is strongly associated with the odds of marrying.

We count this as evidence on the side of “structure” over “culture” in the debates over the decline in marriage. Here’s the main result, showing Black and White women in 172 metro areas (scaled for size), and the difference in sex ratios (the horizontal spread), the difference in marriage rates (the vertical spread), and the statistical effect of sex ratios on marriage (the slopes).

In a nutshell: As you move from left to right, there are more men, and higher odds of marriage. And almost all the White women are up and to the right compared with the Black women. One implication is that this could be one reason why marriage promotion programs in the welfare system aren’t working.

There are a couple of noteworthy innovations here. First, we used the American Community Survey marital events data, which is marriage happening (did you get married in the last year?) rather than just existing (are you married?). This is a better way to assess what might influence marriage. Second, young people, especially single young people who might be getting married, move around a lot. So what is their marriage market? It’s impossible to say exactly, but we define it as the metro area where they lived one year earlier, rather than just where they live now. (This is especially important because the people who move may move because they just got married.)

The paper is on SocArXiv, where if you follow the links you get to the project page, where we put most of the data and code. The paper is under review now, and we’d love to know if you find any mistakes or have any suggestions.

(This began with a blog post four years ago in which I critiqued a NYT Magazine piece by Anne Lowrie about using marriage to cure poverty. Then we presented a first pass at the American Sociological Association in 2014, and I put some of the descriptive statistics in my textbook, and we made a short video out of it, in which I said, “So, larger social forces — the economy, job discrimination, incarceration policies, and health disparities — all impinge on the ability of individuals to shape their own family lives.” Along the way, I presented some about it here and there, while thinking of new ways to measure marriage inequalities.)

Why Heritage is wrong on the new Census race/ethnicity question

Sorry this is long and rambly. I just want to get the main points down and I’m in the middle of other things. I hope it helps.

Mike Gonzalez, a Bush-era speech writer with no background in demography (not that there’s anything wrong with that), now a PR person for the Heritage Foundation, has written a noxious and divisive op-ed in the Washington Post that spreads some completely wrong information about the U.S. Census Bureau’s attempts to improve data collection on race and ethnicity. It’s also a scary warning of what the far right politicization of the Census Bureau might mean for social science and democracy.

Gonzalez is upset that “the Obama administration is rushing to institute changes in racial classifications,” which include two major changes: combining the Hispanic/Latino Origin question with the Race question, and adding a new category, Middle Eastern or North African (MENA). Gonzalez (who, it must be noted, perhaps with some sympathy, recently wrote one of those useless books about how the Republican party can reach Hispanics, made instantly obsolete by Trump), says that what Obama has in mind “will only aggravate the volatile social frictions that created today’s poisonous political climate in the first place.” Yes, the “poisonous political climate” he is upset about (did I mention he works for the Heritage Foundation?) is the result of the way the government divides people by race and ethnicity. Not actually dividing them, of course (which is a real problem), but dividing them on Census forms. (I hadn’t heard this particular version of why Trump is Obama’s fault — who knew?)

How will the new reforms make the Trump situation he helped create worse? Basically, by measuring race and ethnicity, which Gonzalez would rather not do (as suggested by the title, “Think of America as one people? The census begs to differ,” which could have been written at any time in the past two centuries).

Specifically, Gonzalez claims, completely factually inaccurately, that Census would “eliminate a second question that lets [Hispanics] also choose their race.” By combining Hispanic origin and race into one question — on which, as before, people will be free to mark as many responses as they like — Gonzalez thinks Census would “effectively make ‘Hispanic’ their sole racial identifier.” He is upset that many Latinos will not identify themselves as “White” if they have the option of “Hispanic” on the same question, even if they are free to mark both (which he doesn’t mention). Some will, but that is not because anyone is taking away any of their choices.

The Census Bureau, of course, because they always do, because they are excellent, has done years of research on these questions, including all the major stakeholders in a long interactive process that is scrupulously documented and (for a government bureaucracy) quite transparent. Naturally not everyone is happy, but in the end the trained demographic professionals come down on the side of the best science.

