An International Meeting of the American Accounting Association American Accounting Association 2006 Annual Meeting August 6–9, 2006 Washington, D.C.

Stewart Jones
The University of Sydney

David A. Hensher
The University of Sydney

Abstract: In this paper we review a variety of advanced discrete choice models (i.e., nested logit, mixed logit and latent class logit) and assess their potential usefulness in accounting research. Using an empirical illustration from takeovers research, we find that in all cases advanced choice models provide significantly greater explanatory power than standard logit (measured by the overall improvement in the log-likelihood function). We further find that the mixed logit and latent class models have the highest overall predictive accuracy on a holdout sample, while standard logit performed the worst. Moreover, the analysis of marginal effects of all models indicates that use of advanced choice models can lead to fundamentally different behavioural interpretations of the role and influence of explanatory variables and parameter estimates in model estimation.

Back to Session Listing

AAA Home Page