Time check

When, if ever, did this ship get turned around?

After I commented on Time‘s story about young women earning more than young men, Editor-At-Large Belinda Luscombe was good enough to drop in with her comment on my comment. Our exchange led me back to look at the story, and that led me to tinker with the data. And so here we are.

I have three principal beefs with the Time story. My take and some numbers follow.

1. Juxtaposing the overall gender gap in pay with the female advantage among a small slice of the population, and titling the story “At Last, Women on Top.”

Because of the fixation on single numbers, we’re doomed to reflect everything against “the” gender gap, which shows that, as of 2009, full-time working women earn an average of 80.2% of men’s earnings. So Time wrote:

The fact that the average American working woman earns only about 8o% of what the average American working man earns has been something of a festering sore for at least half the population for several decades. … But now there’s evidence that the ship may finally be turning around…

The splashy that some women out-earn men. But the evidence that followed was not about all full-time working men and women, but rather only unmarried workers in the age range 22-30 who have no children and live in large cities — they out-earn similar men by 8%. That could be important. But what does it have to do with “the” gender gap? Even for these women, as Heather Boushey pointed out, what about job segregation and promotions (not to mention child and marriage effects, which really matter).

2. Conjuring a trend from one point in time compared with a presumption about the past.

The story did not reveal a trend, but rather a snapshot of a trend presumed to be in progress. Given what we know about lagging progress toward gender equality — which might or might not be unstuck by the mancession — we should not assume that today’s fact is part of a wave of continuous progress in the direction of equality.

3. Relying on an unpublished analysis by a marketing firm without questioning its assumptions or conclusions.

In this day and age, and with all the corruption going on, who’s got time or patience for peer review? I am definitely not saying that you need a PhD in a relevant social science and your own blog to have something valuable to say on this subject. But this is a well-studied issue, and lots of experts are available to consult, interpret, or critique a splashy new finding — such as those who recently commented for a similar article in the New York Times, or experts on discrimination issues.

Peer review may be a boring system for ensuring conformity to arcane standards while protecting the privileged status of pampered intellectual elites. But those standards encourage researchers to follow common practices and reveal the details of their work, which makes it possible for others to replicate and explore it.

My take

But is the finding important? Time‘s expert attributed this subgroup of women’s advantage “overwhelmingly to one factor: education.” The education explanation was persuasive to Boushey, who wrote: “That’s what the Reach Advisors study shows—that because there are more young women with college degrees, women now outearn young men.”

In that case, education should have been considered in the analysis. I do that a little now. First, women started receiving a majority of four-year college degrees 30 years ago, and female college graduates have been outnumbering men in the 25-34-year-old age group for 20 years:

Source: My graph from Census and Department of Education data.

So, if it’s just more women having college degrees (having exempted all those with messy family issues holding them back), then we might expect that the gender-gap “ship” actually started turning around a while ago.

When I complained to Belinda Luscombe that “there is nothing in the story to show that this narrow subsection of workers hasn’t always shown a female advantage,” she replied: “are you really suggesting that it was ever thus and young women have always out earned men?”

I never say “always” (though I always penalize students who do). But look at this:

Source: My analysis of data from the 1990 and 2000 Decennial Censuses, and the 2008 American Community Survey, provided by IPUMS.

Sure enough, the median earnings for men and women in this odd, unrepresentative slice of the population have been basically equal since at least 1990. So, maybe Time should have done this story 20 years ago.

I can also provide a little support for the education hypothesis. If you take those same samples, but throw out the people who didn’t finish at least 4 years of college, the graph looks like this:

Source: My analysis of data from the 1990 and 2000 Decennial Censuses, and the 2008 American Community Survey, provided by IPUMS.

So, if the female advantage Time found really results from women having more education, the bad news is that the gap among those with college education hasn’t closed much (which could have to do with the stall in gender integration by field of study).

Because the overall gender-gap among college-graduates aged 25-34 is about 81%, these women are doing better relative to men than average college graduates. Why? In the end, I don’t believe the ship has turned around. The main reason is probably the selection of men and women who’ve never married and have no children — the “maternal wall,” as Joan Williams calls it. Is the plan for equality to “die childless at thirty“?

But I don’t much stock in the premise (can’t you tell?). If you’ll pardon me while I put on my demographer’s hat, I am especially concerned with the changing nature of who is in this odd group. People getting advanced degrees — increasingly common over recent decades — are mostly not working full-time till they graduate, so many of them are excluded from this under-30 group. And the age at marriage has increased. And marriage is more common among women than men at young ages. And what about race/ethnicity? Marriage is especially uncommon among young Black women, and young Black men have relatively low average wages (which drove the advantage for women in the NYC study).

And so on. So I don’t want to go too far down the road of explaining this (as if I haven’t already). Because this approach is wrong. We shouldn’t derive an unusual pair of comparison groups and then try to explain the comparison by referring to other facts not in evidence. If you really want to see the relative effects of education, age, marital status, children, race/ethnicity, and so on, you need to do it right.

10 thoughts on “Time check

  1. Ooh this discussion is fun! Little factcheck though. I’m an editor at large. haven’t been a senior editor for three years. (Don’t believe everything you read on the google.)


