Adjusting for Quality in the Benchmarking of Electricity Network Companies

Transcription

1 Adjusting for Quality in the Benchmarking of Electricity Network Companies Endre Bjørndal, NHH Bonn, May 27, 2014

2 Outline Introduction Implementation of quality costs in the Norwegian incentive regulation model Challenges and implications Conclusions 2

3 Incentive regulation of network companies Incentives for efficient organization, operation, and investments Revenue should be independent of the regulated company s own cost Revenue = cost of the marginal company, given the company s output (volume and quality) Operating income: depends also on the company s efficiency and cost Sufficient revenue level to attract both financial and human capital Competitive rate of return on invested capital Accept continual efficiency differences and super-profits Long asset life times and time profiles

4 Regulation, benchmarking and quality Norway Finland Germany Regulation period 5 years 4 years 5 years Individual efficiency requirements Benchmarking frequency Yes Yes Yes Annually Start of reg. period Start of reg. period Quality incentives Yes Yes Yes Quality integrated in benchmarking analysis Yes No No Present German model is similar to the Norwegian model for and Initial revenue cap based on historical costs Individual efficiency requirements 4

5 Yardstick regulation in Norway RCap = 0.6 C* C Annual updates of C and C* C based on accounting values Including capital costs C * based on benchmarking analysis: C * = E C DEA-based cost efficiency with total accounting costs as only input Separate models for distribution and transmission C * calibrated to let averagely efficient companies earn normal rate of return

6 Average efficiency Determines the normal rate of return (for representative company) Should average efficiency be 100 %? How is the industry s average efficiency affected by changes in the benchmarking model? (distribution only)

7 Average efficiency Should we adjust for this? The DEA model is not very strict in the first place A general efficiency requirement for an inefficient industry? Should not be determined by model choices and size of data set NVE has decided to adjust the results such that the industry rate of return over time is approximately equal to the reference rate of return, r NVE => Cost weighted average efficiency = 100 %

8 Accounting based capital costs Choice of capital base / capital cost will influence the measured efficiencies Book values versus replacement values Productivity relatively independent of age Accounting based costs with linear depreciations do not reflect economic costs very well Over-estimated efficiency in old networks, under-estimated in new

9 Evidence of age bias Eff. score based on book values 1 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 y = -0,641x + 0,968 R² = 0, ,1 0,2 0,3 0,4 0,5 0,6 0,7 Book value / Catalogue value Average data for

10 Estimation of quality costs All power interruptions are reported Long (> 3 min) from 1995 Short (< 3 min) from 2006 Companies report energy not supplied (kwh) Cost function (NOK/kW) estimated based on customer surveys and interviews Sintef energi / Pöyry (2012) Covered also voltage variations and rationing Only interruptions are implemented by NVE 10

11 Estimated cost of interruptions Customer group Cost of an interruption lasting r hours (NOK/kW) at the reference point in time All durations Additional adjustments for hour / weekday / month Factor for planned interruptions Agriculture 10.6*r Households 8.8*r hours > 4 hours Manufacturing 55.6*r *r Trade and services 97.5*r *r Public services 14.6*r *r Pulp/paper, energy-intensive industry 7.7*r *r

12 Cost components distribution companies Cost share Industry cost (MNOK) Interest ( 12.7 %) Depreciation ( 19 %) O & M ( 55.8 %) Thermal losses ( 9 %) Quality (VOLL) ( 3.5 %) Average data for

13 Quality costs over time for distribution and regional transmission Dagmar Source: Utvikling i nøkkeltall for nettselskapene (NVE, 2013) 13

14 Quality costs and company size Quality cost per customer (NOK) e+05 No. of customers Lower quality costs for larger companies Distribution companies, average data for

15 Quality costs and location Quality cost per customer (NOK, average) All companies with very low costs are in the south Clusters of companies with high costs Effect of environmental factors? 0 15 Distribution companies, average data for

16 Implicity quality penalty/reward Actual quality costs are deducted from the revenue cap for each company Allowed revenue Rcap VOLL = 0.6 C* C VOLL Net effect of quality adjustment 0.6 (VOLL* - VOLL) Incentive depends on yardstick parameter (0.6) Previous quality incentive ( ) Penalty / reward = expected VOLL acual VOLL Stronger quality incentives than in present model? 16

