COVID-19 mortality rates by race/ethnicity and age

Why are there such great disparities in COVID-19 deaths across race/ethnic groups in the U.S.? Here’s a recent review from New York City:

The racial/ethnic disparities in COVID-related mortality may be explained by increased risk of disease because of difficulty engaging in social distancing because of crowding and occupation, and increased disease severity because of reduced access to health care, delay in seeking care, or receipt of care in low-resourced settings. Another explanation may be the higher rates of hypertension, diabetes, obesity, and chronic kidney disease among Black and Hispanic populations, all of which worsen outcomes. The role of comorbidity in explaining racial/ethnic disparities in hospitalization and mortality has been investigated in only 1 study, which did not include Hispanic patients. Although poverty, low educational attainment, and residence in areas with high densities of Black and Hispanic populations are associated with higher hospitalizations and COVID-19–related deaths in NYC, the effect of neighborhood socioeconomic status on likelihood of hospitalization, severity of illness, and death is unknown. COVID-19–related outcomes in Asian patients have also been incompletely explored.

The analysis, interestingly, found that Black and Hispanic patients in New York City, once hospitalized, were less likely to die than White patients were. Lots of complicated issues here, but some combination of exposure through conditions of work, transportation, and residence; existing health conditions; and access to and quality of care. My question is more basic, though: What are the age-specific mortality rates by race/ethnicity?

Start tangent on why age-specific comparisons are important. In demography, breaking things down by age is a basic first-pass statistical control. Age isn’t inherently the most important variable, but (1) so many things are so strongly affected by age, (2) so many groups differ greatly in their age compositions, and (3) age is so straightforward to measure, that it’s often the most reasonable first cut when comparison groups. Very frequently we find that a simple comparison is reversed when age is controlled. Consider a classic example: mortality in a richer country (USA) versus a poorer country (Jordan). People in the USA live four years longer, on average, but Americans are more than twice as likely to die each year (9 per 1,000 versus 4 per 1000). The difference is age: 23% of Americans are over age 60, compared with 6% of Jordanians. More old people means more total deaths, but compare within age groups and Americans are less likely to die. A simple separation by age facilitates more meaningful comparison for most purposes. So that’s how I want to compare COVID-19 mortality across race/ethnic groups in the USA. End tangent.

Age-specific mortality rates

It seems like this should be easier, but I can’t find anyone who is publishing them on an ongoing basis. The Centers for Disease Control posts a weekly data file of COVID-19 deaths by age and race/ethnicity, but they do not include the population denominators that you need to calculate mortality rates. So, for example, it tells you that as of December 5 there have been 2,937 COVID-19 deaths among non-Hispanic Blacks in the age range 30-49, compared with 2,186 deaths among non-Hispanic Whites of the same age. So, a higher count of Black deaths. But it doesn’t tell you there are 4.3-times as many Whites as Blacks in that category. So a much higher mortality rate.

On a different page, they report the percentage of all deaths in each age range that have occurred in each race/ethnic group, don’t include their percentage in the population. So, for example, 36% of the people ages 30-39 who have died from COVID-19 were Hispanic, and 24% were non-Hispanic White, but that’s not enough information to calculate mortality rates either. I have no reason to think this is nefarious, but it’s clearly not adequate.

So I went to the 2019 American Community Survey (ACS) data distributed by IPUMS.org to get some denominators. These are a little messy for two main reasons. First, ACS is a survey that asks people what their race and ethnicity are, while death counts are based on death certificates, for which the person who has died is not available to ask. So some people will be identified with a different group when they die than they would if they were surveyed. Second, the ACS and other surveys allow people to specify multiple races (in addition to being Hispanic or not), whereas death certificate data generally does not. So if someone who identifies as Black-and-White on a survey dies, how will the death certificate read? (If you’re very interested, here’s a report on the accuracy of death certificates, and here are the “bridges” they use to try to mash up multiple-race and single-race categories.)

