SAMPLE BIAS IN CONTINGENT VALUATION - A COMPARISON OF THE CORRECTION METHODS

This article compares various methods for correcting contingent valuation aggregate benefits when the sample is known to be biased. A sample is defined as the population, and response rates are simulated on the basis of a measure of salience. The simulated response rates suffer from nonresponse bias and selection bias. Coefficient and benefit estimates that result from weighting and self-selection correction approaches are compared with the true coefficient and benefit estimates. Our results indicate that at both low and high response rates the standard approach leads to bias, and either correction approach will reduce the bias in coefficient and benefit estimates.