In my first foray into behavioural economics I have encountered two themes; - Self-reported subjective wellbeing (SWB) is commonly used as a measure of wellbeing. Survey based SWB measures are frequently used as regressors or dependent variables. - Framing effects are also important. How a question or a choice is presented can influence the decisions made, perhaps radically.
Therefore there seems to be an obvious problem with using any particular survey based SWB measure for inference. How do we know that the SWB measure is the result of true SWB and not an artifact from framing?
Even using alternative surveys as a robustness check may not vitiate the problem if the two surveys frame the question similarly. For instance Proto and Rustichini (2015) use two measures of self-reported life satisfaction where people are asked to state their satisfaction on a scale between 0 or 1 (least satisfied) and 10 or 7 (most satisfied). Both these surveys present the choice fairly similarly. If you flipped things, i.e. made 10 least satisfied and 1 most satisfied, the results could be different.
Do you agree that this presents a problem? I have not seen the issue widely discussed (at least, not in the largely empirical papers I have read). How do behavioural economists address it?