Two examples of why “Millennials” is wrong

When you make up “generation” labels for arbitrary groups based on year of birth, and start attributing personality traits, behaviors, and experiences to them as if they are an actual group, you add more noise than light to our understanding of social trends.

According to generation-guru Pew Research, “millennials” are born during the years 1981-1997. A Pew essay explaining the generations carefully explains that the divisions are arbitrary, and then proceeds to analyze data according to these divisions as if are already real. (In fact, in the one place the essay talks about differences within generations, with regard to political attitudes, it’s clear that there is no political consistency within them, as they have to differentiate between “early” and “late” members of each “generation.”)

Amazingly, despite countless media reports on these “generations,” especially millennials, in a 2015 Pew survey only 40% of people who are supposed to be millennials could pick the name out of a lineup — that is, asked, “These are some commonly used names for generations. Which of these, if any, do you consider yourself to be?”, and then given the generation names (silent, baby boom, X, millennial), 40% of people born after 1980 picked “millennial.”

“What do they know?” You’re saying. “Millennials.”

Two examples

The generational labels we’re currently saddled with create false divisions between groups that aren’t really groups, and then obscure important variation within the groups that are arbitrarily lumped together. Here is just one example: the employment experience of young men around the 2009 recession.

In this figure, I’ve taken three birth cohorts: men born four years apart in 1983, 1987, and 1991 — all “millennials” by the Pew definition. Using data from the 2001-2015 American Community Surveys via IPUMS.org, the figure shows their employment rates by age, with 2009 marked for each, coming at age 26, 22, and 18 respectively.

Each group took a big hit, but their recoveries look pretty different, with the earlier cohort not recovered as of 2015, while the youngest 1991 group bounced up to surpass the employment rates of the 1987s by age 24. Timing matters. I reckon the year they hit that great recession matters more in their lives than the arbitrary lumping of them all together compared with some other older “generations.”

Next, marriage rates. Here I use the Current Population Survey and analyze the percentage of young adults married by year of birth for people ages 18-29. This is from a regression that controls for year of age and sex, so it can be interpreted as marriage rates for young adults (click to enlarge).

From the beginning of the Baby Boom generation to those born through 1987 (who turned 29 in 2016, the last year of CPS data), the marriage rate fell from 57% to 21%, or 36 percentage points. Most of that change, 22 points, occurred within the Baby Boom. The marriage experience of the “early” and “late” Baby Boomers is not comparable at all. The subsequent “generations” are also marked by continuously falling marriage rates, with no clear demarcation between the groups. (There is probably some fancy math someone could do to confirm that, with regard to marriage experience, group membership by these arbitrary criteria doesn’t tell you more than any other arbitrary grouping would.)

Anyway, there are lots of fascinating and important ways that birth cohort — or other cohort identifiers — matter in people’s lives. And we could learn more about them if we looked at the data before imposing the categories.

What a recovery looks like (with population growth by age)

If you don’t account for population growth, I don’t get what you’re saying with these employment numbers. I’ll show a simple example, but first a little rundown on Friday’s jobs report.

Here is how CNN Money played the jobs report:

What does it mean, this loss and gain of jobs, returning finally to where we started? Four paragraphs under that happy headline, CNN did points out:

Given population growth over the last four years, the economy still needs more jobs to truly return to a healthy place. How many more? A whopping 7 million, calculates Heidi Shierholz, an economist with the Economic Policy Institute.

So what does “Finally!” mean? The Wall Street Journal ran the headline, “Jobs Return to Peak, but Quality Lags.” On 538 it was, “Women returned to prerecession levels of employment in 2013. Men remain hundreds of thousands of jobs in the hole:”

The Center on Budget and Policy Priorities showed how much better the previous recoveries were:

That’s a good comparison. And CBPP mentioned population growth, too:

…payroll employment has finally topped its level at the start of the recession. Still, with essentially no net job growth since December 2007 but a growing working-age population, many more people today want to work but don’t have a job.

It’s not that no one mentions population growth, it’s that they still lead with the “top line” number. And they all have that horizontal line at the raw number of jobs when the recession started as the benchmark. I don’t know why.

Maybe in some crazy economics world the absolute number of jobs is what really matters for evaluating a recovery, and that explains the fixation on that horizontal line. From a social perspective what matters is the proportion of people with jobs. I could see the logic if you had a finite number of employers that never change, where you could literally count up the jobs at two points in time, and see who added and who subtracted from their payrolls (this is why retail chains report “same-store” trends, so the statistics aren’t confounded by the changing number of stores). But we have zillions of employers, constantly changing and moving around — largely in response to population changes. So that static image seems pointless.