17 Direct compensation for long interruptions Duration (hours) Compensation NOK NOK NOK > per day Does not affect revenue caps Mainly relevant for households? Allocation effect between customers 17

18 Challenges and implications Specification of quality costs Measurement of quality Pricing Specification of other model elements Cost groups Cost drivers Noise 18

19 Specification of quality costs Norway Germany Interruptions covered All > 3 min Companies included All > customers Physical measure Energy not supplied (ENS) SAIDI * #customers Price depends on type of load? Yes No Price depends on time of interruption? Price adjustment for planned interruptions Yes Potential for improvement of German cost data Increase scope? More detailed measurement and pricing is possible, but may increase extent of untruthful reporting No

20 Model specification cost groups Model must include all relevant tradeoffs Maintenance costs Capital costs Norway Use of book values and linear depreciation can lead to overestimation efficiency for companies with old equipment Age bias compensates for higher quality costs in these companies? Germany Uses of annuity-based capital costs No age bias? 20

21 Model specification cost drivers Quality costs can be influenced by exogenous cost drivers This must be controlled for in order to create a level playing field for the companies 21

22 Environmental factors in previous models 1st regulation period ( ) Quality costs not included in benchmarking No environmental factors 2nd regulation period ( ) Quality costs as one of 5 input factors in a cost efficiency model «Expected» quality included as an output variable (cost driver), to control for difficult operating conditions 3rd regulation period from «geography» variables were selected from a long list of candidates Included as output variables in the DEA model Two-stage model from «geography» variables in second stage regression Inital DEA score adjusted based on regression results 22

23 Choice of cost drivers Based on statistical tests But all the relevant cost drivers must have been proposed in order to be tested! Involvement by industry participants Work groups, questionnaires, etc. 23

24 Potential cost drivers - inland companies (NVE, 2006) CapEx O & M Losses VOLL Technical life span of network is reduced by high share of forest in the concession area x x High share of forest increases the need for forest cleaning, and the cost increases with the productivity (fast growth) of the forest Forest in combination with heavy wind / snow is a cost driver x x High winter temperatures cause increased forest growth and the need for forest cleaning x x Low winter temperatures increase the duration of failures and reduce the possibility for doing maintenance work x x Low winter temperatures reduce the possibility for implementing investment decisions in an efficient manner X x

25 Potential cost drivers companies along the coast (NVE, 2006) CapEx O & M Losses VOLL Technical life span of network is shorter in a coastal climate x Corrosion due to saline air along the coast x x x The coast is more exposed to wind, and networks need to be dimensioned accordingly x x Companies serving islands not connected via road to the mainland have higher investment and operating costs x x x Dimensions of masts must be increased in areas with heavy wind x

26 Potential cost drivers companies in urban areas (NVE, 2006) Higher land costs (including land compensation) in urban areas Additional costs of underground cables in urban networks In urban areas much of the construction / maintenance work has to be done at night x x The possibility of laying more than cable in each trench contributes to lower costs in urban areas Laying underground cables in solid rock, asphalt, pavement stone etc. is more expensive than in areas with more favorable soil conditions x x CapEx O & M Losses VOLL The share of underground cables increases capital and failure costs, and reduces operating and maintenance costs x x x x x x

27 Handling of noise Quality costs are more variable than the other cost elements Variability coefficients for cost groups (distribution, ) Quality (VOLL) 0.61 Thermal losses 0.18 O & M 0.13 Depreciation 0.15 Interest 0.13 Companies should be benchmarked on the quality costs that they can influence via their decisions Random variation is irrelevant 27

28 Handling of noise Norway (DEA) Actual observation evaluated against front based on average data (5 yrs) Gives a robust front (cost function) Germany (SFA/DEA) Benchmarking based on data for a single year «Best of four» DEA vulnerable to noise 28

29 Conclusions Quality costs have been implemented as an input factor in the benchmarking of network companies in Norway Strength of quality incentives depends on the yardstick parameter (0.6) A similar implementation in the German model could benefit from improvements in the measurement/pricing of quality, and it could require other changes in the specification of costs and cost driver, as well as adequate handling of noise in the quality data 29