The happiness gap is back

« previous post | next post »

According to Ross Douthat's latest column for the NYT, "Liberated and Unhappy", 5/25/2009:

[A]ll the achievements of the feminist era may have delivered women to greater unhappiness. In the 1960s, when Betty Friedan diagnosed her fellow wives and daughters as the victims of “the problem with no name,” American women reported themselves happier, on average, than did men. Today, that gender gap has reversed. Male happiness has inched up, and female happiness has dropped. In postfeminist America, men are happier than women. [emphasis added]

His warrant for the generic statement that "In postfeminist American, men are happier than women" is Betsey Stevenson and Justin Wolfers, "The paradox of declining female happiness", NBER Working Paper No. 14969, Issued in May 2009. Though this paper was issued by the National Bureau of Economic Research only recently, earlier versions have been around for a while, and were featured in the NYT's Freakonomics column back in the fall of 2007, and discussed here then.

I won't recapitulate that whole discussion; but I will reproduce one of the graphs (click on the image for a larger version):

And I'll ask a simple question: What fraction of graphically and statistically literate people think that the right way to describe the data summarized in that graph is "In postfeminist America, men are happier than women"?

[The y-axis is the proportion of the sample (of men or women) who gave a given answer.]

If you want to read the earlier discussion again, here are the links:

"The happiness gap and the rhetoric of statistics", 9/26/2007
"Gender-role resentment and Rorschach-blot news reports", 9/27/2007
"Why are economists so misleading?", 10/1/2007
"The gender happiness gap: statistical, practical and rhetorical significance", 10/4/2007

The second of those posts has a sample of lapidary web-forum comments. Among my favorites were the woman who commented on the Freakonomics column that "Men are dogs. Dogs are happy. Voila.", and the man who opined at that "Feminists made their bed, now they have to lie in it alone with their cats."

And the last-cited post has a list of other reactions, in both new and old media, to that 2007 orgy of pop Platonism. I guess we're due for another one, thanks to Mr. Douthat, who seems to have decided to follow David Brooks' example in crafting columns that turn small statistical differences into generic statements about groups, accompanied by meditations on the cultural and political implications.

[Update: more here.]


  1. Theo Vosse said,

    May 26, 2009 @ 8:01 am

    You asked what fraction of graphically and statistically literate people think that the right way to describe the data summarized in that graph is "In postfeminist America, men are happier than women"?

    I'm afraid the number might be higher than you think. Nowadays, even 8ms differences in reaction time studies get published, as long as they are (statistically) significant. The difference in the graph can be significant if the sample size is big enough. My point is that statistical significance is no substitute for relevance, interest, etc., as it is used in everyday conversation (sufficiently great or important to be worthy of attention), and I guess thats what you're referring to as well.

    In that case, I hope nobody in his right mind wil want to support that claim.

    [(myl) It's sometimes true that statistically significant differences in sample statistics are a meaningful basis for inference, even though they're small both in absolute terms and in relation to the within-group variance. Small differences in group averages might translate to worthwhile results in a public policy debate, or form part of a scientific argument whose conclusion is interesting. But as you say, the problem arises when a (true but complicated) statement like "the mean value of P for group A was greater than the mean value of P for group B, by 10% of the pooled standard deviation, and this result would occur by chance less than 1 in 20 times" gets translated into a (simple but false) statement like "A's are P-er than B's are". ]

  2. Tim Silverman said,

    May 26, 2009 @ 8:08 am

    There really seems to be nothing serious one can say about this nonsense, so I shall instead confine myself to the following query: If feminism has made these men so happy, why are they complaining about it?

    [(myl) Sympathy for women? ]

  3. Peter said,

    May 26, 2009 @ 9:08 am

    Quite aside from the discussion at hand, I have to say that it took me a while to try and determine what the graph you have posted is actually communicating; for some reason someone has removed the Y-axis label which indicates what the quantity is and gives units. Having looked at the original paper, I can reveal that the Y-axis is "Proportion of population", and thus it is a unit-less quantity. Hence the 'gap' can it seems be interpreted as the distance between the lower and upper series of the graph, for a given gender.