In perspective

So here are some charts to put the recession and recovery in slightly better perspective. These all use the Bureau of Labor Statistics’ Current Population Survey from March 2003 to March 2013 (from IPUMS), the household survey used to track the labor force. I use ages 15 and older, and combine people in school (up to age 24) with those employed (not how most people do it, but a lot of people went to school, or stayed in school, because of the bad job market, and it doesn’t make sense to count them as not simply not employed). The survey excludes people in institutions, like prisons, and on-base military personnel.

To show the basic issue, here are the changes in the non-institutionalized population, age 15+, along with the number of them employed or in school — showing absolute changes relative to 2008, the peak employment year.

The 15+ population increased almost 12 million from 2008 to 2013. People employed or in school was not yet back to 2008 levels in March 2013. So a basic population adjustment is the least you can ask for (and we get that from the BLS with the employment-population ratio, which as of May was up less than one percent in the last 3.5 years, but it’s not the headline number).

What about age shifts? You don’t expect extreme age composition changes in 5 years, but there are different employment trends at different ages, so those affect how many employed people we are short. Here are the trends in work/school, by age and sex:

This makes it look like the 30-49s that are getting crushed. The 50+ community’s gains, however,are deceptive — their population is increasing. In fact, the population of people 30-49 declined 5% during this decade, while the population 50+ increased almost 30%. The younger people have increased their schooling rates, but their population has also grown. If you look at the employment/school rates, you see that among men, it is the younger groups that have done worst:

Women clearly are doing better (partly because in the 20-29 range they’re going to school more). It is amazing that employment rates didn’t fall at all over age 60. This could be because the population increase in that group is all in Baby Boomers just hitting their sixties, but I reckon it’s also people delaying retirement compensating for unemployment.

Now that we have age-specific work/school rates, and population changes, we can easily calculate how many people in each age group would have to be in work/school to get to the number implied by applying the peak-year 2008 rates to the population in each year. Sorry this one is so ugly: I made the last bar for each group pink to show the bottom line, where each group stands in 2013 relative to 2008:

Worst off are the 20-something men, down more than a million worker/students in 2013. Interestingly, women are only better off in the 20-something and 50+ ranges.

Finally, if you sum these figures you get the total, age-adjusted losses, by sex since 2008, as of March 2013:

And that’s your bottom line. The job/school losses stood at 3.3 million for men and 2.4 million for women as of March 2013.*

Really, there are no huge surprises here. In fact, the total population change is not a bad rough adjustment, especially for the short term. But there are some interesting nuances here. And with all the data and computers we have these days, let’s adjust for age and sex.

*I don’t say that’s how many “jobs” we need, because I don’t think “jobs” exist — are created, destroyed, shipped overseas, etc. I think there are employed people, people getting jobs, losing jobs, etc. I don’t see how a “job” exists absent a worker in it (and no, a listing is not a job until they fill it). So we don’t need to “create jobs” after a recession, what we need to do is “hire people.” Pet peeve.

Peak women, labor force participation edition

I had a great visit at the University of Pennsylvania the other day, and gave a talk titled, “What Happened to the Gender Revolution?” It was an elaboration of the op-ed I wrote last fall, in which I sketched out the stall in progress toward gender equality (a recurring theme, not my discovery) and offered some ideas about getting it moving again.

One objection I got during the talk (rather belligerently, from Herbert Smith) was that I was making a big deal out of women’s labor force share peaking at just under half the total, which is a natural place to peak and so we shouldn’t expect it to keep going up.

My first response was that the feminism-has-gone-too-far gang (Hanna Rosin, Kay Hymowitz, Christina Hoff Sommers, etc.) complains as if women’s progress has already shot past 50/50. Although it hasn’t on almost all measures, there’s also no reason why women couldn’t become dominant. Judging from history, one gender dominating the labor market is hardly an impossibility. So women’s labor force share tapering off as it approaches 50% shouldn’t be considered a natural phenomenon.

But second, and for this I blame my presentation, women’s share of the labor force isn’t the best measure because it depends also on men’s labor force participation, too, which has been falling since the 1960s. So maybe it’s best to focus on women’s participation rates instead (it is on this measure that the U.S. has slipped behind many other rich countries).

