Accounting for attention changes in discrete choice experiments and modeling adverse selection in the cattle procurement market

Abstract

This first paper conducts a discrete choice experiment to estimate turfgrass producers' willingness to accept (WTA) values using different logit models and specifications to capture respondents' attention. We first estimate the mixed logit model and a generalized multinomial logit model with and without eye-tracking variables to demonstrate the importance of accounting for individuals' differing levels of attention during an experiment. Our study finds that marginal WTA values are biased when individuals' attention changes are not properly accounted for in the model specification. This finding leads to our second objective, to determine whether attention changes can be fully captured in the absence of eye tracking data by testing six alternative model specifications. All six models are able to detect learning and fatigue effects but are unable to fully capture changes in attention. Of the six alternative models tested, the two models that implement panel data offer more reliable and significant results, suggesting the type of data and model specification used may play an important role in diagnosing attention changes when compared to various heterogeneity models.

In the second paper, a potential adverse selection problem is hypothesized in the beef packing industry. Adverse selection arises from information asymmetry between buyers and sellers, where we suspect feeders may have an information advantage compared to packers regarding cattle quality. Assuming that higher quality cattle are sold through alternative marketing agreements and lower and/or unknown quality cattle are sold through the cash market, we expect to find adverse selection is present in the cash market due to uncertainty regarding quality. We expect to find no or a lower level of adverse selection in the alternative market as premiums and/or discounts alleviates information asymmetries. Using a rare feedlot transaction dataset as well as a private regional aggregate dataset, we begin our analysis employing both Heckman's two-step model and the generalized Roy's model to determine the presence of selection bias. We then calculate cash price differences that a randomly selected lot would receive versus a lot that was sold in the cash market to determine the potential impact quality may have on market prices. We ultimately find the cash market has a negative adverse selection issue where lower quality cattle receive lower prices.