Temporal Stochastic Inflation in Choice-Based Research

W examine the specification and interpretation of discrete-choice models used in behavioral theory testing, with a focus on separating “coefficient scale” from “error scale,” particularly over time. Numerous issues raised in the thoughtful commentaries of Louviere and Swait [Louviere, J., J. Swait. 2010. Discussion of “Alleviating the constant stochastic variance assumption in decision research: Theory, measurement, and experimental test.” Marketing Sci. 29(1) 18–22] and Hutchinson, Zauberman, and Meyer (HZM) [Hutchinson, J. W., G. Zauberman, R. Meyer. 2010. On the interpretation of temporal inflation parameters in stochastic models of judgment and choice. Marketing Sci. 29(1) 23–31] are addressed, specifically the roles of response scaling, preference covariates, actual versus hypothetical consumption, “immediacy,” and heterogeneity, as well as key differences between the experimental setup in Salisbury and Feinberg [Salisbury, L. C., F. M. Feinberg. 2010. Alleviating the constant stochastic variance assumption in decision research: Theory, measurement, and experimental test. Marketing Sci. 29(1) 1–17] and those typifying intertemporal choice and construal level theory. We strongly concur with most of the general conclusions put forth by the commentary authors, but we also emphasize a central point made in our research that may have been lost: that the temporal inflation effects observed in our empirical analysis could be attributed to stochastic effects, deterministic influences, or an amalgam; appropriate inferences depend on the nature of one’s data and stimuli. We also report on further analyses of our data, as well as a meta-analysis of HZM’s Table 1 that is consistent with our original findings. Implications for, and dimensions relevant to, future research on temporal stochastic inflation and its role in choice-based research are discussed.