Here are the labor force participation rates for women by age, education, race/ethnicity, and marital status, from 1962 to 2013, from the Current Population Survey, with men for comparison. The dots show the peak year for each trend (click to enlarge).

Women’s overall share of the labor force hit 46% in 1994, and has spent the last 20 years within a point of that (as both men’s and women’s rates fell). But if you look at all these groups it’s clear that doesn’t represent the simple slide of women into the home plate of equality. Every line here rose for decades before hitting a peak between 1996 and 2001. And they peaked at different levels: Women with BA degrees peaked at 85%, Black women peaked at 80%, Hispanic women peaked at 68%. Married women peaked at 75%, single women at 82%. And so on.

Maybe all these trends are not being driven by the same underlying forces. But I’m pretty sure it’s not a complete coincidence.

That number you want, it is not precise (women’s labor force edition)

Everyone wants a number. You want to know if the number is different from last year, or 100 years ago. Numbers are great. But the number you’re using is usually a statistic, a number calculated from a sample drawn from a population. You want a good number, you need a good sample. And a big one. And that’s going to cost you.

Who didn’t love the news recently that single British men ages 18-25 change their bedsheets only four times a year? Really? Really. How does anyone know this? Ergoflex, a memory-foam mattress distributor. At least UPI had the decency to report, “No survey details were provided,” although somehow Time found out the sample size was 2,004 (men and women, all ages). Rubbish, I reckon, or bonkers, or whatever. No one can resist a number; methods details don’t make it into the tweet version of the press release.

Here’s a more answerable question: What is the labor force participation rate for married, college graduate women with children, ages 25-54 in the United States? I’d say 76.1% — plus or minus a percentage point — based on the gold standard for labor force data collection, the Current Population Survey, easily analyzable these days for free with the IPUMS online tool.That’s from a sample of 60,000 households with a 90+% response rate, at a cost of umpteen million taxpayer dollars (well spent).

Here’s the trend in that number from 1990 to 2012, with 95% confidence intervals, based on the sample size, as calculated by IPUMS:

As more women have gotten college degrees, and the CPS sample has been enlarged, the sample size for this trend has grown and the error bars have shrunk, from a spread of almost 3 points to just less than 2. Still, there are only 8,265 of these women in the sample.

Only! Hold that up to a Gallup or Pew poll and compare confidence intervals when they start dividing and subdividing their samples. (Nothing against them — they give us the information we need to know how much variance there is in the estimates they put out, and then most people [+/- 51%] ignore it.)

There aren’t many one-year changes in this trend that are statistically significant at conventional levels. Of course, with this sample size you could say with confidence the labor force participation rate was higher in the late 1990s than the early 1990s (but check the survey redesign in 1994…), and higher again in the late 2000s than in the early 2000s. But were 2007 and 2002 sample flukes? And if so, what about 2012?

What about if you want a slightly smaller subgroup, say, Black married, college graduate women with children, ages 25-54. That’s a reasonable question. Here’s the trend (note the y-axis scale changed):

Now the sample size is a couple hundred and the confidence intervals are more than 6 points wide; there isn’t a pair of years in the trend that doesn’t have overlapping confidence intervals. And look at 2007 and 2012 — Black women are blipping in the opposite direction from the larger group in each of those years. Yes, if you put the whole Black trend in the blender with a time trend you have a significant decline of about a fifth of a point per year on average (and a sliver of this change is because of the increasing tendency of college graduates to be in grad school and not working — there are 13 of them in 2012, dragging down the participation rate by 0.6%). But don’t hang a lot on one year.

So, my advice for doing simple description:

• Eyes on the prize: who cares what the exact number is? Is it a lot or little, going up or going down, higher or lower than some other group? That’s usually what matters.
• Stick to data with reported methods
• Know the size of your subsamples, try to get confidence intervals
• Don’t fixate on (or report) small changes or differences (don’t use that second decimal place if the margin of error is 6%)
• For trends, pool data from multiple years, or report moving averages
• Spend tax money on surveys, not war

Opting out and jumping in

Because the opt-out revolution was a myth, it’s hard to know where to start talking about what became of the the “opt-out generation.”

