First the link to this week’s complete list as HTML and as PDF.
***
Looking at the comment by Gilbert et al. and the reply by Anderson et al. I shall disregard both their arguments’ merits and take Gilbert et al.’s most optimistic numbers of 85 % and 66 % at face value. What do they mean? The unspoken and unproven but implicit and generally accepted meaning of “statistical significance”
is, that only 5 % of results occur by chance and 95 % (or at least well over 90 %) should be replicable. This holds for the borderline significance of p=0.05. Most studies I read are less than p=0.01 or even p=0.001. So the implicit claim is that at least 99 % of them ought to be successfully reproduced. They aren’t. Something very basic is very much wrong with the results and the reporting in the psychological and sociological sciences.
***
For me the surprising bit in the convincing study by Gluth & Fontanesi and Hein et al. is not the result but part of the background. If you divide subjects into selfish and prosocial based on their contribution towards a neutral partner, it is the selfish ones that show empathy towards a partner, who has suffered, and the prosocial ones who reciprocally reward a partner, who has been nice to them before. I would naïvely have expected exactly the opposite outcome.
***
Kanazawa bases his studies on sound theory and states plausible hypotheses. I would therefore expect his results to be sound and most of them are. But his latest result in Li & Kanazawa looks a bit too much like (conscious or unconscious) data milking for the desired result.
Population density at the state level is not directly observable and should, according to their own hypothesis, not directly influence subjective wellbeing. It does and its effect is far stronger than that at the county or local level. Contrary to what they state in their conclusion this does not support their hypothesis but counters it. What they disregard is, that population density at the state and county levels is not arbitrary but there are sound underlying reasons, why many people live there and not somewhere else. These reasons and the resulting competition and e.g. pricing will probably turn out to be the true drivers of the result. And then a locality with exactly the same density may either be a rural and leafy outskirt in a dense state or a town aggregation in a low density one. This could explain a lot of why local density looks so unimportant here.
The more surprising result is the one that states intelligent people are happier, if they have fewer friends – not less affected and less unhappy, but actually happier. The answer may lie in their measure for friendship. Let’s look at the age group here. At 22 years of age the more intelligent people tend to be students or at the start of a successful career. If the latter they will work long hours and may well be satisfied in spite of only seeing their friends at weekends. Students on the other hand tend to see their friends every day, but their interactions do not conform to those measured here. Young employees too tend to find their friends among colleagues. For those stuck in boring and less satisfying jobs the meaningful exchange tends to happen after hours and will be recorded here. As one more confounder, universities are not evenly distributed and having been accepted at a good one may feel a lot different from living in the same general area outside academic life.