A Samuelsonian validation test for happiness data

There is growing interest in the use of subjective well-being data, such as survey questions about happiness and life satisfaction. The existing validation tests determine whether these subjective measures have a positive correlation with objective measures of well-being, such as suicide rates and frequency of smiling. We propose an alternative test consisting of three steps: using regression analysis to infer preferences from subjective well-being data; using those estimated parameters to predict how a rational utility-maximizer individual should have acted; and comparing predicted behavior with actual behavior. This validation test can be compelling for economists, because it compares decision utility (i.e., preferences inferred from behavior) to reported utility (i.e., preferences inferred from self-reported well-being). We provide an application of this test based on a model of food consumption, estimated with one of the most widely used measures of subjective well-being: life satisfaction. We find that, across individuals, a one percentage point increase in the actual expenditure share is associated with a 0.76 (SE 0.196) increase in the share predicted by life satisfaction. Additionally, life satisfaction performs significantly better than other objective and subjective measures of well-being (e.g., income, satisfaction with income). The evidence suggests that life satisfaction offers some useful information about experienced utility.