My solution to this is make denominators more or less the way race/ethnicity was defined before multiple race identification was allowed. I put all Hispanic people, regardless of race, into the Hispanic group. Then I put people who are White, non-Hispanic, and no other race into the White category. And then for the Black, Asian, and American Indian categories, I include people who were multiple race (and not Hispanic). So, for example, a Black-White non-Hispanic person is counted as Black. A Black-Asian non-Hispanic person is counted as both Black and Asian. Note I did also do the calculations for Native Hawaiian and Other Pacific Islanders, but those numbers are very small so I’m not showing them on the graph; they’re on the spreadsheet. Note also I say “American Indian” to include all those who are “non-Hispanic American Indian or Alaska Native.”

This is admittedly crude, but I suggest that you trust me that it’s probably OK. (Probably OK, that is, especially for Whites, Blacks, and Hispanics. American Indians and Asians have higher rates of multiple-race identification among the living, so I expect there would be more slippage there.)

Anyway, here’s the absolutely egregious result:

This figure allows race/ethnicity comparisons within the five age groups (under 30 isn’t shown). It reveals that the greatest age-specific disparities are actually at the younger ages. In the range 30-49, Blacks are 5.6-times more likely to die, and Hispanics are 6.6-times more likely to die, than non-Hispanic Whites are. In the oldest age group, over 85, where death rates for everyone are highest, the disparities are only 1.5- and 1.4-to-1 respectively.

Whatever the cause of these disparities, this is just the bottom line, which matters. Please note how very high these rates are at old ages. These are deaths per 100,000, which means that over age 85, 1.8% of all African Americans have died of COVID-19 this year (and 1.7% for Hispanics and 1.2% for Whites). That is — I keep trying to find words to convey the power of these numbers — one out of every 56 African Americans over age 85.

Please stay home if you can.

A spreadsheet file with the data, calculations, and figure, is here: https://osf.io/ewrms/.

Race/ethnic intermarriage trends, 2008-2018

Rising, with gender differences.

Since 2008 the American Community Survey has been asking respondents whether they got married in the previous 12 months. Using the race/ethnicity of spouses (when they are living together), you can estimate the proportion of new marriages that cross racial/ethnic lines.

Defining such “intermarriage” is not as simple as it sounds. Some people have multiple racial or ethnic identities. Some people marry across national-origin lines within panethnic groups (e.g., Mexicans marrying Puerto Ricans). Is a Black+White Dominican marrying a White Mexican, or a Black+White person marrying a Black person, “intermarriage”? In these estimates I drop people who are not Hispanic and specified more than one race, then combine Hispanic origin and race into one, mutually exclusive 5-category variable: White, Black, American Indian, Asian/Pacific Islander, Hispanic (of any race). In other words, intrapanethic marriage (Mexicans marrying Puerto Ricans, or Filipinos marrying Koreans) is not intermarriage. I’m not saying this is the best way; it combines conventional categories with convenience. I combine same-sex and different-sex marriages.

To present the results, I separate men and women (you’ll see why), and estimate predicted probabilities of intermarriage at the mean of controls for age and age-squared, using logistic regression with normalized weights. My Stata code is on the Open Science Framework; help yourself. (I previously did something very similar for states and metro areas.)

The results are in figures, with each race/ethnic group presented on its own scale (check the y-axes). I don’t present American Indians because the samples are small (about half the API sample) and the multirace group is large.

Results

Click the images to enlarge

Framing social class with sample selection

A lot of qualitative sociology makes comparisons across social class categories. Many researchers build class into their research designs by selecting subjects using broad criteria, most often education level, income level, or occupation. Depending on the set of questions at hand, the class selection categories will vary, focusing on, for example, upbringing and socialization, access to resources, or occupational outlook.

In the absence of a substantive review, here are a few arbitrarily selected examplar books from my areas of research:

• Mothering While Black: Boundaries and Burdens of Middle-Class Parenthood, by Dawn Marie Dow
• The Tumbleweed Society: Working and Caring in an Age of Insecurity, by Alison Pugh [my review]
• Negotiating Opportunities: How the Middle Class Secures Advantages in School, by Jessica McCrory Calarco.
• For the Family?: How Class and Gender Shape Women’s Work, by Sarah Damaske

This post was inspired by the question Caitlyn Collins asked the other day on Twitter:

#soctwitter, I'd love thoughts on how to operationalize social class categories for sampling in qualitative research. Yes, multidimensional evals with income, ed, occupation.. but I want best practices for actual schema to be used as filters for fieldwork, w people, face to face.