    [(myl) No, the "gap" is the difference between male happiness and female happiness, which would be the amount by which the upper blue-gray line (the proportion of men who say they are "very happy") is above the upper pink line (the proportion of women who say they are "very happy"); or the amount that the lower blue-gray line is below the lower pink line. (Sorry if I've misunderstood and that's what you meant.)

    But, you say, the upper blue-gray line *isn't* consistently above the upper pink line. In fact, in four of the last five years plotted, it's below it. Exactly, I respond.]

  4. Mark P said,

    May 26, 2009 @ 10:23 am

    What's with the category "not too happy"? Is it really desirable to be "too happy"?

    [(myl) Oh, come now — "not too X" is a standard (if somewhat informal) way of saying something like "rather worse than middling".]

  5. Tlönista said,

    May 26, 2009 @ 10:38 am

    While Douthat's effort is pretty feeble, using a study like this to prove anything about feminism would be insurmountably difficult. From a generic poll on happiness, how would you be able to tell whether it's feminism that makes the difference, rather than being in a recession / the shift in the urban/rural divide / having a war on / concurrent movements like those for civil rights, disability rights, or gay liberation?

    [(myl) Actually, if you read the Stevenson/Wolfers paper, you'll see that they bring in a wide range of data, from other aspects of the GSS and from other sources as well, to evaluate the contributions of various factors, including some of those that you list, to the trends that they identify. (Though it's not clear from Douthat's column that he's basing his thoughts on the finer details of their analysis.) The problem is not that the contribution of different factors to the effect can't be evaluated, the problem is that there's not much of an effect to start with. ]

    It's like the chart that jokingly compares global warming to piracy.

    Say Douthat really wanted to make a go of it, rather than grabbing at the nearest study to justify his pre-existing beliefs. It's probably better to examine our progress on particular things women's rights activists are pushing for, like gay marriage, parental leave, better health care, equal pay, and more humane policies for immigrants and sex workers. But that would be a terrible lot of work, more than you could demand of a humble NYT op-ed columnist who doesn't even know that feminism is still a going concern.

    (Also, like Peter said, the lack of labels on that Y-axis is pretty confusing.)

  6. Jan Schreuder said,

    May 26, 2009 @ 10:46 am

    Two David Brookses on the NYT staff? I am sorry Mark, but you will have to double your watch dog efforts. I rely on you so I can happily skip Brooks and Douthat.

  7. Bloix said,

    May 26, 2009 @ 10:50 am

    In 2008, just over 1% of the US adult population was in prison or jail – over 2.3 million people. Approximately 7% of them were women. That is, roughly 1.9% of adult men are incarcerated; and roughly 0.14% of adult women are incarcerated. Were inmates surveyed in these studies? And if not, how would the slight differences between men and women change if one assumes that a high percentage of inmates, regardless of gender, would respond that they "are not too happy"?

  8. Dan Lufkin said,

    May 26, 2009 @ 11:15 am

    I just finished reading Douthat's piece for myself and I think that he carries the argument much better than LL has given him credit for. He notes that the data's not overwhelming and that the ambiguity lends itself to a broad-brush treatment, which he then proceeds to dish up. He outlines the arguments for both the traditionalist and the feminist views, erects and knocks down a few straw men on both sides, and points to one aspect — balance of work and family — where both sides have a common interest.

    [(myl) The more nuanced discussion is nice, but it doesn't change the opening, which states baldly (and I submit, falsely) that "In postfeminist America, men are happier than women." ]

    I find it hard to tease out any of Douthat's pre-existing personal beliefs here. He's produced a workmanlike essay that brings an interesting observation to our attention, given it some context and invited us to think about it. I don't see any evidence that he doesn't know that feminism is still a going concern. He specifically says that American society doesn't make enough accommodation to the specific challenges of motherhood. The poor guy has just a few column-inches to make his case and I think he's done a good job of it.

