Women's happiness and pundits' accuracy

« previous post | next post »

Following up on yesterday's discussion of Ross Douthat's column on women's liberation and women's unhappiness, I thought that some people might find useful to look at the underlying data in a more quantitative way. So I downloaded the whole General Social Survey dataset from here, and pulled out the columns corresponding to the variables "year", "sex", and "happy": some  summaries are below, and if you want to do your own analysis of this subset of the data, a .csv file is here.

(My .csv file includes "age", and also "hapmar", which answers the question "Taking things all together, how would you describe your marriage?" There are 5,359 other variables in the full GSS dataset — see the documentation here.)

Data is available for 27 of the 37 years between 1972 and 2008 (1972, 1973, 1974, 1975, 1976, 1977, 1978, 1980, 1982, 1983, 1984, 1985, 1986, 1987, 1988, 1989, 1990, 1991, 1993, 1994, 1996, 1998, 2000, 2002, 2004, 2006, 2008). If we sum up all the responses across years, we get these proportions of answers to the question "Taken all together, how would you say things are these days — would you say that you are very happy, pretty happy, or not too happy?"

Very happy Pretty happy Not too happy
Male 31.2% 56.7% 12.1%
Female 32.4% 55.1% 12.5%

If instead we take proportions year-by-year and then average the proportions, we get a slightly different result (since the number of respondents differs from year to year):

Very happy Pretty happy Not too happy
Male 31.4% 56.6% 12.0%
Female 32.8% 54.8% 12.4%

Either way, there are a few more women, proportionally, who report their happiness as being in the extreme categories, and a few more men who assign themselves to the middle category. But either way, the differences are tiny — we certainly shouldn't describe this by saying that "men are happier than women".

On the other hand, what Ross Douthat actually said was, "In postfeminist America, men are happier than women". So maybe we should just look at the more recent data, and compare it to the earliest available data.

In the responses for 1972, 1973, and 1974, the overall proportions were:

Very happy Pretty happy Not too happy
Male 31.9% 53.0% 15.1%
Female 37.0% 49.4% 13.6%

In the responses for 2004, 2006, and 2008, the proportions were:

Very happy Pretty happy Not too happy
Male 29.8% 56.1% 14.0%
Female 31.2% 54.9% 13.9%

The best way to describe this, I think, would be to say something like:

In the early 70s, women self-reported their happiness at levels somewhat higher than men did. Specifically, 5.1% more of the women reported themselves "Very happy", while 1.5% fewer reported themselves "Not too happy".

30-odd years later, in the mid 00s, women's self-reported happiness was closer to men's, though it was still slightly higher. 1.4% more of the women reported themselves "Very happy", while 0.1% fewer reported themselves "Not too happy".

As a description of these facts, Douthat's assertion that "In postfeminist America, men are happier than women" is, at best, bizarrely off base.

The Stevenson and Wolfers paper fits a statistical model, which validates the impression also visible in a plot of the data that women's reported happiness has declined slightly over the past few decades, both absolutely and relative to men's. Stevenson and Wolfers argue that these changes are unexpected and require some explanation: women's economic situation has improved; economists stereotypically believe that happiness ought to be a measure of one's economic situation; but women's self-reported happiness has gone down somewhat; so what gives?

You can read their paper to evaluate the various alternative explanations (and other sources of data) that they consider. My point here is simply that the effect is not a very big one, at least from the perspective of a non-economist; and in any case, it's nonsensical to describe it as a situation where women as a group are "unhappy" while men as a group are "happy"; or where  women are (almost) uniformly unhappier than their counterparts were four decades ago.

You may object that Stevenson and Wolfers don't say this; and Douthat doesn't exactly do so either, depending on how you interpret generic phrases like "men are happier than women". But go to the readers' comments on the NYT site, and you'll see reactions like these (each paragraph is a separate comment):

I haven't read Stevenson & Wolfer's paper (& I can't because I'd have to pay for it, & that would make me even less happy), but I'd be surprised if they argued, as Douthat seems to suggest, that all women are less happy than were all women 40 years ago. Common sense says that can't be true.

Most of the women I know are highly educated, affluent, and socially involved. Also, most are still physically attractive, for whatever that is worth. But, as a group, they are not only miserable, but they make people around them miserable.

Men were unhappier before. Women insisted that men "had it all". Men knew better but women would never listen to men. Heaven forbid. Men were seduced into wars, farm work, factory labor, being the protectors and earlier deaths by being given the status of "being a man". Women forgot all the real power they had by getting half of humanity to carry this burden for them. So, now they're miserable, eh? Justice, I'd say.

What no one wants to admit is that women's "liberation" was a disaster, ironically making the sexes confused and frustrated. I wish I had been born 10 years earlier; I missed out. I never understood what was so bad about staying home. Clean and run errands in the morning, and watch soap operas and cook in the afternoon. Sounds great to me.

Until we stop pretending that women are the same as men, women's unhappiness will continue to deepen, and we will see an unceasing parade of grim, lemon-bitten mugs as we stroll our nasty, egalitarian streets.

