More sex might increase happiness, but not if an economist tells you to

I couldn't let this one pass without comment. Quartz highlighted an article in the next issue of the Journal of Economic Behavior and Organization by George Loewenstein, Tamar Krishnamurti, Jessica Kopsic, and Daniel McDonald (all from Carnegie Mellon University), entitled "Does increased sexual frequency enhance happiness?" (gated, I don't see an ungated version anywhere).

The literature to date on the relationship between happiness and sexual activity has been purely correlational - no study has been able to demonstrate the causal links between sexual activity and happiness. Understanding the difference is important, because if we observe that happier people have more sex this could be caused by any of three things: (1) more sex makes people happier; (2) happier people tend to have more sex; or (3) some other variable both increases happiness and increases sexual activity (Loewenstein et al. suggest 'liking for sex' as one candidate variable). Alternatively, even if the correlation is observed maybe there is no relationship between sex and happiness, and the observed correlation arises because of random chance (although given the number of studies showing this correlation, this explanation seems the least plausible). So, teasing out the causal effects is important in terms of working out which of these alternative explanations is most correct.

Loewenstein et al. set out to do this using a field experiment:

Couples... were then stratified by age and sexual frequency before being randomized to one of two groups: the control group, who received no instructions on sexual frequency during the 90-day experimental period, and the treatment group, who were asked to double their baseline weekly frequency of sexual intercourse.

The main outcome variable was a 'positive mood scale', which was composed of a number of positive and negative emotion items from the PANAS (Positive and Negative Affect Schedule), combined into a single variable. The expectation is that, if more sex increases happiness then the treatment group should be happier at the end of the treatment period than the control group.

Loewenstein et al. found that the treatment group did increase their sexual frequency, but did not quite double frequency (sexual frequency increased by 40% on average). So far, so good. However, in terms of happiness:

...those induced by the experimental condition to have more sex displayed a lower mood during the course of the experiment than those in the control group. The point estimate of the negative impact of treatment condition on mean mood was 0.2 S.D.s.

So, you might conclude that if an economist tells you to have more sex, this makes you less happy. However, that conclusion would be based on incomplete analysis, and I have to raise one gripe with this paper. In further analysis, they go on to state:

Specification IV controls for years married, a key variable that was imbalanced between the two groups. With the addition of this control, the coefficient on treatment is still negative, but no longer significant at the 0.1 level.

If a variable is not statistically significant, then by definition you can't tell whether the 'true' coefficient value is positive or negative (or zero), and hence the sign of the coefficient is meaningless. So, rather than concluding that the treatment (more sex) made participants less happy, it is more correct to conclude that it had no significant effect. Thanks to my former colleague Bridget Daldy, who drummed that point into me over many years.

The overall conclusion then? More research needed. One last point: Loewenstein et al. have offered to provide all of the data from their research to anyone interested in investigating further. I don't have the time, but it might make for an interesting honours or Masters dissertation project...

[HT: Marginal Revolution]