巴西REGULATORY SCHEMES FOR THE BRAZILIAN MARKET OF ELECTRICITY -Artigo_002_Serrato

巴西电力市场监管

REGULATORY SCHEMES FOR THE BRAZILIAN MARKET OF ELECTRICITY TRANSMISSION

Maria da Conceição Sampaio de Sousa,

Eduardo Serrato, eserrato@http://wendang.chazidian.com

Introduction

Regulatory schemes should be designed to stimulate competition and improve efficiency in the transmission and distribution segment, in the electrical sector organized as a natural monopoly. Yet, in Brazil, the Regulatory Agency for Electricity (ANEEL) uses an ad-hoc scheme to remunerate the firms engaged on electricity transmission. In spite of the huge amounts involved in those reimbursements there is no evaluation of this scheme. Such an analysis is particularly relevant as the access to the transmission networks constitutes the base of a successful liberalization of the electricity markets. However, the presence of sub-additivity of costs, congestion, vertical scale economies with power generation and externalities brought about by the existence of loop flow among transmissions networks make difficult the regulatory design in this sector [1]. To account for these elements, different regulatory schemes have been proposed; they include long-term financial-transmission-right as well as those that incorporate incentive systems which permit to tackle with asymmetric information and strategic behavior. Our paper is inserted in this debate. Its objective is to evaluate the regulatory process in the electricity transmission sector in Brazil by combining contract theory and DEA (Data Envelopment Analysis) calculations. On that account we use a multi-product, multi-input regulatory cost-based framework to assess the performance of 14 firms operating in this segment during the period 2004-2007. We compute C-MDEA (Multiple Data Envelopment Analysis) efficiency and Malmquist productivity indexes to derive financial transfers to the firms and compare them with the ones actually paid by the regulator to those same firms.

Methodology: The Dynamic Yardstick Competition Model and the M-DEA Method

The Dynamic Yardstick Competition model [2] constitutes an extension of the yardstick competition model where the performance of a regulated utility is compared against that of a group of comparable utilities. Consider a principal that delegates the production of p outputs to an agent i (or Decision Making Unit). The agent will transform the input vector1 x∈R+xto produce the outputy∈R+x. Input and output prices are given, respectively by

qxandpi∈R+y; t = 1,...,T. The dynamic regulation model is a blueprint incentive model wi∈R+pqp

巴西电力市场监管

linked to a conditional revenue cap. The DEA-Yardstick is given by:

bti=cti+ρi?[CDEA(yti|wti)?cti]t=1,...,T [1]

where bti is a committed revenue cap, cti is DMU’s actual cost and ρi is an incentive parameter. Here, the regulator estimates a CtDEAcost function [3] for the firms and uses this information to determine revenues during the regulatory period. To avoid the ratchet effect, the DEA cost function is computed for the utility peers thus excluding the DMU analyzed. Hence, the cost efficiency is

CtDEA?i(y|w)=minx,λw?x

s.a.x≥∑∑(λjs?xjs+λi0?xi0)

j≠is=0t?1

y≤∑∑(λjs?yjs+λi0?yi0) [2]

j≠is=0t?1

λ∈Γ(r)

j≠i,s=0,...,t?1

iE0We compute historic efficiency scores,for each firm as:

iiCDEA(y0|w0)E= [3] iiw0?x0i0

The cost function CDEA is generated taking into account the whole set of information, including the ones of the analyzed utility. Assuming that the proportion of the initial inefficiency that the firm should suppress each year is δ, the cost norm becomes: i(1?δ?(1?E0))t

iE0 [4]

δ should be defined to assure that the suppression total of the inefficiency along the regulatory period do not exceed the initial inefficiency for the i-th firm: i(1?δ?(1?E0))T

≥1iE0 [5]

Inserting [4] into [1] we have the dynamic yardstick revenue cap, RYi, that will be used for the analysis of the regulatory schemes in the Brazilian market of electricity transmission in Brazil:

DEA?i?(yti|wti)itCti??R=c+ρ??(1?δ?(1?E0))?ct=1,...,T [6] t?iE0??Yitit

巴西电力市场监管

2.1 The M-DEA Model

As we have a limited number of firms and data is available only from 2004, the restricted size of the data base lead to the well known curse of the dimensionality. Moreover, the methodology uses a historic efficiency for each firm based on information prior to the regulatory contract, to be held constant during the negotiation periods t (t = 1,...,T), This requires the segmentation of the data base in two subsets, the first one representing information before the contract is formalized ( t=1) and the second that refer to the period subject to the regulatory process. This limited information reduces the discriminatory power of the usual DEA calculations. For that reason we used instead an extension of the DEA techniques, the M-DEA methodology [4]. Below, we will briefly describe this approach.

The approach MDEA computes efficiency indexes for different combinations of inputs and outputs. This procedure gives efficiency spectra (frequency distributions) for each DMU, from which efficiency ranking can be extracted, together with confidence intervals. The method identifies the largest sets of Q inputs and P outputs as follows.

