http://dx.doi.org/10.2139/ssrn.5258503

Obtaining valid answers from respondents has been a central concern of stated preference (SP) studies. In contrast, sample representativeness is the main focus of current public polling accuracy debates, largely due to declining survey response rates. This paper adapts the total survey error framework to SP research to illustrate the trade-offs between measurement and representativeness errors. We then introduce the multi-level regression and post-stratification (MRP) modelling approach to welfare estimation using discrete choice models. Using two water resource valuation surveys, we demonstrate how MRP can help researchers to (i) generate population-relevant welfare measures from non-representative samples, (ii) estimate preference heterogeneity and distributional impacts across people, and (iii) evaluate the impacts of removing potentially invalid responses using a consistent target population frame. We propose MRP as a complementary modelling approach for non-market valuation data and discuss its opportunities and limitations.

This is a preprint article, it offers immediate access but has not been peer reviewed.

http://dx.doi.org/10.2139/ssrn.5258503