Testing whether marginal values of tropical beach attributes are stable across CVM and choice experiments designs

Background/Question/Methods Several articles have tested the equivalence of Contingent Valuation Method (CVM) and Choice Experiment (CE) capabilities to yield equivalent estimates of total willingness to pay for a given recreation scenario. A few statistically test equivalence of coefficients on the attributes. However, to our knowledge, none directly test the equivalence of the marginal values of the attributes. We also compare the width of confidence intervals around the estimates of marginal values when using a main effects orthogonal design versus a quasi-orthogonal design. In addition, we compare CVM and CE results to marginal values estimated using CVM and the natural variation in attributes from the visitor’s most recent trip. Initially, fifty-six in-person interviews with open-ended questions were conducted in four beaches in northeast Puerto Rico to identify relevant attributes. The following were most frequently identified: wave height, absence of trash, crowding, and water clarity. These results were used as input to design two CVM and CE questionnaires. From December, 2010 to July, 2011, 214 CVM and 213 CE in-person surveys were administered in five beaches in Puerto Rico: Carolina, La Pocita, La Monserrate, Costa Azul, and Seven Seas. 

Results/Conclusions We find that, with the exception of the current trip CVM design that relies only upon the natural variation in the four attributes, that CVM and CE yield statistically equivalent marginal values. A comparison of marginal value estimations shows that CE and CVM confidence intervals overlap for each of the two statistically significant attributes, trash and water clarity. Further, the finding of statistical significance at the 5% level is not influenced by whether a pure orthogonal design or quasi-orthogonal design is used.  However, the narrower width of the confidence interval estimates based on the orthogonal design show the efficiency gain over the quasi-orthogonal design. Crowding and wave height were not statistically significant attributes, regardless of method and presentation format. Even though both were frequently mentioned in the pre-test as two of the four most important attributes to beach recreationists, the various model results consistently did show their lack of significance. These findings should help instill additional confidence in environmental economists that economic values from stated preference methods are fairly robust to the choice of whether to use CVM or CE.