    Second the motion on the Y-axis, though. But blame it on the page makeup editor's careless cropping, not on Douthat.

    [(myl) The graph came from the 2007 version of the Stevenson and Wolfers paper, and it never had a y-axis label, as far as I can recall — the scale was explained in the text.]

  9. Dan T. said,

    May 26, 2009 @ 11:16 am

    That graph only goes back to the 1970s (not the 1960s as in the alleged premise of a pre-feminist baseline), and the most recent data point there actually has a higher figure for women than men in the "very happy" graph.

  10. Mark P said,

    May 26, 2009 @ 11:17 am

    What struck me was the two categories shown, "Very happy" and "not too happy." I thought it odd that there were two (subjective, as they say) categories that limited response to those particular choices. Now I see in what I assume is the original paper that there was a third category, "pretty happy." I also see a lot more, including questions with more possible choice categories, but I don't have time right now to read the entire paper. I did note that it is an analysis of a long-term study not conducted by the authors, so the authors don't have much choice about the questions or responses.

    My only other comment (aside from what I just said) is that I wonder if other people dislike that kind of survey as much as I do (like the satisfaction surveys from businesses), and, consequently, give as little thought to their responses. I don't know whether it's worse to have too few possible responses or too many. Surely someone has studied that?

  11. Sili said,

    May 26, 2009 @ 11:23 am

    So putting men in prison makes them happier?

    For the record, I sleep alone with my cat.

  12. Benjamin Zimmer said,

    May 26, 2009 @ 11:26 am

    Stevenson and Wolfers were interviewed on the WNYC show "The Takeaway" last week, and they made no effort to counter the popular generic reading of their study. On the contrary, they seem to be running with it.

  13. Graham said,

    May 26, 2009 @ 12:12 pm

    Hmm. I'll bite then.

    If you posit a category of "Happy" that includes both "Very Happy" and the residual "Pretty Happy", I'd accept the conclusion of men being happier.

    Plot the Men/Women lines for "Happy" and you'd clearly see the Men line above the Women line for every single year from the late 80s to the present except for two or three years in the 2000s. It doesn't seem unreasonable to conclude that "In postfeminist America, men are happier than women" from that.

    [(myl) Suppose we draw randomly from "male" and "female" populations with the proportions of "happiness" you specify (i.e. the complement with respect to 1 of the GSS response proportions for the category "not too happy", for the years from 1980 to present). What proportion of the time do you expect the "male" to be "happy" while the "female" is not? How small does the difference need to be before it bothers you to make a generic statement about the groups as a whole?

    (Added later): For the past six years, as discussed here, women's proportion in the combined "Very happy" and "Pretty happy" categories was 86.1%, while men's was 85.9%. This is not a meaningful difference — though the number of resondents is so large that it's (probably) statistically significant — and I submit it would be at best misleading, especially in a popular article, to describe this result by saying that "women are happier than men".

    For the 20 years since 1980 for which the GSS has data, women's overall "happiness" proportion — your sum of "very happy" and "pretty happy" — is 87.5%, while men's proportion is 88.3%. Again, a very small difference, this time in favor of the males. And again, to describe this by saying that "men are happier than women" will predictably result in the preposterous misunderstandings evident in the public commentary on the Freakonomics article in 2007, and on Douthat's column more recently. ]

  14. Mark P said,

    May 26, 2009 @ 12:17 pm

    I forgot to answer the question. I think anyone who draws that conclusion is engaging in wishful thinking.

