existence of unobservable correlation across choices. Two different method-
ologies have been used. Hansen (1987), Ondrich & Wasylenko (1993) and
Guimaraes, Rolfe & Woodward (1998) estimated a two-step limited informa-
tion nested logit. The difficulty here resides in the identification of the upper
levels as they may constitute unrealistic scenarios for the decision-maker.
Moreover, it is sometimes difficult to conceive of regional characteristics
that affect upper level location choices in ways different from the elemental
choices. Consequently, most authors [e.g. Bartik (1985), Woodward (1992),
Luker (1998), Levinson (1996) and Head, Ries & Swenson (1999)] have at-
tempted to control for the IIA violation by introducing dummy variables
for larger regions.7 Both approaches, however, and importantly, are unsat-
isfactory because they are only valid if one is willing to assume that the IIA
assumption holds within subsets of the choice set (lower level nests for the
nested logit solution and larger regions for the dummy procedure).
A recent strand of empirical research has modeled the firm location deci-
sion problem using Poisson (count) models and microlevel spatial data sets
[Papke (1991), Wu (1999), Coughlin & Segev (2000)and List (2001)]. These
Poisson studies approached the location problem differently than the CLM.
They relate the number of new plants being opened at a particular site to
a vector of area attributes. The Poisson regression is particularly advanta-
geous in dealing with large spatial choice sets. Thus, what was perceived as
a drawback in the CLM model becomes an advantage in the context of count
models.8 At the same time, the authors claim that extensions of the Poisson
regression model can be used to address known problems that surface when
applied to location studies. In particular, this is the case of the overdisper-