Efficiency and Regulation of Electricity and Gas Distribution Companies

Transcription

1 Efficiency and Regulation of Electricity and Gas Distribution Companies How to use efficiency measurement in regulation? anel Tooraj Jamasb: Benchmarking and Regulation in Energy Industry: An Overview Einar Hope: Benchmarking and regulation: itfalls and issues in electricity regulation Massimo Filippini and Mehdi Farsi: Regulation and Measuring Cost Efficiency with anel Data Models: Application to Electricity Distribution Utilities

2 Regulation and Measuring Cost Efficiency with anel Data Models: Application to Electricity Distribution Utilities Annual conference of the International Association for Energy Economics rague 2003 Massimo Filippini and Mehdi Farsi Center for Energy olicy and Economics Swiss Federal Institutes of Technology ETH Zentrum, WEC CH-8092 Zurich, Switzerland

3 Outline Goals: Estimation of inefficiency scores in electricity distribution utilities using different methods Study the implications of the empirical results in performance-based regulation Contents: Introduction Empirical Analysis Conclusion 3

4 Efficiency concept roduction Efficiency Cost Efficiency Scale Efficiency Technical Efficiency Allocative Efficiency Relevant for regulation 4

5 Inefficiency measurement methods Simple Indicators Average cost; artial productivity indicators; TF Operations Research Data Envelopment Analysis; Free Disposal Hull Econometric Approach Cost Frontier Regression 5

6 TC TCobs Observed cost, Y i Cost Frontier Eff i TCfro EFF i = Frontier Cost TC obs TC fro 1 Frontier cost function: TC=f(Y) Model heterogeneity with more explanatory variables: TC=f(Y; Z,...) Y 6

7 Cost Frontier Analysis Type of Data Cross Section anel Data Displaced OLS (Deterministic cost function) Displaced OLS (Deterministic cost function) Stochastic Frontier (half-normal, truncated etc) Fixed Effects Model (no assumptions on correlation) Random Effects Model (no distribution assumption) Max. Likelihood Estimation (half-normal, truncated etc) 7

8 Empirical Analysis Model specification -Explanatory variables -Functional form -Hypotheses Data Collection Sensitivity Analysis Estimation with Econometric Methods Ineffiency Measures Use in Regulation 8

9 Model Specification C = C( Y, K, E, L, LF, CU, AS, HGRID, DOT, DW, T ) C : Total costs Y : Output (total number of kwh delivered) K,L,E: rices of capital, labor and input power LF: Load factor CU: Number of customers AS: Size of the service area of the distribution utility HGRID: Indicator for high-voltage transmission network DOT: Indicator for auxiliary revenues (> 25%) DW: Indicator for forest coverage (>40%) roblem: definition of the capital price 9

10 10 Functional form T DW DOT HGRID LF CU AS Y T L L K K LF CU AS Y β δ δ δ β β β β β β β = ln ln ln ln ln ln C ln

11 Data 59 Swiss electricity distribution utilities eriod => Unbalanced panel with 380 observations Data sources: (Unpublished) financial statistics on electric utilities (Swiss Federal Office of Energy) Mail survey Area statistics (Swiss Federal Office of Statistics) 11

12 Cost efficiency analysis Minimum Maximum Average Median 95 ercentile Number of Firms OLS RE (GLS) RE (MLE) FE

13 Correlation between efficiency ranks DOLS RE (GLS) RE (MLE) FE DOLS 1 RE (GLS) RE (MLE) FE

14 Implication for some typical companies Company Inefficiency Score OLS RE (GLS) RE (ML) FE A B C D E The companies are adopted based on the ranking obtained from the RE (GLS) model: A: median; B: most efficient; C: least efficient; D: 1st quartile; E: 3rd quartile. 14

15 roblems in application of efficiency indicators in regulation: From an econometric point of view it is not straightforward to specify which model is the best one. Only after a careful econometric analysis the efficiency indicators can be used in regualtion. We should be careful in using directly the level of inefficiency scores in regulation, for instance in a price cap regulation formula. 15

16 Conclusions The estimated frontier cost function can be a useful instrument for yardstick and price cap regulation. Advantages: Heterogeneity factors can be incorporated in setting yardstick and price cap regulation (fair benchmarking). The regualtor and firms should be aware that the results can be influenced by the model specification, functional form and the econometric approach. The use of intervals instead of point estimators for efficiency indicators could be a possible improvement. 16

17 THANK YOU FOR YOUR INTEREST

18 OLS Estimated functions are well behaved. Regression Results Random-Effects (GLS) Coeff. Std. Err. Coeff. Std. Err. Coeff. Std. Err. Coeff. Std. Err. lny lncu lnas lnlf lnl lnk HGRID DOT DW T Constant Almost all the parameter estimates are statistically highly significant and have the expected sign. Some coefficients across the models are similar. Random-Effects (MLE) Fixed-Effects

19 Implication in an example of price-cap regulation rice cap: t + 1 = t (1 + CI TF X + Z) t = 19 CI = 1% TF = 1% Z = 0 X = Inefficiency Termof regulation= 5 years Company (based on GLS) rice Cap (Cents) OLS RE (GLS) RE (ML) FE Median Most Efficient Least Efficient st Quartile rd Quartile

20 rediction errors Type of prediction Out-of-sample 1-year-ahead Model Average absolute error Maximum absolute error 95 percentile absolute error Average prediction bias Number of predictions OLS RE FE OLS RE FE Errors are given in percentage of the actual costs.

21 Stochastic Frontier Model with anel Data OLS : lnc it = lnc ( yit, wit) + v it v it ~ iid(0, σ 2 v ) Deterministic ( y, w ) + u u 0 : ln C ln C it = it it it it ( y, w ) + u + v u 0 Stochastic : lnc lnc it = it it i it i Different Specifications of u i : half-normal, truncated,... Maximum Likelihood No particular specification of u i with anel Data: Random-effects model Fixed-effects model 21

22 Effect of outputs and output characteristics on total costs OLS RE (GLS) RE (MLE) FE lny lncu lnas lnlf Estimated functions are well behaved. All the parameter estimates are statistically significant and have expected signs. Some coefficients across the models are similar. 22

23 A possible improvement in the practical application of cost frontier analysis: Regulated Company Cost Information ropose a price Benchmarking Regulator redict Cost Interval Negotiation rice within the interval No Ye s Approval