Thursday, October 15, 2009

Declining States of Happiness

Read this post on the Freakonomics blog and had to comment upon it. The post refers to a paper noting the decline in the reported levels of happiness of American women over the past few decades. Apparently this paper was recently attacked in an op-ed piece by a journalist who questioned its methodology. Here one of the researchers defends his work.

The study itself may be perfectly reasonable--I have not read it. The comments about it annoyed me, as did the complete dismissal of the criticisms leveled by the journalist, some of which seemed to be petty but others of which seemed to be practical enough.

I don't know if my comment will pass blog muster or not. Here's what I wrote:

In regards to #33, the original poster is saying several things about the flawed nature of self-reported happiness surveys as sources of hard empirical data. The political screed you seem to have provided on your own.

(a) People (and societies) don’t all define happiness in the same way. Individual and social expectations play a huge role in determining happiness. Over time, it seems reasonable to argue that these expectations can change within a society. (After all, the survey is arguing that levels of happiness can change–why not definitions of happiness as well?)

(b) When revealing personal information on anonymous surveys, people may be motivated to misrepresent themselves. If they feel they SHOULD be answering in a particular fashion, they have a higher incentive not to give an honest answer that paints them in a negative light.

(c) Surveys asking people to self-report their degree of happiness have the same challenges as any survey asking about someone’s emotional state, as opposed to something more concrete like their height or salary. Even if everyone defined happiness in the same way and answered honestly, what happens if you catch a survey respondent on a bad or good day?

These issues make me very suspicious that a statistical analysis of such self-reported happiness data can provide results that in and of themselves explain much of anything about the phenomena they observe. For example, what if contemporary women continue the pattern of OVER-reporting happiness as has been proposed for the earlier generation of respondents? The gap would actually be even greater than proposed.

What such a study can do is raise important questions. Have womens’ EXPECTATIONS for their personal lives and their social roles changed in the past few decades, and how? Or are the EXPERIENCES of modern women contributing to a decline in happiness? Perhaps both? Do men and women in our society bring fundamentally different perspectives to the question of happiness, or do they define it in the same way? Those are difficult and interesting questions, but this study is just a stepping stone toward them. [end comment]

Now, a statistician may argue that they have mathemathical methods allowing them to account for any outlying deviations in credible responses to the happiness survey.

I say bullshit to that. The entire enterprise is flawed because it takes one incredibly vague, unconfirmed data point (How happy are you?) and extrapolates from there.

Don't believe me? Imagine asking a bunch of people how nice they are and then trying to make grand assumptions about the shift in compassion in American society. Would that hold as much credibility as asking people their views of capital punishment or examining hard data on how much they contributed to charities?

No comments: