Other ecosystem differences represented by land, soil and water share (dissimilairty)
indexes do not significantly affect regulatory congruence as illustrated in figure 2. Of the 24
coefficients on land, soil and water share indexes, only 4 have the expected sign with statistical
significance (table 2). In 2002 (table 3), only the coefficients on water share index are
significant in List 2. These results prompted additional tests on the relevance of land, water and
soil indexes for regulatory congruence. Restricting their coefficients to zero or only a subset to
zero (e.g., land and water share) did not alter the results on other included variables. Since some
of these restrictions are rejected and the efficiency losses are minimal, we retain all 3 variables in
List 1 through 4. We suspect that the information embodied in land and water indexes is likely
captured in agronomic and precipitation indexes, respectively.
Agronomic Dissimilarities and Regulatory Congruence: With regard to agronomic dissimilarities
biased in the direction of the base state (negative quadrant), the coefficients on field crop indexes
have the expected sign with statistical significance in List 1, 2, and 3 (table 2). When
dissimilarities are positive, again the field crop and irrigated land share indexes mostly have a
significantly negative coefficient (List 1, 3, and 4). However, those on the irrigated land share
index are mixed when dissimilarities are biased toward the base state. Overall, only one of the
twenty coefficients on agronomic dissimilarities has the unexpected sign with statistical
significance. The results for 2002 are similar to those in table 2 (1997). As in the case of
ecological dissimilarities, the most common results suggest that the relationship between
agronomic characteristics and regulatory congruence is best illustrated by figure 2.
Table 4 presents a summary of the direction of the effects of ecological and agronomic
dissimilarities on NXWS and NXW regulatory congruence. A blank space in table 4 indicates
lack of statistical significance. As noted earlier, we mostly obtained positive coefficients when
20