The dependent variable, percent overlap, takes positive values, but the construction of
dissimilarity indexes of explanatory variables allowed for differences among states take on
values between negative and positive infinity. A negative (positive) value of the dissimilarity
index suggests that the base state’s indicator is relatively higher (lower) than that of the
comparator state. Consider the case of temperature differences, where a negative dissimilarity
index implies that the base state’s average temperature is higher than that in the comparator state.
Therefore, positive dissimilarities can be hypothesized to have different effects on overlaps than
the negative indexes. There is some scientific evidence supporting such differences in effects for
ecological variables. For instance, Brown, Stevens, and Kaufman (1996) indicate that an
introduced species’ relationship with others may differ depending on whether the new
environment is warmer or colder relative to its native environment. To further illustrate,
consider figure 2, where overlap is plotted on the Y-axis and the dissimilarity is represented by
the positive and negative quadrants of the X-axis. Larger dissimilarity leads to lower overlap on
either quadrants. Therefore, we set up slope and intercept dummies to allow the coefficient on
any explanatory variable change between negative and positive realizations of the dissimilarity
index. We follow up with a test of the restriction that the coefficient is the same regardless of the
sign of the dissimilarity index. In the case of lobbying dissimilarities, the positive and negative
dissimilarity indexes simply reflect relative strength of an interest-group between any two states.
The general model is then:
(12) Lj = D+α0 + D-δ0 + β( D+X. ) + γ'( D- X. ) + εy
where D+= [D1+,D2+,...,D1+2] is a set of dummy variables which take value 1 when Xij > 0, e.g.,
D1+ =1lfXl1j >0;else=0. Similarly, when Xlj ≤ 0, dummy variables in
17