The median voter, setters, and non-repeated construction bond issues

Abstract

The median voter model pervades the literature concerning the estimation of demands for collective expenditures but with some unanswered criticism. The general issue addressed in this paper is whether observed collective expenditure levels and the characteristics of observed majorities are more adequately described by the median voter model or the Romer and Rosenthal setter model. On a sample of non-repeated hospital construction referenda bond issues, the median voter model and the Romer and Rosenthal setter model are compared in both full-information and uncertainty contexts. Little empirical support of the setter model is found. Indeed, the evidence supports the absence of expenditure maximizing behavior. The specific conclusions are: (1) the standard median voter model explains quite well for this sample of non-repeated bond issues, (2) the explanatory power of the median voter model appears to vary across both voting arrangements and types of reversion threats, (3) the explanation of hospital expenditures which have been judged ‘too large’ may be due to how benefits and costs are distributed among voters, and (4) the preponderance of large majorities for hospital construction, in the face of evidence that such facilities typically become a burden on the general tax fund, is consistent with setters offering the median ideal in the presence of a reversion below the ideal.