  15. Bloix said,

    May 26, 2009 @ 12:29 pm

    Sili- If everyone in prison is "not too happy," which seems like a reasonable assumption, and you don't survey them, then you are reporting the both the male and female "not too happy" results lower than they should be. But the skewing for men will be large, while the skewing for women will be very small. My question is. what happens if you add estimated results for prisoners to the survey results – that is, you increase the percentage of men who are "not too happy" by about 1.9% and increase the percentage of women who are "not too happy" by about 0.14. As I read the graph reproduced in the post, the difference in the "not too happy" lines between men and women is less than 1%, so if prisoners were included the men's line would be higher than the women's line.

  16. Bloix said,

    May 26, 2009 @ 12:35 pm

    And with respect to Mark P's comment: assuming feminism has had any impact at all, one of the first effects you would expect would be that women would become more honest about admitting their unhappiness. One of the feminist criticisms of patriarchal culture is that women must acquiesce in their own oppresson, so that being unhappy or dissatisfied is a sign that a woman is either morally bad ("ungrateful") or that she's mentally unstable. Therefore, you'd expect, if feminism has been successful, that there would be fewer unhappy women pretending to be happy.

    So, assuming there is a statistically significant decline in women's self-reported rates of unhappiness, one explanation that you'd have to eliminate before you conclude that women actually are more unhappy is that there's been an increased willingness among women to admit to unhappiness.

  17. Otter said,

    May 26, 2009 @ 12:36 pm

    I believe Tlönista's is the pithiest summary to date of Douthat's approach to editorial writing (shared by his predecessor Kristol, and many other conservative writers): grabbing at [anything] to justify his pre-existing beliefs. There's no reason conservatives shouldn't be represented on op-ed pages, but the paper of record ought to be able to find one who really wanted to make a go of it. Not to mention one who can read a graph.

  18. Ben said,

    May 26, 2009 @ 12:57 pm

    I think there's an unfortunate crossover between the term "statistically significant" and the everyday use of the word "significant" that drives a lot of the absurd over-generalization and over-extension we see in science reporting in the mainstream media. In everyday use "significant" DOES imply relevant, broad, and meaningful and people don't understand that "statistically significant" doesn't necessarily carry the same implications. So I propose we replace "statistically significant" with something else, ha.

  19. Ben said,

    May 26, 2009 @ 1:07 pm

    Also, I'd like to point out one thing in Douthat's defense. Here is the abstract for that paper:

    "By many objective measures the lives of women in the United States have improved over the past 35 years, yet we show that measures of subjective well-being indicate that women's happiness has declined both absolutely and relative to men. The paradox of women's declining relative well-being is found across various datasets, measures of subjective well-being, and is pervasive across demographic groups and industrialized countries. Relative declines in female happiness have eroded a gender gap in happiness in which women in the 1970s typically reported higher subjective well-being than did men. These declines have continued and a new gender gap is emerging — one with higher subjective well-being for men."

    Yes it would be nice if columnists read the actual paper and came to their own conclusions about what the research actually shows. But come on, that abstract reads like an overwrought PR-office press-release. What responsibility does the journal (in this case the National Bureau of Economic Research) have in writing its abstracts in a more scientific and balanced way that gives proper context?

  20. Rae said,

    May 26, 2009 @ 1:38 pm

    Is it possible (not having read the original article) that the small, albeit statistically significant, difference lies not in the proportion of women versus men who are "very happy" versus "not too happy," but in the proportion of women versus men who are willing to say that they are "not too happy?" Perhaps women's willingness to admit to less-than-total life satisfaction is what has increased, rather than our overall misery.

  21. Picky said,

    May 26, 2009 @ 1:45 pm

    I love this stuff about Taking all things together, how would you say things are these days full point no query. Well, you know, sort of OK, really, but nothing special, you know? …

    Opinion surveys are really impressively scientific, aren't they? Anyway …

    Don't the graphs show we're all trending towards unhappiness, m or f?

    Whoever thought feminism was about happiness?

    Who decided that we were all agreed we were now living post-feminist?