The historical story could be shortened down to this: Women’s employment rates — and those of married mothers with young children in particular — rose quickly in the 1970s and 1980s, but that growth stalled in the 1990s. Since then, the trends are mostly flat. No successful attempts to impose grander themes on the more recent trends come to mind, but feel free to reference them in the comments if you know of any.*

I’ll show one older chart before showing you what’s new, inspired by the Judith Warner cover story in the NYTimes magazine last week. Back in 2007 I made the following graph to contribute to this briefing paper. Using the March Current Population Surveys, it shows the employment rates for married mothers of young children divided into four groups according to how much income their households had apart from their own earnings:

Wives with access to little other income (who need jobs most) and those with access to lots of other income (who need them least) have the lowest employment rates. But the similar shapes of these curves reinforces the take-home message: regardless of economic need — as represented by other income in the household — married mother’s tendency to be employed peaked in the mid-to-late 1990s and then stalled or tapered downward into the 2000s. That’s not a revolution (for most people nothing changed after the mid 1990s), and it’s not about rich professionals, but it is a serious divergence from several decades of rapid increase. Since then, this narrative has strengthened, and we now have a full-blown situation with stalled progress toward gender equality (for many posts on this, see also the Hanna Rosin tag).

However, we can be more specific to capture Lisa Belkin’s and Judith Warner’s opting-out and jumping in concepts. Let’s use these definitions:

• Opting out: The movement from any employment in one year to being out of labor labor force in March of the following year.
• Jumping in: The movement from no employment in the previous year to being in the labor force the following March.

To approximate the groups that inspired the NYTimes, I apply these definitions to 25-54-year-old, married, college-educated women with children, using the March CPS from 1976 to 2012, to get these trends (shown with approximate timing of recessions):

One can see long-term and short-term trends in the figure. In the long run, opting out has become much less common, dropping from nearly 18% in a given year to less than 3% in 2012. In the long run, the opt-out revolution is a bust. And the tendency to jump in for those who weren’t employed in a given year (I don’t know if it’s back in, with these data), has been pretty flat at between 8% and 10%, with increases after most recessions.

Notice some short-term trends, however. From 1995 to 2003, the opt-out trend stopped heading downward. That happened again from 2006 to 2012. And over the 2000s there was a modest increase in the jump-in rate, from a low point of 6.4% to over 10%. These fluctuations contribute to the overall trend of labor force participation for 25-54-year-old, married, college-educated women with children, which fell from 1997 to 2003, and then rebounded some until 2009:

Note that opting out and jumping in aren’t the only things that matter. A woman who starts employment at 24 and never leaves till 65 — or one who never has a job — wouldn’t contribute to either trend.

The opting out and jumping in stories aren’t crazy, they’re just exaggerations of fluctuations in the trends, and they distract us from the bigger picture. That said, there are good stories to be told in here, and both of these provided fodder for improving our understanding.

In the long run, the long-term trends matter more. And that is pretty clear: big increase in labor force participation for this group of women from the 1970s to the mid-1990s, stall since. The fact that opting-out shows a continued decline, however, is an interesting wrinkle I’m not prepared to explain.

UPDATE: Reeve Vanneman sent along the figure he describes in the comments below. The definitions are a little different (not just college educated mothers, employment instead of labor force participation), but the trend goes back to 1963, showing the increase in jump-ins and drop in opt-outs in the 60s and 70s. Note, however, that where Reeve et al. found exits plateauing by the early 2000s, my figure above shows opt-outs started declining again after that.

*For background, I recommend Pamela Stone’s book Opting Out: Why Women Really Quit Careers and Head Home; a Council on Contemporary Families briefing paper from 2007; and the paper by Christine Percheski in American Sociological Review from 2008, “Opting Out? Cohort Differences in Professional Women’s Employment Rates from 1960 to 2005.”

Women’s Employment and the Decline in Marriage Are No Longer Related

Originally published on TheAtlantic.com.

For a few decades, women’s rising share of the workforce probably led to fewer women getting married. But that’s not the case anymore.

CBS

It is common knowledge—and true—that marriage rates are falling and unmarried parenting is becoming more common (nicely illustrated here). On the other hand, it is also common knowledge—but not true—that women’s employment rates have continued to rise in the last two decades (as illustrated here.)

In the long run of history, there is little doubt these trends are related: As women’s economic independence increased with better job opportunities, marriage became more optional and fewer women got (or stayed) married. But in the medium run, on the scale of a few decades rather than long eras, it’s not that simple.

Here are the trends in marriage and labor force participation for women using U.S. Census data going back to 1900.

Source: My analysis of Census data from IPUMS.

