The stability of stated preferences for public goods: Evidence from recent budget games

We have investigated the stability of the individual response in recent budget games based on survey data, which is an important requirement for the reliability of this instrument. Budget games have gained popularity due to the problems encountered with alternative methods to determine preferences for public goods, such as the analysis of actual public expenditure date using median-voter theory or similar approaches. The short-term test-retest correlations (within an interval of one month) turn out to be rather low, typical around 0.3. No explanation of the test-retest differences could be found from the usual socio-economic and political characteristics of the respondents or from information characteristics of the survey design. Also, the pattern of budget-game outcomes for different countries and different periods is rather similar. The cumulative evidence suggests that the survey response to budget games is generated to a large extent by very general notions on the (un)desirability of public goods: defence is bad, education and health care are good. This implies that outcomes are often not related to the actual level and structure of public expenditure or revenues. As a result, the individual responses, even to the more sophisticated budget games, are subject to large uncertainty margins. Our results should warn researchers and, even more important, policy makers against giving too much weight to stated preferences for public expenditure or taxation levels obtained from budget games. Of course, further research is needed to obtain the precise limits of the instrument, including laboratory experimental economics.