— Caitlyn Collins (@CaitlynMCollins) June 27, 2020

She followed up by saying, “Social class is nebulous, but precision here matters to make meaningful claims. What do we mean when we say we’re talking to poor, working class, middle class, wealthy folks? I’m looking for specific demographic questions, categories, scales sociologists use as screeners.” The thread generated a lot of good ideas.

Income, education, occupation

Screening people for research can be costly and time consuming, so you want to maximize simplicity as well as clarity. So here’s a way of looking at some common screening variables, and what you might get or lose by relying on them in different combinations. This uses the 2018 American Community Survey, provided by IPUMS.org (Stata data file and code here).

• I used income, education, and occupation to identify the status of individuals, and generated household class categories by the presence of absence of types of people in each. That means everyone in each household is in the same class category (a choice you might or might not want to make).
• Income: Total household income divided by an equivalency scale (for cost of living). The scale counts each adult as 1 person, each child under 18 as .70, and then scales that count by ^.70. I divided the resulting distribution into thirds, so households are in the top, middle, or bottom third. Top third is what I called “middle/upper” class, bottom third is “lower class.”
• Education: I use BA degree to identify households that have (middle/upper) or don’t (lower) a four-year college graduate present. This is 31% of adults.
• Occupation: I used the 2018 ACS occupation codes, and coded people as middle/upper class if their codes was 10 to 3550, which are management, business, and financial occupations; computer, engineering, and science occupations; education, legal, community service, arts, and media occupations; and healthcare practitioners and technical occupations. It’s pretty close to what we used to call “managerial and professional” occupations. Together, these account for 37% of workers.

So each of these three variables identifies an upper/middle class status of about a third of people.

For lower class status, you can just reverse them. The except is income, which is in three categories. For that, I counted households as lower class if their household income was in the bottom third of the adjusted distribution. In the figures below, that means they’re neither middle/upper class nor lower class if they’re in the middle of the income distribution. This is easily adjusted.

Venn diagrams

You can make Venn diagrams in Stata using the pvenn2 add-on, which I naturally discovered after making these. If  you must know, made these by generating tables in Stata, downloading this free plotter app, entering the values manually, copying the resulting figures into Powerpoint and applying the text there, then printing them to PDF, and extracting the images from PDF using Photoshop. Not recommended workflow.

Here they are. I hope the visuals might help people think about for example, who they might get if they screened on just one of these variables, or how unusual someone is who has a high income or occupation but no BA, and so on. But draw your own conclusions (and feel free to modify the code and follow your own approach). Click to enlarge.

First middle/upper class:

Then lower class:

I said draw your own conclusions, but please don’t draw the conclusion that I think this is the best way to define social class. That’s a whole different question. This is just about simply ways to select people to be research subjects. For other posts on social class, follow this tag, which includes this post about class self identification by income and race/ethnicity.

---

Data and code: osf.io/w2yvf/

The arriving divorce decline

In “The Coming Divorce Decline” I showed the U.S. divorce rate falling from 2008 to 2017, and predicted that, because the married population was being stocked with increasingly non-divorce-prone marriages, the rate would continue to fall. After the first draft (based on 2016 data), divorce fell in 2017, providing the first support for my prediction before the paper was even “published” (accepted for Socius). Now the 2018 data is out, and divorce has become less common still.

Here’s a quick update.

Based on the number of divorces reported in the survey each year, by sex, and the number of married people, I calculate the refined divorce rate, or the number of divorces per 1,000 married people. That fell another 3% for both women and men in 2018, to 15.9 and 14.3 respectively (the rates differ because these are self reports and women report more).

When I run the model from the paper again on the new data (on women only), I can show the drop in the adjusted odds of divorce, updating Figure 1 of the paper (the 2018 change in an unadjusted model is significant at p=.06; adjusted is p=.14, the adjusted change from 2016 is significant at p=.002).

