The GCIMs, the earnings-and the margin sharing plans, the price caps and the service quality plans are included in this analysis.
This section reports on the estimation of the regression coefficients for the reduced-form model outlined above, and evaluates the conjectures regarding the impact of the GCIMs on the two consumer rates.
The coefficient on the GCIMs has the expected sign in the regressions with the utility-type fixed effects, but remains statistically insignificant.
This work is thus content with studying the cumulative effect of the GCIMs, and abstracts from correctly matching the commodity costs to the price reductions in multiple time periods.
In summary, after controlling for the suspected selection biases due to the unobserved heterogeneity across states and utilities, we fail to find the effect of the GCIMs on utilities' consumer prices.
Although we cannot control for the suspected relationship between gas rates of different consumer classes without appropriate instrumental variables in the system of equations, we can examine whether the correlation of residuals in the above residential and commercial regressions has changed when GCIMs were implemented.
If this coefficient is significantly different from zero, implementation of the GCIMs constituted a structural change in the relationship between the residential and commercial prices.
Although we could still worry about the joint determination of prices, we can safely assume that the same principles determined the relative prices both before and after the introduction of the GCIMs. APPENDIX 3.
(2.) Minnesota Public Utilities Commission's 1999 report, for instance, reads: "[At the other utilities, without GCIMs in place,] customers on average received lower gas costs without providing their utility with $10.5 million incentive to perform."
Implementation of the GCIMs across the US Gas Cost Incentive Mechanisms By Year Year 1992 0 0 1993 1 1 1994 8 5 1995 8 10 1996 10 21 1997 14 26 1998 15 28 1999 14 29 2000 16 30 2001 17 33 2002 17 34 2003 17 33  States with GCIMs  LDCs with GCIMs Source: State PSC surveys Note: Table made from bar graph.
The difference between the utility's average costs and the benchmark (sometimes excluding a certain dead band) proxies for the savings that are attributable to the GCIM, which are shared between the shareholders and ratepayers at a predetermined rate.
In all but the last regression above, the GCIM has an unexpected coefficient, which is not significantly different from zero.