Bridging between Hypothetical and Incentivized Choice
General, Cognitive and Mathematical Psychology
Statistics and Econometrics
Final Report Abstract
The project adapted and developed dependent Poisson processes as the probabilistic structure underlying preferential choice in marketing and economics. The resulting model reflects how the decision maker’s cognitive processing translates into evidence supporting the choice of a particular alternative from a set. In contrast to existing approaches, the model defines counterfactual choice probabilities under different levels of choice effort, as a function of incentives to identify the best alternative from a set described along multiple attributes each. By modeling correlated evidence accumulation for choice alternatives in a set and bounds to the amount of information processing, the proposed model can rationalize satisficing behavior, i.e., the idea that a chosen alternative is not necessarily identified as the best but only as "good enough.” Relatedly, the model rationalizes that choice between alternatives that require trading-off advantages and disadvantages across different dimensions or attributes of an alternative endogenously results in more information processing and is more error-prone than the choice between alternatives that involve fewer trade-offs. Finally, the model motivates increasing sensitivity to price as a function of making choice consequential and thus can explain what is known as hypothetical bias in market research aimed at forecasting actual demand from behavior observed in discrete choice experiments. At a technical level, we contribute an algorithm to estimate the proposed model overcoming the intractability of the implied likelihood function. The algorithm is computationally intensive but reasonably fast, thanks to parallel computing on a graphics card.
Publications
-
"Essays on Cognitive Decision Models and Causal Inference in a Bayesian Framework,” doctoral dissertation, Goethe University, Frankfurt
Laghaie, Arash