For other takes on the latest data, see this report on the marriage-divorce ratio from Valerie Schweizer, and this on geographic variation from Colette Allred, both at the National Center for Family and Marriage Research.

---

• The data and code for the paper are available here. This update uses the same code with one new year of data.
• If you like my new Stata figure scheme (modified from Gray Kimbrough’s Uncluttered) you’re welcome to it: here.
• Slides from my presentation this fall at the European Divorce Conference are here.
• Divorce posts are gathered under this tag.

Family diversity, new normal

Family diversity is not just a buzzword (although it is that), and it’s not just the recognition of diversity that always existed (although it is that). There really is more actually-existing diversity than there used to be.

In The Family, I use a figure with five simple household types to show family conformity increasing from 1900 to a peak in 1960 — and then increasing diversity after that. I’ve updated that now for the upcoming third edition of the book.

In 2014, I wrote a report for the Council on Contemporary Families called “Family Diversity is the New Normal for America’s Children,” which generated some news coverage and a ridiculous appearance with Tucker Carlson on Fox & Friends. A key point was to demonstrate that the declining dominant family arrangement after 1960 — the male-breadwinner-homemaker family — was replaced by a diversity of arrangements rather than a new dominant form. Here I’ve updated one the main figures from that report, which shows that “fanning out from a dominant category to a veritable peacock’s tail of work-family arrangements.”

For this update, I take advantage of the great new IPUMS mother and father pointers to identify children’s (likely) parents, including same-sex couple parents who are cohabiting as well as those who are married. Census doesn’t collect multiple parent identifiers in the Decennial Census or American Community Survey, and IPUMS has tackled the issue of how to best presume or guess about these with a consistent and well-documented standard. In this figure, 0.42% of children ages 0-14 (about 250,000) are living in the households of their same-sex couple parents. I also rejiggered the other categories a little, but the basic story is the same.

I published a version of this figure for K-12 educators in Educational Leadership magazine in 2017. I wrote:

Today, teachers need to have a more inclusive mindset that recognizes the diversity of family structures. Although there are reasons for concern about some of the changes shown in the data, the driving factors have often been positive. For example, changes in family roles reflect increased educational and occupational opportunities for women and greater gender equality within families. Fathers are expected to play an active role in parenting—and usually do—to a much greater degree than they did half a century ago.

My advice to teachers is:

The key points of diversity in family experiences that teachers should watch for are family structure (such as who the student lives with), family trajectories (the transitions and changes in family structure), and family roles (who cares and provides for the student). Using principles from universal design, teachers can promote language and concepts that work for all students. Done right, this is an opportunity to broaden the learning experience for everyone—to teach that care, intimate relationships, and family structures can include people of different ages, genders, and familial connections.

So that’s my update.

New working paper: The rising marriage mortality gap among Whites

I wrote a short working paper on U.S. mortality trends for the last decade. You can go straight to the paper on SocArXiv, or the code and output, if you want the full version.

The issue is that premature mortality has been rising for Whites, partly because of the opioid epidemic and also from suicide and alcohol, and also from other causes related to stress and hardship. (See, e.g., Case and Deaton, and Geronimus.) And a recent NCHS report showed that mortality nationally declined much more for married people since 2010.

So I got the Mortality Multiple Cause Files from the National Center for Health Statistics, for two years: 2007 and 2017. These are a complete set of death certificates, which include race/ethnicity, marital status, and education. I linked these to the American Community Survey, to create age-specific mortality rates by age, sex, marital status, and education, for non-Hispanic Whites, Hispanics, and Blacks, in the ages 25-74 (old enough to finished with college, but too young to die).

The basic result is that virtually all of the growth in premature death is among Whites, and further among non-married Whites. (Whites still dies less than Blacks, and more than Hispanics, at each age and marital status.)

Here is the figure of age-specific mortality rates, by race/ethnicity, sex, and marital status for 2007 and 2017. At the bottom of each column I calculated “marriage mortality ratios,” which are how much more likely single people are to die than married people. Note these death rates are deaths per 10,000, but they’re on a log scale so you can see changes where rates are very low.