  22. Mossy said,

    May 26, 2009 @ 2:37 pm

    I feel compelled to mention (either darkly or flippantly; not sure which) Nora Ephron's analysis:
    The major concrete achievement of the women's movement of the 1970's was the Dutch treat.

  23. marie-lucie said,

    May 26, 2009 @ 2:42 pm

    I agree with Bloix: before the advent of feminism many North American women would have been ashamed to admit even in private that they were not blissfully happy being "a wife and mother", even if they were drowning their sorrows in alcohol or having trysts with door-to-door salesmen.

  24. PortlyDyke said,

    May 26, 2009 @ 2:47 pm

    I concur with Bloix's point. I'm 52, and the cultural entrainment I experienced as a young girl was all about "A lady doesn't complain. A lady doesn't show her temper. A lady suffers in silence." Look at the portrayals of mothers and wives in 50s and 60s TV — even when one of these icons was pissed off, she put on a gracious smile and played nice.

    One of the greatest liberations resulting from feminism for me was the realization that I had a right to complain.

  25. Bill Walderman said,

    May 26, 2009 @ 3:04 pm

    "the opening, which states baldly (and I submit, falsely) that 'In postfeminist America, men are happier than women.'"

    And his conclusion is an even worse distortion of the data, whatever they mean (if anything): " . . . 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."

  26. DW said,

    May 26, 2009 @ 3:17 pm

    =Ross Douthat (once more)

    [(myl) Thanks — I seem to have a problem with that 'o'.]

  27. Bloix said,

    May 26, 2009 @ 4:31 pm

    People who collect health statistics are conversant with the problems of over- or under-reporting based on a respondent's unwillingness to admit to departure from cultural norms. For example, women under-report the amount they smoke more than men do, and women from conservative cultural backgrounds (e.g, church-going Hispanic women) reliably under-report their smoking more than other women. You can determine how much a person actually smokes by doing a blood test for nicotine metabolites, so health statisticians collect all sorts of data from small samples of people that they can take blood from, and then create correction factors to be applied to much larger samples that are merely surveyed.

    But there's no blood test for happiness yet, so patterns of mis-reporting in conformance with cultural norms, although you'd expect to find them, must be difficult or impossible to measure.

  28. Just Sayin' said,

    May 26, 2009 @ 11:04 pm

    What's with the category "not too happy"? Is it really desirable to be "too happy"?

    [(myl) Oh, come now — "not too X" is a standard (if somewhat informal) way of saying something like "rather worse than middling".]

    I would say that is quite bounded by region and culture. "Fair to middling" is the equivalent of "not too happy" that I thought of and corresponds to "moderately good". And the expectation of exposure or sharing of emotional state as opposed to being closed about disappointment might have resulted in a change of meaning over time. I like those Likert 1-5 scales ….. "I am happy" Strongly disagree ….. etc

  29. Trond Engen said,

    May 27, 2009 @ 4:15 am

    Whenever I see one of these surveys of happiness I think that what they've really measured is semantic content — diachronically of the word 'happiness' or, in the case of Pan-European surveys, cross-linguistically of the translations of the concept. The difference between men's and women's language is just another parameter.

  30. Rick S said,

    May 27, 2009 @ 8:10 am

    Bloix, Rae, and PortlyDyke all remarked that the rise of feminism has increased women's willingness to admit unhappiness, but I think there is a related effect they all just barely missed: that of consciousness-raising. Women are not only more likely to admit unhappiness, they are more likely to be aware of it rather than accepting/discounting it as their natural lot in life. By giving women new opportunities and goals to shoot for, feminism naturally increases their dissatisfaction and frustration with their continued oppression.

    Not that I'm saying that that's what the study reveals. I doubt it reveals anything meaningful at all, given the fallacy of its "post-feminist" initial assumption.

  31. Rick S said,

    May 27, 2009 @ 8:13 am

    Sorry, I didn't mean the study was flawed, I meant Douthat's interpretation was.