There never was a mystery to all this. Women wanted to be liberated and they got what they wanted, but that didn't make them happy. They were unhappy with what they got and proceeded to make the men around them unhappy.

Unfortunately your piece, as has been the custom in this country for 2 generations, does not ask why men are happier and instead focuses entirely on women.

This is exactly what happened the last time that the Stevenson-Wolfers work was touted in the NYT: many if not most readers took generic statements about "men" and "women" to characterize general properties of the groups, or at least of most members of the groups, whereas the effect under discussion is a shift of a few percentage points, mostly accomplished by shifting the opinions of around 5 women in a hundred from "very happy" to "pretty happy".

It seems to me that researchers should do more to counteract this kind of predictable distortion when their work that is likely to be widely discussed. And they bear an especially heavy responsibility when their work deals with emotional issues like race or sex, and when they work hard to promote it in the media.

As for the distorters, represented in this case by Mr. Douthat, I'm at a loss.

I remain hopeful that the web's fact-checking capability will gradually lower the level of bullshit in the mass media — but the effect is slow at best.  On the whole, I'm less worried about women's failure to become happier than about columnists' failure to become smarter.

[Note that I've put a .pdf copy of the Stevenson/Wolfers paper on my web site, where you can read it for free. I'm sorry for violating their copyright, and I'll take it down if someone insists who has standing to do so; but it seems to me that when a piece of work on important public issues is promoted heavily over a period of years — and especially when it's based on publically-funded data — the public ought to have open access to it.]


  1. parvomagnus said,

    May 27, 2009 @ 9:58 pm

    The first NYT comment posted seems to be showing just this type of skepticism toward blanket statements about women.

    [(myl) Indeed. But it also shows that the commenter understood Douthat to be making such blanket statements, which was why I included that comment in the list. ]

  2. Tlönista said,

    May 28, 2009 @ 12:41 am

    Thanks for uploading the paper—despite the copyright violation. Whenever the press seizes on a bit of research, almost invariably the original paper is inaccessible to everyone but academics.

  3. jo lumley said,

    May 28, 2009 @ 6:32 am

    First, thank you for such an interesting post.

    Secondly, I just thought I would mention, for any statistics dunces like me who are trying to play along at home, that it looks like for the variable "Sex" in the dataset, 1=male, 2=female, for "happiness", 0, 8 and 9 are missing values and 1, 2, and 3 are "not too happy", "pretty happy" and "very happy" respectively. When I assigned values in this way, I was able to arrive at a table identical to the very first one in this post.

    [(myl) Thanks for clarifying this, and apologies for not giving the decoder key myself. I believe (from the information given here) that "0" means that the question wasn't asked, and "8" means that the subject declined to give one of the available answers. I'm not sure what "9" means.

    The overall numbers of responses in each category are:

    1 15,399
    2 26,974
    3 5,945
    0 4,383
    8 18
    9 324

    So whatever "8" and "9" exactly mean doesn't matter very much; and if "0" means that the subjects in question didn't get that particular question (out of the many, many questions in this survey), then its distribution by sex and year can be ignored. ]

  4. Bill Walderman said,

    May 28, 2009 @ 8:40 am

    Douthat's concluding statement is the most bizarre and outrageous distortion of the Stevenson/Wolfer paper in his entire column:

    ". . . ours is a kinder, gentler, more forgiving country than it was 40 years ago. But for half the public, it’s an unhappier country as well."

    I'm not exactly sure how to interpret that, but if "half the public" refers to American women, he seems to be saying that all American women are unhappier today than all American women were 40 years ago! And if "half the public" doesn't refer to the female segment of the population, then exactly what point is he trying to make in the rest of the column?

  5. Mark P said,

    May 28, 2009 @ 8:44 am

    I think the problem is not just drawing sweeping and mostly invalid conclusions from this type of data, but the casting of the conclusions in a particular way. Douthat says that women are unhappier in post-feminist America. If you think women are unhappier, why pick feminism as the reason? Why not say that women are unhappier in post-Vietnam America? Or women have tended to become unhappier as Honda automobile sales have increased? I'm sure there is an almost unlimited number of phenomena that the happiness scale could be correlated to. If a columnist (or a researcher) picks feminism in the absence of evidence to support it, I think it says more about the columnist than it does about the reasons for women's happiness.

    If you believe, as you said economists apparently do, that happiness depends on economic conditions, and if you think women's economic conditions have changed as a result of feminism and nothing else, I suppose you might think that womens' happiness has changed as a result of feminism. But I still say the more important result is the general decline in the "very happy" category for both men and women. Might that indicate causes that would apply to both men and women instead of to women alone?

    And, by the way, I found a pdf of the paper yesterday at upenn.edu:


  6. Chris said,

    May 28, 2009 @ 9:15 am

    Stevenson and Wolfers argue that these changes are unexpected and require some explanation: women's economic situation has improved; economists believe that happiness is just a measure of one's economic situation; but women's self-reported happiness has gone down somewhat; so what gives?

    It seems to me that the obvious interpretation is that the economists (which ones believe that? surely not *all* of them? is there a little more essentialism at work here?) are refuted. Isn't there happiness by SES data that shows a much weaker effect than might naively be expected?