In the figure you can see how much the marriage mortality ratio jumped up, for Whites only. Now, at the most extreme, single White men age 35-39 are more than 4-times more likely to die than married White men (that’s in the bottom left).

Then I zoom into Whites specifically, and do the same thing for four levels of education:

In the lowest education group of Whites (the far left), mortality rates for married and single people increased similarly, so the marriage mortality ratio didn’t increase. However, for the other education levels, death rates increased for single people more than married people, so the ratio increased (across the bottom). Even among White college graduates, there were increases in mortality for single people. I did not expect that.

My bottom line is that marriage is taking an ever-more prominent place in the social status hierarchy, and now we can add growing mortality inequality, at least among Whites, to that pattern.

Early version, comments welcome!

The Coming Divorce Decline, Socius edition

“The Coming Divorce Decline, ” which I first posted a year ago, has now been published by the journal Socius.  Three thousand people have downloaded it from SocArXiv, I presented it at the Population Association, and it’s been widely reported (media reports), but now it’s also “peer reviewed.” Since Socius is open access, I posted their PDF on SocArXiv, and now that version appears first at the same DOI or web address (paper), while the former editions are also available.

Improvement: Last time I posted about it here I had a crude measure of divorce risk with one point each for various risk factors. For the new version I fixed it up, using a divorce prediction model for people married less than 10 years in 2017 to generate a set of divorce probabilities that I apply to the newly-wed women from 2008 to 2017:

…the coefficients from this model are applied to newly married women from 2008 to 2017 to generate a predicted divorce probability based on 2017 effects. The analysis asks what proportion of the newly married women would divorce in each of their first 10 years of marriage if 2017 divorce propensities prevailed and their characteristics did not change.

The result looks like this, showing the annual probability falling from almost 2.7% to less than 2.4%:

The fact that this predicted probability is falling is the (now improved) basis for my prediction that divorce rates will continue to decline in the coming years: the people marrying now have fewer risk factors. (The data and code for all this is up, too).

---

Prediction aside: The short description of study preregistration is “specifying your plan in advance, before you gather data.” You do this with a time-stamped report so readers know you’re not rejiggering the results after you collect data to make it look like you were right all along. This doesn’t always make sense with secondary data because the data is already collected before we get there. However, in this case I was making predictions about future data not yet released. So the first version of this paper, posted last September and preserved with a time stamp on SocArXiv, is like a preregistration of the later versions, effectively predicting I would find a decline in subsequent years if I used the same models — which I did. People who use data that is released on a regular schedule, like ACS, CPS, or GSS, might consider doing this in the future.

---

Rejection addendum: Sociological Science rejected this — as they do, in about 30 days, with very brief reviews — and based on their misunderstandings I made some clarifications and explained the limitations before sending it to Socius. Since the paper was publicly available the whole time this didn’t slow down the progress of science, and then I improved it, so I’m happy about it.

Just in case you’re worried that this rejections means the paper might be wrong, I’m sharing their reviews here, as summarized by the editor. If you read the current version you’ll see how I clarified these points.

* While the analyses are generally sensible, both Consulting Editors point out the paper’s modest contribution to the literature relative to Kennedy and Ruggles (2014) and Hemez (2017). The paper cites both of these papers but does not make clear how the paper adds to our understanding derived from those papers. If the relatively modest extension in the time frame in this paper is sociologically consequential, the paper does not make the case clearly.

* There is more novelty in the paper’s estimates of the annual divorce probability for newly-married women (shown in Table 3 and Figure 3), based on estimating a divorce model for the most recent survey year, and then applying the coefficients from that model to prior years. This procedure was somewhat difficult for the readers to follow, but issues were raised, most notably the question of the sensitivity of the results to the adjustments made. As one CE noted, “Excluding those in the first year of marriage is problematic as newlyweds have a high rate of divorce. Also, why just married in the last 10 years? Consider whether married for the first time vs remarried matters. Also, investigate the merits of an age restriction given the aging of the population Kennedy and Ruggles point to.”