  2. This is probably indicative of the difference between journalists and academics. We report snapshots. We write every day, so we often write about little incremental changes or interesting new tidbits. That’s what Chung’s findings are. They need to be put in context for non-experts, hence the reference to the gender gap. ANd they need to pique people’s interest, so they get a snazzy headline, one that does not tell always the whole story. Academics (a) produce work more far-reaching in scope and they do it less often (b) have an informed readership (c) have a readership that the writer can assume is interested. (And (d) have a readership that’s tiny, but let’s not pile on…)
    So you know, what you lose on the swings you gain on the roundabout &c.


  3. Professor Cohen:
    We thought it might be worth jumping in as the firm that conducted this analysis. We concur with a number of your points, and respectfully should clarify a key issue.

    First off, it’s probably important to understand what this income dynamics analysis was…and what it wasn’t.

    We’re a strategy and market research firm focused on emerging shifts in the consumer landscape. The inquiry was triggered by some client work on a community development project a few years ago where we noticed that single women were buying 40% of the homes, a dramatic shift from past patterns. Since then, we’ve been finding similar patterns in other industries and geographic areas. Researching the drivers behind that shift triggered a multi-year research project digging into whether there’s anything different about a generation of young women that’s about 1.5 times more likely to graduate from college than young men…and income dynamics were an obvious part of the inquiry since that tends to drive consumer spending patterns.

    It’s critical to note that this was never designed as a study to prove or disprove wage discrimination in the workplace. Unfortunately, we’ve seen a number of interests on all sides of the spectrum try to take or refute the data for that purpose, when there are far better inquiries on that specific issue from various academics and advocacy organizations as you point out. For us, this is a much simpler matter of trying to determine how various shifts might impact consumer markets.

    But this leads to my main issues with this post, which we hope to respectfully convey. We actually concur that if the point was to prove or disprove wage discrimination or job segregation, there are other research pathways to accomplish that objective. Given the nature of our work, this research simply wasn’t designed for that purpose.

    At the same time, all the academic work on related issues that we reviewed wasn’t able to help us assess potential impact on consumer markets either, simply because those examinations weren’t designed for our needs. But I’m not going to assert that those approaches were wrong. We found significant value in many of them, but not necessarily for what we needed to find out. (As an aside, my wife is a professor who recently released some research that’s generated attention in academic circles on the history of gender discrimination in the financial arena, and while our research approaches are rather different, I’m not going to dare say that difference means hers is wrong!)

    A related issue we have is the assertion that “this approach is wrong” since it’s an “odd, unrepresentative slice of the population.” If it were attempting to represent the entire populace, it would be wrong, but instead, it’s simply the specific cohort that we found most interesting from the perspective of a comprehensive income dynamics analysis churning through a ridiculous amount of American Community Survey microdata (and the Census data that preceded the ACS). At least from a consumer market perspective, it’s not an insignificant slice when the number of single women in their 20s increased dramatically in the past decade. So with this combination of issues coming together for this slice of the population, their impact is starting to get notice…at least in the consumer marketplace.

    As many of the academics and some of the interest groups concluded when considering this data and implications, many of us agree that it’s a primarily driven by educational gains…there are far more young women than young men with the educational credentials to enter today’s knowledge-based work force. But yes, there’s a ton of stuff underneath.

    Regarding your data on men with BAs earning more than women with BAs, that’s a pretty good example where the data is generally accurate…but the real devil is in the details. To one of the points that you mention, there’s certainly the question about job segregation. After pulling data on 2,000 communities across the U.S. and appending various demographic data to those communities, and when comparing means versus medians, it certainly became clear to us when looking on a more granular level that there are some communities dominated by industries (e.g., software and finance) where the overwhelming majority of top earners remain male. Likewise, the minority issue that you allude to is another rather interesting topic that was evident in the analysis. But job segregation and social stratification are areas that other experts are far better suited to discuss with other research more relevant to those topics.

    So onto the question of what happens as this increasingly well-educated cohort moves through their next life stages? It’s a huge wildcard worth watching, and we’ll certainly continue to track the data over time to see how that informs what happens in the consumer market, as I’m sure others will examine the data for issues such as job segregation or wage discrimination.

    We look forward to crossing paths at some point down the road since you examine some issues that we like to track, although we hope that you don’t find it off-putting that even private-sector researchers sometimes read academic work as well 🙂

    As for Belinda Luscombe’s TIME article, we accept that her job is not to crank out copy to serve the needs of the academic community or the consumer marketing community. We suspect that her objective is to prompt thinking and consideration of interesting issues among as many people as possible. Based on this dialog, maybe she nailed it in that respect?


  4. James — your points are well taken. The Time article did not say what the original purpose of the analysis was. So I do not have enough information to say you did what you did “wrong.” What was wrong was using those results to draw some of the conclusions that people did — enabled if not actually provoked, I think (and Belinda I know disagrees), by the headline and framing of the article.

    It is an interesting problem in research on inequality, maybe especially in sociology, that we focus on the jobs people have and their earnings, more than other aspects of their lives that may be more directly indicative of their happiness, well-being, or success — which we assume follow from income. This has improved with the introduction of more studies of health outcomes, for example. But we should study consumption patterns more as matters of inequality. That we don’t is only partly a function of available data (I’ve used the Consumer Expenditure Survey a little, and it’s not pretty). Anyway, I’d like to see more of it.

    Thanks for taking the time to respond.


Comments welcome (may be moderated)

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

WordPress.com Logo

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

Twitter picture

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

Facebook photo

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

Connecting to %s