Race that Latino

The most recent report on the research I found was a presentation by Nicholas Jones and Michael Bentley from the Census Bureau. This is my source for the research on the new question.

First, why combine Hispanic with race? You have probably seen the phrase “Hispanics may be of any race” on lots of reports that use Census or other government data. The figure below is from the first edition of my book, using 2010 data, in which I group all 50 million Hispanics, and show the races they chose: about half White, the rest other race or more than one race (usually White and other race). Notice that by this convention Hispanics are removed from the White group anyway, just because we don’t want to have people in the same picture twice (“non-Hispanic Whites” is already a common construction).

The “may be of any race” language is the awkward outcome of an approach that treats Hispanic as an “ethnicity” (actually a bunch of national origins, maybe a panethnicity), while White, Black, Asian, Pacific Islander, and American Indian are treated as “races.” The distinction never really made sense. These things have been measured using self-identification for more than half a century, so we’re not talking about genetics and blood tests, we’re talking about how people identify themselves. And there just isn’t a major categorical difference between race and ethnicity for most people — people of any race or ethnicity may identify with a specific national origin (Italian, Pakistani, Mexican), as well as a “race” or panethnic identify such as Asian, or Latino. And now that the government allows people to select multiple races (since 2000), as well as answering the Hispanic question, there really is no good justification for keeping them separate. As you can see from my figure above, when we analyze the data we mostly pull all the Hispanics together regardless of their races. The new approach just encourages them to decide how they want that done, which is usually a better approach.

Of course, Asians and Pacific Islanders have been answering the “race” question with national origin prompts for several decades. There was no “Asian” checkbox in 2000 or 2010 (or on the American Community Survey). So they have been using their ethnicity to answer the race question all along — that’s because for some reason you just can’t get “Asian” immigrants, especially recent immigrants — that is, people from India, Korea, and Japan, Vietnam, and so on — to see themselves as part of one panethnic group. Go figure, must be the centuries of considering themselves separate peoples, even “races.” So, a new question that combines the more ethnic categories (Mexican, Pakistanis, etc.), with America’s racial identities (Black, White, etc.), just works better, as long as you let people check as many boxes as they want. This is what the “race” question looked like in 2014. Note there is no “Asian” checkbox:

As a general guide, the questionnaire scheme works best when (a) everyone has a category they like, and (b) few people choose “other.” That is the system that will yield the most scientifically useful data. It also will tend to match the way people interact socially, including how they discriminate against each other, burn crosses on each other’s lawns, and randomly attack each other in public. We want data that helps us understand all that.

Through extensive testing, it has become apparent that, when given a question that offers both race and Hispanic origin together, Latino respondents are much more likely to answer Hispanic/Latino only, rather than cluttering up the race question with “some other race” responses (often writing in “Hispanic” or “Latino” as their “other race”). If I read the presentation right, in round numbers, given the choice of answering the “race” question with “Hispanic,” in the test data about 70% chose Hispanic alone; about 20% chose White along with Hispanic, and 5% choose two races. In fact, the number of Latinos saying their only race is White probably won’t change much; the biggest difference is that you no longer have almost 40% of Latinos saying they are “some other race,” or choosing more than one race (usually White and Other) which usually just means they don’t see a race that fits them on the list.

In the end, the size of the major groups (Hispanics and the major races) are not changed much. Here’s the summary:

In fact, the only major group that will shrink is probably the non-group “multiracial” population, which today is dominated by Hispanics choosing White and “some other race.”

It’s really just better data. It’s not a conspiracy. It’s not eliminating the White race or discouraging assimilation of Hispanics. In short, keep calm and collect better data. We can fight about all that other stuff, too.

