BIAS IN DISCRETE RESPONSE CONTINGENT VALUATION

The empirical literature on discrete response contingent valuation has found that seemingly innocuous changes in the statistical models estimated result in significantly different point estimates of willingness to pay. This paper hypothesizes and tests several potential explanations for these results. First it investigates and compares the biases inherent in single-bounded and double-bounded maximum likelihood estimation procedures and examines how they react to various bid designs and sample sizes. Then it examines the presence and identification of ''outliers'' in binary choice data and how these outliers influence estimation. Finally, it presents an alternative approach to addressing the issue of outliers which explicitly acknowledges the possibility of upwardly biased response probabilities. (C) 1995 Academic Press, Inc.