A More Flexible Model or Simply More Effort? On the Use of Correlated Random Parameters in Applied Choice Studies
The random parameter logit model has become the dominating model for analyzing stated choice data in environmental valuation. The unrestricted version of the model with correlated random parameters, however, is rarely applied. An important advantage of this specification is that the correlations bet...
| Autores: | , |
|---|---|
| Tipo de documento: | artigo |
| Data de publicação: | 2018 |
| País: | España |
| Recursos: | Universidad del País Vasco |
| Repositório: | Addi. Archivo Digital para la Docencia y la Investigación |
| OAI Identifier: | oai:addi.ehu.eus:10810/64407 |
| Acesso em linha: | http://hdl.handle.net/10810/64407 |
| Access Level: | Acceso aberto |
| Palavra-chave: | agri-environmental measures choice experiment correlated parameters random parameter logit model |
| Resumo: | The random parameter logit model has become the dominating model for analyzing stated choice data in environmental valuation. The unrestricted version of the model with correlated random parameters, however, is rarely applied. An important advantage of this specification is that the correlations between the parameters are not restricted to zero. These correlations can arise due to a behavioural phenomena or scale heterogeneity. One consequence of this might be that derived willingness-to-pay or to-accept estimates are under- or overestimated, providing decision makers with incorrect estimates. We compare both model specifications using data from a study about farmers’ willingness to accept compensation for implementing agri-environmental measures in Brandenburg, Germany. For this data both model specifications - with and without correlated random parameters - provide similar willing-to-accept estimates, but the model with correlations performs better despite the higher number of parameters. As our findings could be case study specific, we want to encourage especially applied researchers to estimate also specifications with correlated random parameters. Applying only models with uncorrelated random parameters can result in biased estimates and thus provide incorrect information to decision makers. |
|---|