In the article, they make it sound like they’re so excited about having completely validated the correlation between self-reported happiness and objective measures. But the graph shows R = 0.6 — meaning that the claimed relationship accounts for only about 60% of the variability in the data.

So, there’s one or more other factors involved that, together, are nearly as strong as the objective measures at predicting self-reported happiness levels.

The other thing that bothers me, here, is that the y-coordinates of the points on the graph are misleading. What’s plotted is the average happiness score reported for that state — but is the variance among happiness scores similar among states?

That variance implicitly “smears” those points vertically — the higher the state’s variance, the longer the “smear”, which further damages the degree to which the linear relationship connects the variables.

Why don’t these articles give us links to the actual studies?