We recently discussed two claims made by Arthur Brooks in the New York Times regarding political affiliation and happiness. In both cases, Brooks based his claims on General Social Survey data and we disputed him.
In 2012, Brooks wrote:
People at the extremes are happier than political moderates. . . . none, it seems, are happier than the Tea Partiers . . .
But here’s what the data show:
Sociologist Jay Livingston, who sent us this graph, writes:
The GSS does not offer “bitter” or “Tea Party” as choices, but extreme conservatives are nearly three times as likely as others to be “not too happy.”
Political junkies might be interested to learn that conservative women are particularly blissful: about 40 percent say they are very happy. That makes them slightly happier than conservative men and significantly happier than liberal women.
Again, here’s the GSS graph sent to us by Livingston:
I wouldn’t call conservative women “particularly blissful.” They’re no more likely to be happy than extremely liberal women, and they (the conservative women) are much more likely to describe themselves as “not too happy.” Indeed, one could with just as much justification describe conservative women as “particularly unhappy.”
[Note: The two graphs above use nearly-opposite color schemes! Don’t blame me, blame the default software that Livingston was using. Anyway, just keep that in mind when comparing the plots.]
When I blogged on all this, I speculated that Brooks made the mistake of working with the cumulative General Social Survey file, which aggregates the survey responses since 1972. As Livingston points out, it is reasonable to believe (and is borne out by the data) that the unhappiness level of self-described extreme conservatives has increased a lot since Obama’s election in 2008. And this is no small thing, given the tea party phenomenon that has been associated with angry conservatives.
But it seems that there is a different explanation for what happened, an explanation I had not thought of, as Brooks’s collaborator Andrew Quinn explains in a blog post at the American Enterprise Institute. Quinn writes:
Brooks doesn’t discuss the narrow subgroups such as “extremely liberal,” “slightly liberal,” “extremely conservative,” and so on. He touches on four broader categories: conservative women, conservative men, liberal women, and liberal men.
We chose to zoom out and take this wider view for several reasons. One is statistical: in a single year of survey data, the sample sizes for some of the niche subgroups are quite small. Aggregating them into fewer and larger categories makes broad trends more salient. . . .
Given this, Dr. Brooks and I synthesized the GSS responses into four categories:
All men who identify as liberal
All men who identify as conservative
All women who identify as liberal
All women who identify as conservative
We then produced an aggregate “very happy” score for each group by constructing a simple weighted average of its subgroups.
Quinn goes to the trouble of performing those calculations for the 2012 GSS alone and for the combined 2008-2012 GSS, and, as he points out, “the groups’ relative rankings are identical.” The conservative women (combining “slightly conservative,” “conservative,” and “extremely conservative”) have the highest rate of describing themselves as very happy.
That’s all good. Looking at all the data, though–in particular, that that big fat red bar at the far right of the second graph above (the “not too happy” extremely Republican women), it’s hard for me to buy the summary statement that conservative women are “particularly blissful.” Particularly likely to say they are either “very happy” or “not too happy”–particularly extreme in their happiness statements!–but that’s a bit different.
The real message, though, is that survey responses can show a lot of different things. I think this discussion provides a strong argument for presenting a graph rather than trying to pull out patterns using sentences and paragraphs. A graph shows the big patterns, it also shows the complexity of the situation, in particular, a group such as conservative women who are more likely to say they are “very happy” and more likely to say they are “not too happy.” Instead of arguing over whether it’s appropriate to describe such a group as “particularly blissful,” we can just show people the data. I appreciate Quinn’s openness in sharing his analysis, and I hope that, in the future, media outlets such as the Times will make it easier for essayists and reporters to include graphs where appropriate. We’re used to thinking of graphs as a big production, but a well-designed graph can be formatted to take up not much more space than a paragraph of text.
P.S. I still see no justification for Brooks’s earlier claim that “People at the extremes are happier than political moderates. . . . none, it seems, are happier than the Tea Partiers . . .” To the extent that the GSS is answering these questions, I’d think that “extremely conservative” would be the best match for “Tea Partier.” And, in any case, “people at the extremes” would have to refer to “extremely liberal” and “extremely conservative,” and as we can see from the top graph above, those groups have about the same level of “very happy” as the general population but (on the conservative side) are much more likely than others to be “not to happy.” So I still don’t know whassup with that.