19
Finally, we also decided to include in the model the state per capita GDP as a
controlling variable. This variable might be a good way of measuring the level of autonomy
of the states vis a vis the national government. Thus, we expect that the greater the GDP per
capita the less inclined and constrained the governor will be to follow the national
executive’s preferences. We can also test for the effect of organizational isomorphism. It is
expected that the diffusion of innovations would follow a pattern by which it would spread
from the richest and most institutionally developed states to the more underdeveloped
states. According to the first hypothesis, we predict negative coefficients for this variable.
For the second hypothesis, we predict positive and significant coefficients.
Table 3 - Logit Estimation on the Creation State Regulatory Agencies
Models
(1)
(2)
(3)
(4)
(5)
(6)
Constant
-.182
(.605)
-3.417
(2.113)
-6.690
(3.961)
-12.691
(6.902)
-11.847
(7.082)
-5.813
(4.299)
Privatization
2.890
(1.197)**
5.349
(2.248)**
5.336
(2.513)**
6.403
(3.226)**
5.874
(3.197)*
5.269
(2.502)**
Volatility
1.735
(.989)*
2.308
(1.379)*
2.426
(1.394)*
2.778
(1.758)*
2.570
(1.539)*
Transferences
.364
(.294)
.332
(.307)
.291
(.523)
.374
(.371)
Number of
Effective Parties
.594
(.450)
.774
(.490)
GDP
-.710
(.868)
-.378
(.601)
% Prediction
77.77
88.88
88.88
88.88
92.59
88.88
Log likelihood
-11.3197
-8.3731
-7.0832
-5.9978
-5.1374
-6.7661
Nº observations
27
27
27
27
27
27
Standard error in parenthesis
Significance: 1% ***, 5% **, and 10% *
The decision to set up state IRAs
As we can see from the Table 3, the key results in all regressions are the positive
and significant coefficients for Privatization and Volatility variables in all six models tested
with different combinations of variables. These coefficients show that even when
controlling for other variables that may affect the Governor’s decision of whether to
construct an independent regulatory agency, an increase in the privatization and volatility
also increases the probability that the governor will delegate powers to a regulatory agency.
All different models presented a high percentage of prediction which suggests a good fit of
the model. However, the model 5, the one that was included all variables, appears to be the
highest one. Although the other variables were not statistically significant, they behaved
Convention