I’m sure Gonzalez doesn’t really think this will “eliminate Hispanics’ racial choices.” He’s dog-whistling to people who think the government is trying to reduce the number of Whites by not letting Hispanics be White. His statements are factually incorrect and the Washington Post shouldn’t have printed them. (I don’t know how the Post does Op-Eds; when I wrote one for the NY Times it was pretty thoroughly fact-checked.)

MENA

The Migration Policy Institute estimates there are about 2 million MENAs in the U.S. now, about half of them immigrants. This is a pretty small population, mostly Arab-speaking immigrants and their descendants, and more Christian (relative to Muslim) than the countries they left. This is especially true of the more recent immigrants, which don’t include a lot of Iranians (who aren’t Arab).

Census could have instead defined them by linguistic origin (Arab), and captured most, but they instead are going with country of origin, which is consistent with how the other race/ethnic groups are identified (for better or worse). Their testing showed that this measure captures most people with MENA ancestry, encourages them to identify their ancestry, cuts down on them identifying as White, and cuts down on them using “some other race.”

The difference is dramatic for those identifying as White, which fell from 85% to 20% in the test once a MENA category was offered. Would it be better if they just identified as White? I’m really not trying to shrink the count of Whites, I just think this is more accurate. I don’t care about the biology of Whiteness and whether Iranians are part of it, for example (and don’t ever say “Caucasian,” please), I care about the experience and identity of the people we’re talking about — as well as the beliefs of the people who hate them and those who want to protect them from discrimination. Counting them seems better than shoehorning them into a category most of them avoid when given the chance.

Here’s one version of the proposed new combined question, from that Census presentation:

Yuck

Why not Mike Gonzalez to run Census? Unbelievably, he probably knows more about it than Trump’s education and HUD department heads know about their new portfolios.

But that’s just one odious possibility. It makes me kind of sick to think of the possible idiots and fanatics Trump might put in charge of the Census Bureau, after all this work on research and testing, designed to get the best data we can out of a very messy and imperfect situation.

What else would they do? Will they continue to develop ways to identify and count same-sex couples? The Supreme Court says they can get married, but there is no law that says the Census Bureau has to count them. What about multilingual efforts to reach immigrant communities? This has been a focus of Census Bureau development as well. And so on.

It is absolutely in Trump’s interest, and the interests of those who he serves (not the people who voted for him), to reduce the quality and quantity of social science data the government produces and enables us to produce.

How do Black-White parents identify their children?

In 2015 the American Community Survey yields an estimate of 66,913 infants who have one Black parent and one White parent present in the household. (Either parent may be multiracial, too.)

What is the race of those infants? 73% of them were identified as both White and Black by whoever filled out the Census form.

(Note “other” doesn’t mean they specified “other,” it just means they used some other combination of races.)

These are children age 0 living with both parents, so it’s a pretty good bet they’re mostly biological parents, though some are presumably adopted. This is based on a sample of 507 such infants. If you pooled some years of ACS there is plenty to study here. Someone may already have done this – feel free to post in the comments.

That’s it, just FYI.

Sexual minority counts

One of the big happenings at the Population Association of American (PAA) conference, just completed, was news of progress toward collecting better data on sexual diversity.

Photo of PAA 2014 by Philip Cohen from Flickr Creative Commons

Call it weakness if you like, but in this area I am prone to viewing modernity as a march of progress from a dark past toward a half-full glass of bright future, with popular politics driving widening notions of human rights, motivating legal reforms, compelling the adoption of state bureaucracies to progressive social reality, and gradually incorporating us into a new world order more or less of our own creation.

That last part – about the bureaucracies incorporating the public – might not be the most complicated, but it is still pretty thorny. (And from here till the next subhead it gets technical.)

Good news bad news

The good news is that we have great new data collections coming along. Virginia Cain from the National Center for Health Statistics reported on their new sexual orientation question for the National Health Interview Survey, the largest federal health survey (the paper doesn’t seem to be available yet). This is already yielding important data on health disparities for sexual minorities, which is vital for policy responses to inequality.

