Estimation bias in choice models with last choice feedback

The study compares two estimation methods for choice models with last choic e feedback, using simulated and real data. The first method ignores the imp act of unobserved heterogeneity on observed choices via presample choices, while the second method approximates this impact by a stochastic relationsh ip. In panels with less than 10 choices per panelist, the first method over states the impact of last choice on current choice and understates the impa ct of intrinsic preferences (i.e., brand intercepts). The second method per forms significantly better than the first method. Under both methods, an in crease in the number of heterogeneous coefficients in the model tends to in crease the bias in the estimates. The largest bias occurs when lagged choic e coefficients are heterogeneous. (C) 1999 Elsevier Science B.V. All rights reserved.