103 Journal of Marketing Research Vol. XXXII (February 1995), 103–110 ROLAND T. RUST and NAVEEN DONTHU* No retail store choice model, no matter how many relevant variables it might include, can realistically expect to model all the variation in store choice. There are always some variables that are left out, because they are difficult to measure, they have not yet been conceptualized in theory, or their estimated parameter stability suffers when an excessive number of predictors are included. Because these omitted variables can be correlated with geographic location, model misspecification error may itself be correlated with location. Estimating the geographically localized misspecification errors therefore suggests itself as a method for estimating (and predicting) the effects of these omitted variables. The authors show that spatial nonstationarity of the model parameters may also be expressed as an instance of omitted variables and therefore be addressed using their method. They show, using both a simulation study and an empirical natural experiment, that estimating the geographically localized misspecification error can appreciably reduce prediction error, even when the predictor model is reasonably well specified. Capturing Geographically Localized Misspecification Error in Retail Store Choice Models Selecting the best possible site for a retail store (or service facility) is often a critical factor in that store’s eventual suc- cess (Achabal, Gorr, and Mahajan 1982; Ghosh and Craig 1983). The site selection problem is more complicated for a chain, which must worry about whether the new store will cannibalize its existing stores by stealing their customers. In either case, site selection cannot be accomplished effective- ly without knowledge of where the customers are and how they choose their store, as well as knowledge of the geo- graphic configuration of existing stores. Because the geographic configuration of the existing stores is easily obtained, the most important information that must be determined by primary research is where the cus- tomers are located geographically and how they choose a store. Methods for flexibly estimating the geographic customer density from sample scatter plot data have been proposed in recent years (Donthu 1991; Donthu and Rust 1989; Rust and Brown 1986). There is also extensive literature on consumer choice models, much of it based on application of logit (e.g., Batsell and Lodish 1981; Gensch and Recker 1979; Guadag- ni and Little 1983; McFadden 1980). Consumer choice models, when applied to retail attraction, have been in com- mon usage for many years (e.g., Fotheringham 1980; Huff 1964; Louviere 1984). Retail attraction models typically predict store choice based on a number of variables. However, it is a practical impossibility to include all of the variables that affect choice. There are several reasons for this. First, some vari- ables may be very difficult to measure, and thus they are not practical to include. Second, some variables that affect choice may not have been conceptualized or identified by the researcher. Third, even if it were possible to identify and measure all relevant predictors, it would not be advisable, because the use of too many variables leads to parameter in- stability and a decline in predictive accuracy (Inagaki 1977; Rust and Schmittlein 1985). The result is that a well-specified retail attraction model will include what are thought to be the most important vari- ables, but it will necessarily omit many others. The in- evitable presence of omitted variables means that the typical model is misspecified (in this case, underspecified). This re- sult is inevitable, and the most careful model building and testing will never be able to completely specify the model without suffering parameter instability and declines in pre- dictive accuracy, which will more than offset the benefits. *Roland T. Rust is Professor of Marketing and Director of the Center for Services Marketing, Owen Graduate School of Management, Vanderbilt University. Naveen Donthu is Associate Professor of Marketing, Georgia State University.