Although the contingent valuation method has been widely used to value a diverse array of non-market environmental and natural resource commodities, recent empirical evidence suggests it may not accurately estimate real economic values. The hypothetical nature of environmental valuation surveys typically results in responses that are significantly greater than actual payments. Economists have had mixed success in developing techniques designed to control for this “hypothetical bias.” This paper highlights the role of experimental economics in addressing hypothetical bias, and identifies a gap in the existing literature by focusing on the underlying causes of this bias. Most of the calibration techniques used today lack a theoretical justification, and therefore these procedures need to be used with caution. We argue that future experimental research should investigate the reasons hypothetical bias persists. A better understanding of the causes should enhance the effectiveness of calibration techniques.

This paper reports results from a contingent valuation based public good experiment conducted in the African nation of Botswana. In a sample of university students, we find evidence that stated willingness to contribute to a public good in a hypothetical setting is higher than actual contribution levels. However, results from regression analysis suggest that this is true only in the second round of the experiment, when participants making actual contributions have learned to significantly lower their contribution levels. As globalization expands markets, and economies such as Botswana's continue to modernize, there is a growing need to understand how hypothetical bias will influence the valuation of public goods.

We compare two approaches to mitigating hypothetical bias. The study design includes three treatments: an actual payment treatment, a contingent valuation (CV) treatment with a follow-up certainty question, and a CV treatment with a cheap talk script. Our results suggest that both the follow-up certainty treatment and the cheap talk treatment produce willingness-to-pay (WTP) estimates consistent with the actual payment treatment. However, the follow-up certainty treatment provides response distributions at all offer amounts that are statistically similar to the actual payment treatment, while the cheap talk treatment provides similar responses only at some offer amounts. Furthermore, the cheap talk treatment is effective only for inexperienced individuals. We conclude that the follow-up certainty approach is more consistent than the cheap talk approach for eliminating hypothetical bias.

The Orbit demand model allows the magnitude of the calibration to stated purchase intentions to vary based on the magnitude of the stated quantities. Using an empirical example of stated trips, we find that the extent of calibration varies substantially with less correction needed at small stated trips (-25%) but larger corrections at higher quantities of stated visits (-48%). We extend the Orbit model to calculate consumer surplus per stated trip of $26. Combining the calibrations in stated trips and value per trip, the Orbit model provides estimates of annual benefits from 60% to 111% less than the count data model.

In this paper we compare results from an in-store field experiment and a mail survey choice experiment (CE) to investigate CE's capacity in predicting grocery store market share. For the comparison, we used three seafood products: freshwater prawns, marine shrimp, and lobster. CE estimates were obtained via four econometric models: the conditional logit, the random parameter logit, the heteroskedastic extreme value, and the multinomial probit. We found that the level of control in the grocery store experiment and the choice of econometric model influenced the capacity of CE to predict grocery store market shares.

Hypothetical bias is a pervasive problem in stated-preference experiments. Recent research has developed two empirically successful calibrations to remove hypothetical bias, though the calibrations have not been tested using the same data or in a conjoint analysis. This study compares the two calibrations in a conjoint analysis involving donations to a public good. Results find the calibrations are biased predictors of true donations but that calibrated and uncalibrated models together provide upper and lower bounds to true donations.