We advocate a more formal structural approach for comparing WTP for non-market or pre-test-market goods conveyed by fundamentally different preference elicitation mechanisms. Seven independent samples of respondents were asked to value the identical good. Elicitation methods include one actual purchase and six widely used hypothetical choice formats. Using a common underlying indirect utility function (and stochastic structure) allows data for different elicitation methods to be used independently, compared pair-wise (as in much of the earlier literature) or pooled across all samples in one unified model with heteroscedasticity across elicitation methods. Our differences in estimated WTP for the individual models are typical of earlier findings. However, pooled-data models that allow for heteroscedasticity reveal that while there are substantial differences in the amount of noise in the different samples, a common underlying systematic component of the preference structure cannot be rejected for at least four (and possibly five) of these seven elicitation methods.