The changing household age range, U.S. 1900-2017

One way to understand daily interaction, and intergenerational resource exchange, is just to look at the structure of households. This doesn’t tell you everything that goes on in households, but it gives some strong clues. And we can measure it going back more than a century, thanks to IPUMS.org’s collection of Census microdata.

In 1900, the most common situation for an American was to live in a household where the age difference between the oldest and youngest person was about 38 years. Now the most common situation is an age range of 0 — either living alone, or with someone of the exact same age. And there are a lot more people living in households with only similar-aged adults, with age ranges of less than 10.

In between 1900 and 2017, life expectancy increased, the age at first birth increased, and the tendency to live in multigenerational households fell and then rose again. So the household structure story is complicated, and this is just one perspective.

But it’s one indicator of how life has changed. Line up your household from youngest to oldest, look to your left and look to your right — how far can you see?

 

Data and Stata code (for all decades 1900-2000, then individual years to 2017) are available on the Open Science Framework, here.

Let’s raise the legal age of marriage in Maryland

Today I sent the following letter to the Maryland House Judiciary Committee, which is scheduled to hold a hearing on these bills tomorrow. Under current law in Maryland, marriage is permitted as young as age 15 with parental consent and evidence of pregnancy or childbirth, and age 16-17 with one or the other, and these exceptions are granted by county clerks rather than judges. By my calculations, from 2008 to 2017, based on the American Community Survey, the annual marriage rate for girls ages 15-16 was 5 per 1000 in Maryland, behind only Hawaii, Nevada, and West Virginia. HB 855 would raise the age at marriage to 18, while HB 1147 would establish an emancipated minor status, requiring review by a judge, under which 17-year-olds could marry. For more on the effort to end child marriage in the U.S., visit the Tahirih Justice Center site.

---

March 6, 2019

To the House Judiciary Committee:

I write in support of Maryland House Bill 855, concerning age requirements for marriage; and House Bill 1147, concerning the emancipation of minors.

My relevant background

• I am a Professor of Sociology, and family demographer, at the University of Maryland, College Park, where I have been on the faculty since 2012. I also earned my PhD at the University of Maryland, College Park, in 1999, and I live in Silver Spring.
• I have written two books and many peer-reviewed articles on family sociology, including on topics related to marriage and divorce, family structure, gender inequality, health and disability, infant mortality, adoption, race and ethnicity, and the division of labor.
• I have served as a consultant to the U.S. Census Bureau on the measurement of family structure, and testified before Congress on gender discrimination.

My support of the bills

In general, the rise of the age at marriage and childbearing in U.S. have been positive developments for women and children, allowing mothers to devote more years of early adulthood to education and career development, which is beneficial to both adults and their children.

Very early marriage in particular is detrimental to women’s opportunity to finish high school. More urgently, research and service work shows that very early marriage is usually unwanted, coerced, or forced. Very young women should not be expected to protect themselves legally or socially from such impositions, which are usually from older men and dominant family members. Very early marriage often follows statutory rape or other sexual assault, compounding rather than mitigating the harms of these crimes against children. Rather than protect a young woman, very early marriage instead provides protection from scrutiny for her abuser(s), and makes state intervention on her behalf all the more difficult to accomplish in the following years. The privacy and discretion we bestow upon families has benefits, of course, but it also makes the family a dangerous place for the victims of abuse.

Research, including my own, unequivocally shows that very early marriage leads to the highest rates of divorce. I have written several papers on divorce rates in the United States (see references). For illustration, here I used the same method of analysis, and present only the relationship between age at marriage and incidence of divorce. As you can see from the figure, divorce rates are highest by far – estimated at 2.5% per year – for women who married before age 18. This is about twice as high as divorce rates for those who marry in their 30s, for example. (These estimates hold constant other factors; data and code are available here.) The evidence is very strong.

I only reluctantly support increasing state restrictions on women’s freedom with regard to family choices, but in the case of marriage before adulthood I see the restriction as a protection from the exploitative behavior of others, rather than an imposition on young women’s rights.