  32. bianca steele said,

    May 27, 2009 @ 9:19 am

    Rick S's comment suggests a possible research topic: do women's magazines from the 1930s to the present show any trends w/r/t discussion of women's dis/satisfaction? Though it isn't obvious how to measure that: a story about a dissatisfied housewife in the 1950s might be read as "I'm glad that didn't happen to me," whereas in the 1970s a similar story was probably read as "look what patriarchal values did."

  33. Dan Lufkin said,

    May 27, 2009 @ 9:37 am

    For a book on whether happiness can be quantified, let me recommend Eric Weiner's The Geography of Bliss. He visited Iceland (#1) and Moldova (#40) and reports what he found on the hoof.

  34. Bloix said,

    May 27, 2009 @ 10:39 am

    My guess is that "not too happy" was carefully chosen in order to avoid "unhappy" or "not happy." It's asking a lot of someone to admit straight out to a stranger that they are "unhappy" – I expect the researchers decided that they would be more likely to get truthful answers if you used "not too happy" as a choice instead.

  35. Mark P said,

    May 27, 2009 @ 11:21 am

    Well, whatever the state of women's happiness, I think that the most significant feature of the plot is a general downward trend in "very happy" responses for both males and females.

    In my line of work I see plots like this all the time, for entirely different types of data from this study. If I saw a plot like this for the types of data I work with, my conclusion would be that for practical purposes, the lines are identical, even if there is a statistically significant difference. I would note the downward trend. I would ask about the rest of the population.

  36. Mark P said,

    May 27, 2009 @ 11:24 am

    I should note that if you know that there are only three categories, you can infer what's happening in the third category, but you have to go to the original paper to find out that there are three and only three categories.

  37. Tim Silverman said,

    May 27, 2009 @ 2:24 pm

    If we're speculating on hypothetical reasons for the pattern, we could equally well point to the the apparent spike in women's relative happiness that occurred in the early to mid seventies, just as feminism was getting going in a big way, and suggest that it was at that time that women were (depending on which women and which part of their lives we're talking about) either experiencing a historically unusual improvement in their position relative to men, or could reasonably hope to experience such improvement in the near future. Then subsequently, as the gains of feminism (and related changes) were realised and started to hit the usual combination of diminishing returns, acclimatisation, resistance to change, and unanticipated negative consquences, their happiness levels reverted to the more normal levels experienced by men over this period. All purely speculative, of course.

  38. Graham said,

    May 27, 2009 @ 3:11 pm

    On Bloix's point: Note that the incarceration rate in the USA has risen dramatically in the period under discussion (from 0.14 percent of the population in 1972 to 0.742 percent in 2005). Since the incarcerated are almost entirely male (93%), poor (and therefore unhappy), and unreachable by telephone pollsters, it seems likely that this would account for at least part of the gender gap evident in the survey (by essentially removing the unhappiest 1.1% of the males from the sampled population).

  39. Bloix said,

    May 27, 2009 @ 5:09 pm

    Actually the unhappiest 2% of males – as of 2008, the incarceration rate was over 1% of the adult population, meaning about 2% of the adult male population – and this study purports to measure self-reported happiness among adults. And as I read the graph, it more than accounts for the entire gender gap.

    BTW, I wouldn't say that the incarerated are likely to be unhappy because they're poor. I'd say they're likely to be unhappy because they're in prison. There may be the occasional serene soul who is reconciled to his state, but I expect that most people in US correctional facilities are "not too happy" for very good reasons.

  40. W. Kiernan said,

    June 3, 2009 @ 4:36 pm

    Nice graph. I got a laugh out of how regularly dips in the pink line (hey I laughed at the girly pink in the pink line too!) precede dips in the manly gray line. And that's a fact: if Mama's not happy, nobody's happy.

    By the way, why should we Americans be emburdened with pronouncing Douthat's name whatever damn foreign way he pronounces it, instead of the natural English-speaker's phonetic pronunciation "doubt that"?

RSS feed for comments on this post