In the long run of the past 111 years, there certainly are more employed women and more single women. But the trends only moved strongly in the same direction for the three decades from 1960 to 1990, when the percent of women not married more than doubled from 18 percent to 43 percent and the percent in the labor force almost doubled from 41 percent to 76 percent. In the last two decades labor force participation has frozen while the percent not married has jumped another 7 points.

Here is the trick: Despite the real connection between non-marriage and employment—in which women don’t feel as strong a need to be married if they are employed—the lion’s share of rising employment has been among married women. Women’s employment opportunities made non-marriage more viable but also changed marriage. As the employment rates of married and non-married women grew more similar, the decline of marriage has made less of a difference to the total employment rate. Moving women from married to single doesn’t do much anymore. Here are the employment trends:

The American Stall
So we need to understand the stalled rise in employment because it may be the key to understanding progress toward gender equality generally.

In a previous post I suggested that stalled progress resulted from feeble work-family policy, anti-feminist backlash, and weak anti-discrimination enforcement. A recent analysis by economists Francine Blau and Lawrence Kahn lends support to the first: work-family policy. Economix writer Catherine Rampbell highlighted the paper, which tracked employment rates over 22 wealthy countries for two decades. During that time, U.S. women fell from sixth to 17th in labor force participation rates—rising just one percentage point while women in the average country increased 12 points.

Here are the labor force participation rates for the 22 countries for 1990 and 2010. Dots to the left of the blue line show countries where women’s labor force presence increased; dots to the right show decreases. At the extreme, for example, Ireland saw a jump from 45 percent to 72 percent.

Source: My chart from the Blau and Kahn paper.

What happened? One big change was the advance of several work-family policies. The average number of weeks of guaranteed parental leave increased from 37 to 57 in these countries, with the U.S. adding only a 12-week rule under the Family Medical Leave Act (covering only half the workforce). The average country on this list now provides a guaranteed 38 percent of parents’ wages while they’re on leave, while the U.S. provides none. Seven of the countries now protect a right to part-time work, and three-quarters guarantee equal treatment for part-time workers. Public spending on child care as a proportion of GDP increased by more than a third outside the U.S., and the average country now spends more than four-times as much as the U.S.

Together, based on the experience of these countries, Blau and Kahn estimate these changes account for more than a quarter of U.S. women’s slippage relative to other countries. That’s not everything, but it’s a substantial bite. If we had kept up with the average country’s policies, U.S. women would have had an 82 percent labor force participation rate, putting them at 11th on the list instead of 17th.

On the Other Hand
Not all work-family policies are the same. One way to divide them is between those that protect time out of paid work (parental leave, part-time protections) and those that protect time in paid work (especially state-supported childcare). As Blau and Kahn note, U.S. women have much lower rates of part-time work than those in most other rich countries, but we also have higher rates of women in professional and managerial jobs. That might be because employers in those countries are reluctant to hire or promote women who are expected to take time out of the labor force when they have children—which is exactly the goal of some of our low-fertility peer countries. How, and whether, such policies can improve family life while also promoting gender equality is the subject of a rich debate—which unfortunately remains in the realm of the hypothetical here in the U.S.

Married women learning that paid work pays

The economist Raquel Fernández has a new paper out called, “Cultural Change as Learning: The Evolution of Female Labor Force Participation over a Century” (published version here, free version here). If I understand it, though, “female” labor force participation really only refers to married women. Correct me if you understand this better than I do and I’m wrong, but I think that’s a problem for the theory.

The basic point is that married women learn from the experience of others, producing a generational change in employment rates as positive experiences transmit to younger cohorts. As it became more culturally acceptable for married women to have jobs, the cultural effect accelerated, but it reached a saturation point resulting in the stalled progress toward higher employment rates among (married) women. Here are the trends she uses:

The normative survey question she relies on is about whether it’s OK for a woman to work “if her husband can support her.” The S-shape of labor force participation rates is supposed to be consistent with the cultural transmission theory (rather than being caused by, for example, anemic work-family policy, anti-feminist backlash, or hollow anti-discrimination enforcement).

But I don’t see anything in the paper about increasing non-marriage (now about twice as common as in 1960), or about labor force participation rates for single women. Shouldn’t economists be concerned about that kind of selection issue? In fact, labor force participation rates for single women have stalled, too, as my figure shows:

I don’t think attitudes toward married women’s work — or anything about marriage alone — are going bear the burden of explaining two decades of stalled progress into the labor force for both single and married women. I’m happy to have cultural explanations as part of the mix here, but I don’t think this one will do it.