    [(myl) In fairness to economists in general, and Stevenson and Wolfers in particular, there's a large literature on the relationship (or lack thereof) of happiness measures to absolute, relative, or expected material circumstances; and S&W discuss at length what they call "a richer consideration of the psychology behind happiness". One of the motivations of their research, as I understand it, is the neoclassical view that economic agents are rational maximizers of utility, along with the observation that women have done better, over the past few decades, on various measures related to this hypothetical utility, so that women might be expected to self-report as happier. The fact that they (apparently) don't is then a fact worthy of investigation; hence S&W's paper.

    (I suppose that this must be somehow related to the apparently growing field of "happiness economics", though I don't know enough about the discipline-internal developments to place it with any confidence. ]

  7. Tom said,

    May 28, 2009 @ 10:12 am

    I'm hard-pressed to come up with any women's historian who would say that the feminist cause has advanced since the mid-1970s. (The majority of evidence of women's "progress" in the first paragraph of the study is dubious at best.) Moreover most political historians consider the 1980s to be the start of a *conservative* turn in American politics. A better explanation for the decrease in women's happiness is that the 1970s represents the high-water mark of the modern feminist movement and that women's rights have steadily eroded since then.

  8. J. W. Brewer said,

    May 28, 2009 @ 3:42 pm

    I wonder if the problem here is specific to the fuzziness of "happiness" and its quantification rather than with English semantics and a comparatively innumerate population, because, for example, I do not think average speakers of English with no formal training in statistics have any trouble understanding sentences like "Men are taller than women" as being a claim about statistical distributions, with the bell curves overlapping — rather than a claim that the shortest man is taller than the tallest woman. Some might pedantically say that the sentence *should* be "Men are, on average, taller than women," but they have no problem understanding it correctly without that express qualifying language.

    Take another example, where people don't have the clear evidence of their own eyes, and which involves the politically loaded intersection of race, sex, and educational attainment: The ETS asserts without any qualifying language that "Blacks are the only group where females scored higher than males on the SAT I Verbal." (www.ets.org/Media/Research/pdf/PICGENDER.pdf at the page numbered 19, which is page 21 of the pdf.) Is it really likely that this will be misunderstood by the average reader as something more sweeping and absolute than a claim about overlapping statistical distributions? (As it happens, the difference in "average" score between black males and black females — unclear at quick read whether mean or median — is only a "few points," which from squinting at the graph probably means less than 5 on a 200-800 scale.)

    [(myl) Yes, I think that many people do misinterpret such statements. Anecdotal evidence is available in the comments on various sites to stories about the Stevenson and Wolfers paper. ]

  9. Bill Walderman said,

    May 28, 2009 @ 3:50 pm

    "But it also shows that the commenter understood Douthat to be making such blanket statements, which was why I included that comment in the list."

    But doesn't he actually make just such a blanket statement in the last sentence of his column? See my previous comment.

    [(myl) His last sentence ("But for half the public, it’s an unhappier country as well") is subject to the same sort of ambiguity (or vagueness?) as "women are unhappier now than they were in 1972", isn't it? On one way of reading such a sentence, it's true if the distribution of happiness levels of today's women is even a tiny bit shifted towards lower values. But most people's interpretation will be that this tells us something about typical women, or most women, or perhaps nearly all women. ]

  10. MC said,

    May 29, 2009 @ 12:39 pm

    It may not even be as much as a 5% change. I used a cut point of 1984/1985, and female non-smokers (smoking can affect happiness) go from 37.5% very happy in 1972-1984 to 35.0% very happy 1985-2008. Male non-smokers go from 35% to 35.7% . So, I do see a bit of a trend, but it's very very small and the differences between men and women within each year don't appear to be significant. Most of the difference to me seems to be that women took up smoking in this period (which they did until 1985).

    (Note also that I didn't weight for the complex survey design, so estimates aren't really reliable).

  11. MC said,

    May 29, 2009 @ 12:55 pm

    never mind. The periods are actually 1972-1984/1985-1994 for the smoking questions.

  12. Steve Roth said,

    June 1, 2009 @ 9:40 am

    If there is any validity to the statement regarding women's happiness, it is perhaps most easily attributed to the fact that women had to psend more time at work. Women's labor force participation jumped 50% from the 70s to the 00s–from 40% of women working to 60%.

    Thanks for running the numbers.

  13. digg » Blog Archive » Can Health Care Lead to Happiness? said,

    June 26, 2009 @ 11:21 pm

    […] "Padadox of Declining Female Happiness" at 100% face value (Another professor at UPenn offers a critique of the language of the report and particularly the way it was picked up by Ross Douthat here). As […]

  14. Anna Kincaid said,

    September 23, 2009 @ 10:04 pm

    The original paper by Betsey Stevenson and Justin Wolfers is free at:

  15. Happy With Your Bikini Wax? | GLORIA FELDT said,

    September 27, 2010 @ 2:28 pm

    […] of power over women. But only women are the subject of media speculation. While other experts have debunked Buckingham’s metrics as well as his conclusions, even a novice could spy a trope that has relentlessly fostered women’s […]

RSS feed for comments on this post