Tim Vizard from the UK Office of National Statistics also reported on his agency’s new sexual identity question, which has been tested for several years on a few hundred thousand people each year. The latest numbers show 1.5% of adults self-identifying as lesbian, gay, or bisexual. They get these low numbers because they ask a very simple, narrow question, only on sexual identity rather than sexual attraction or sexual behavior (see other studies for the range of estimates).* Importantly, less than 4% of the UK respondents are refusing to answer, and the question is not affecting overall response rates – two big fears in the statistical agencies that appear to be receding with these and other results. Here’s how they ask it (semi-confidentially, so that in theory a husband and wife taking the survey together could both tell the interviewer they’re gay without either knowing what the other said):

The other good news is that the U.S. Census Bureau is making great strides (which I first praised here), on several tracks. First, they are working on the same-sex married couple data from the American Community Survey (ACS). At present they only release aggregate estimates of same-sex couples, differentiating between those that are married versus cohabiting (explained here).

A big reason we don’t have more data is the bad news: In another paper (just an abstract is posted, but you can ask the authors for a copy), Census analysts Daphne Lofquist and Jamie Lewis reported on their investigation into possible errors in the same-sex couple data the ACS has collected.

The background is that in a 2011 paper (linked here) Census analysts showed that a lot of seemingly same-sex couples were actually different-sex couples in which someone’s sex was miscoded.** If even a tiny percentage of different-sex couples make a mistake on the form – say, 1-in-1000 – then you would roughly double the number of same-sex couples. And they do. The paper used name-gender associations to reveal that, for example, in Texas 29% of supposedly male-male couples had one partner with a name that was used by women 95% of the time in that state – probably women accidentally marked as male.

But that 95% cutoff is a conservative estimate of the error. In the new analysis Lofquist and Lewis went further and checked same-sex couples against their Social Security records to see what sex they had recorded there. The result was shocking: 72.5% of the same-sex couples had a member whose sex didn’t match the Social Security record. Yes, some people change their sex/gender, and some people’s Social Security Records are wrong, but not that many. The much more likely culprit is simply a tiny number of straight people mismarking the sex box (there are some other technical possibilities, too).

The great thing about just asking people their marital status and sex is that you can count gay and lesbian couples without changing anything about the form (such as asking about sexual identity or orientation). That’s what all the people want who think I’m backward for worrying about couple-sex gender terminology. “C’mon!” they say, “Why do you have to label marriage as homogamous or heterogamous – just call it marriage!” Maybe someday, but at the moment that approach is producing an accuracy-crushing level of noise in the same-sex couple data.

Fortunately, Census is also moving forward with other improvements to fix this. The most important change is probably to the basic relationship question, which will soon look something like this, with couples labeled “opposite-sex” or “same-sex,” and the gender-neutral “spouse” added beside “husband/wife.” This will allow Census to check those couples that are reported as married to see if their same/opposite relationship identification matches what they reported for their sexes:

If we end up with a question like that, which seems most likely (the Census testing and development is quite far along), then we should be able to much more reliably identify same-sex couples (both married and cohabiting).

We’ll get used to this

That proposed new relationship question has 17 categories. That’s a long way from these six, in 1960 (the whole series of Census forms is here):

That goes to show you that family diversity is a state of collective mind as well as a structural reality. Building bureaucratic bins into which we pour data describing the various aspects of our lives is one of the defining elements of modern life. Eventually, I am pretty sure people will become disciplined by the new bureaucratic reality, and identities will calcify around checkboxes. That’s life under the modern state. (Even most haters, once they realize the data is being collected, will want to answer the questions accurately so they don’t get counted as gay – although, just as a few people refuse to answer race questions, there will be holdouts.)

* Identifying transgender people is much more complicated and difficult. The number of required questions and categories increases as the size of the groups in question grows smaller. This is feasible for smaller, more targeted surveys, but not in the immediate cards for the big ones (see Gary Gates’s presentation at PAA for more on this).

** I’m pretty sure Gary Gates was the first person to identify this problem, but can’t remember which paper it was in.