A Survey-Based Procedure for Measuring Uncertainty or Heterogeneous Preferences in Markets
Executive Summary — People who buy retail prescription drugs, invest funds, or participate in auctions rarely have complete information about the product they are buying. Often the only auction information participants have is the number of bidders, observed bids, and product characteristics. If data from an auction, for instance, is a function of bidder behavior, then external survey data may help in testing hypotheses about bidding behavior. Researchers often avoid using surveys because they consume time and effort, but Yin presents a survey design technique and econometric tool to deal with a general population of survey respondents. Her application tested eBay online auctions selling personal computers. Key concepts include:
- Survey data may be a good complement for market data, especially for auctions, as a measure of uncertainty or different preferences.
- Survey data may be more valuable than other methods of evaluation because it exploits the human ability to assess complex sets of information.
- A survey may be implemented more quickly with a larger number of respondents, even if they are inexperienced, than with a smaller number of experienced respondents, by correcting for survey bias.
This paper shows how surveys can be used to generate a measure of the amount of information and/or heterogeneity of preferences within a market. This measure can be employed as a regressor in empirical work where variance in the dependent variable (e.g., auction prices, retail price dispersion, or investment choices in stocks, R&D, or education) might be explained by uncertainty about the value of the item being sold or the returns to investment choice and/or heterogeneous preferences in the market. The effects of incomplete information and heterogeneous preferences are usually relegated to the error term, which a) confounds these effects with other drivers of the error term and b) could lead to heteroskedasticity at best or omitted variable bias at worst. Furthermore, by specifically modeling the effect of this uncertainty or dispersed taste, one can estimate policy implications such as the effect of publicly introducing information into the market or selecting the pool of agents to change the distribution of preferences. I demonstrate the validity and usefulness of my survey-based procedure by using it to measure the mean and dispersion of private information signals in eBay online auctions for personal computers. I exploit a mixture of respondents with and without experience on eBay. The use of inexperienced respondents permits the survey to be implemented more quickly and with a larger number of respondents than if it were restricted to experienced respondents. The use of experienced respondents allows me to correct for potential bias from using more noisy, inexperienced responses.