At present in Maryland, exceptions allowing marriage before age 18 – based on pregnancy and/or parental consent – are granted without adequate legal review. Together, HB 855 and HB 1147 would set the minimum age at marriage in Maryland to 18, with an exception only for court emancipated minors of age 17. This would improve the state’s protection of young women from unwanted, coerced, forced, or ill-advised marriages without unduly restricting the freedom to marry for younger women (age 17), who may be emancipated by a court after a direct application and careful review of circumstances.

I urge your support for these bills. I would be happy to provide further information or testimony at your request.

Sincerely,

Philip N. Cohen

References

Cohen, Philip N. 2015. “Recession and Divorce in the United States, 2008-2011. Population Research and Policy Review 33(5):615-628.

Cohen, Philip N. 2018. “The Coming Divorce Decline.” SocArXiv. November 14. https://osf.io/preprints/socarxiv/h2sk6. To be presented at the Population Association of America meetings, 2019.

Decadally-biased marriage recall in the American Community Survey

Do people forget when they got married?

In demography, there is a well-known phenomenon known as age-heaping, in which people round off their ages, or misremember them, and report them as numbers ending in 0 or 5. We have a measure, known as Whipple’s index, that estimates the extent to which this is occurring in a given dataset. To calculate this you take the number of people between ages 23 and 62 (inclusive), and compare it to five-times the number of those whose ages end in 0 or 5 (25, 30 … 60), so there are five-times as many total years as 0 and 5 years.

If the ratio of 0/5s to the total is less than 105, that’s “highly accurate” by the United Nations standard, a ratio 105 to 110 is “fairly accurate,” and in the range 110 to 125 age data should be considered “approximate.”

I previously showed that the American Community Survey’s (ACS) public use file has a Whipple index of 104, which is not so good for a major government survey in a rich country. The heaping in ACS apparently came from people who didn’t respond to email or mail questionnaires and had to be interviewed by Census Bureau staff by phone or in person. I’m not sure what you can do about that.

What about marriage?

The ACS has a great data on marriage and marital events, which I have used to analyze divorce trends, among other things. Key to the analysis of divorce patterns is the question, “When was this person last married?” (YRMARR) Recorded as a year date, this allows the analyst to take into account the duration of marriage preceding divorce or widowhood, the birth of children, and so on. It’s very important and useful information.

Unfortunately, it may also have an accuracy problem.

I used the ACS public use files made available by IPUMS.org, combining all years 2008-2017, the years they have included the variable YRMARR. The figure shows the number of people reported to have last married in each year from 1936 to 2015. The decadal years are highlighted in black. (The dropoff at the end is because I included surveys earlier than those years.)

Yikes! That looks like some decadal marriage year heaping. Note I didn’t highlight the years ending in 5, because those didn’t seem to be heaped upon.

To describe this phenomenon, I hereby invent the Decadally-Biased Marriage Recall index, or DBMR. This is 10-times the number of people married in years ending in 0, divided by the number of people married in all years (starting with a 6-year and ending with a 5-year). The ratio is multiplied by 100 to make it comparable to the Whipple index.

The DBMR for this figure (years 1936-2015) is 110.8. So there are 1.108-times as many people in those decadal years as you would expect from a continuous year function.

Maybe people really do get married more in decadal years. I was surprised to see a large heap at 2000, which is very recent so you might think there was good recall for those weddings. Maybe people got married that year because of the millennium hoopla. When you end the series at 1995, however, the DBMR is still 110.6. So maybe some people who would have gotten married at the end of 1999 waited till New Years day or something, or rushed to marry on New Year’s Eve 2000, but that’s not the issue.

Maybe this has to do with who is answering the survey. Do you know what year your parents got married? If you answered the survey for your household, and someone else lives with you, you might round off. This is worth pursuing. I restricted the sample to just those who were householders (the person in whose name the home is owned or rented), and still got a DBMR of 110.7. But that might not be the best test.

Another possibility is that people who started living together before they were married — which is most Americans these days — don’t answer YRMARR with their legal marriage date, but some rounded-off cohabitation date. I don’t know how to test that.

Anyway, something to think about.