But Arnold's prediction and the happiness literature are actually compatible.
I am not going to be trapped into talking about happiness research as if it makes behavioral predictions. Happiness research cannot make behavioral predictions at all. It consists of taking meaningless surveys, and the most it can do is make predictions about the "findings" of other meaningless surveys.
There is a classic paper in economic methodology by Nobel Laureate Tjalling Koopmans called "Measurement without Theory," which is an indictment of the "leading indicators" approach to macroeconomic statistical analysis.
I call happiness research "measurement without measurement," because the researchers are not really measuring anything. If you ask somebody to rate their happiness on a scale, you have no idea what the answer means. Are respondents reporting a feeling, or an evaluation of how they think they ought to feel? Are they reporting something instantaneous, or something that also combines looking backward and/or forward? Are they attempting to report their feeling/evaluation relative to themselves at some other time in life? Relative to other people at some time in their lives?
There is a classic problem in economics with comparing interpersonal utility. Bernie Saffran used to say that you could never disprove the hypothesis that all of the world's resources should be devoted to giving him ice cream. How do we know that the marginal utility to him of the millionth scoop of ice cream isn't higher than the marginal utility of one bread crumb to everyone else?
To my knowledge, happiness research has not solved this methodological problem. You have no more basis for saying that "Married people are happier than single people" than you do for claiming that Bernie's marginal utility from the millionth scoop of ice cream exceeds my marginal utility from the first bread crumb.
Happiness research is fascinating for the same reason that astrology is fascinating. And it is equally scientific.