1. Choose, sequentially, different subsets of inputs and outputs such as q∈{1,...,Q}e

?Q????qp∈{1,...,P}. As we have ??? possibilities to choose a subset containing q inputs (from a

?Q?Q∑q=1??=2?1?q???total of Q), there are possible choices. Similarly, there are Q

?P?P?=2?1∑p=1??p??? output choices. P

QPΩ≡(2?1)(2?1) scores for all combination of inputs and outputs, 2. Compute DEA

each one corresponding to a specific input output set.

3. Define the final DEA efficiency score, for a given DMU, as the average on ω (1,...,?) from

the computed scores:

C

Data MDEA=∑Cωω=1ΩMDEA/Ω [7]

Firm’s costs, for the period 2004 to 2007 for seven public utilities (federal and state owned companies), six private firms and one public firm that was privatized during this period (CTEEP). Yearly data were taken from BMP (Balancete Mensal Padronizado), a detailed financial statement that concessionaires must deliver periodically to ANEEL. To compute efficiency scores (Eti), we considered only controllable costs, such as personnel, subcontracts,

巴西电力市场监管

rental, equipment and general purchase, which includes all the expenses that companies had to

manage in order to operate and maintain their transmission lines. On the other hand, we

considered as “actual costs” all controllable and not controllable costs (which include financial

expenses, depreciation, taxes, etc), since they are supposed to be completely covered by the

regulatory revenue (RAP). In general, public firms also run electricity production plants and

lower tension transmission lines that are not covered by RAP. Obviously, costs related to these

businesses were not considered. Costs are in 2004 prices. Annual wage costs were deflated by

the Consumer Price Index (IPCA). The other operational expenses were deflated by a

wholesale price index, the IGP-M (Índice Geral de Preços). The other information used in this

paper was taken from public reports, delivered annually by the transmission companies.

Table 1 describes inputs and outputs for electricity transmission. A measure for

transported electricity, that constitutes the output of the transmission firms, is the power of the

line. However, this information is not available. For that reason, we computed a proxy for total

power that take into account topological and technical characteristics of the transmission lines

and use this variable as a measure of transported electricity. A line power is proportional to the

square of its tension (P α V2). We assumed that all transmissions lines have the same physical

configuration, same reactance and phase angle equal to 90º (which results in maximum potency

transmission). Moreover, the transmissions lines have been unitized, that is each circuit = 1Km.

Under those hypothetical conditions, the energy transportation capacity (ETC) for a given firm

is computed as:

ETC = 69KV x 69KV x length of the line 69KV + 88KV x 88KV x length of the line

88KV + ... + 750KV x 750KV x length of the line 750KV.

We also used as outputs the substations’ capacity (sum of transformers ratings) and the network

density. As for inputs, we used in this analysis: total controllable costs, financial costs, the

length of the lines and the quantity of substations. Table 1: Inputs and Outputs – Transmission System – 2004-

Length of the lines Total length of the lines 69 to 750 KV Report

Annual # Substations Total of substations Report

BMP – Total Controllable Costs Operational Expenditures ANEEL

BMP – Financial Costs monetary variations) ANEEL

2Report (Lines) [(tension) x length]

Annual Capacity of the Substations MVA (mega volt -ampere) Report

Annual Network Density Length of the Line /Area Report

巴西电力市场监管

Table 2 presents descriptive statistics for inputs and outputs. Notice the heterogeneity of

the data base in which small and new utilities coexist with huge companies. The high standard

deviations as well as the big differences between minimum and maximum values illustrate this

point.

Table 2: Descriptive Statistics – Inputs and Outputs – Transmission System – 2004-2007

Standard Error

Length of the lines Km 6.800,50 7.096,55 253,00 18.894,00 # Substations 42,29 44,15 0 125,00 Total Controllable Costs (R$x106544,92

Financial Costs (R$x106209,05

2Capacity of Energy Transportation (Lines) V4.064,10

Capacity of Energy Transformation 92.978,00 (Substations)

[(Km) 72,61 Network Density /(Km2x103)]

Results

Efficiency results are summarized on Table 3. As we suppose that during the period analyzed

the scale of production is variable we used the constant returns to scale (CRS) MDEA model.

Notice here, the relative stability of the efficiency scores across the period analyzed, as attested

by the small variation of mean scores. However, such aggregate result is misleading as the

efficient paths are quite different across firms. 2 Indeed, the analysis of Figure 1 shows three

distinct groups of utilities: the new private firms (TSN, ETEO, EATE, ECTE, ETEP, and

EXPANSION), the public ones (CEEE, COPEL, CEMIG, ELETRONORTE, ELETROSUL,

FURNAS, CHESF) and the CTEEP, privatized in 2006. The superior performance of the first

group reflects the substantial cost reductions obtained primarily by subcontracting services

required for the transmission process, which results in lean operational and administrative

structures. Moreover, they provide only electricity transmission thus allowing them to grasp the

benefits derived from specialization.

Table 3: Efficiency Scores - Descriptive statistics - MDEA Efficiency Scores – Constant

Returns to Scale (CRS) -2004/2007

Years Minimum 1º QuartileMedian Mean 3º Quartile Maximum

0,386 0,521 0,586 0,607 0,659 0,833 2004

0,383 0,506 0,566 0,599 0,680 0,843 2005

0,345 0,510 0,580 0,603 0,679 0,831 2006

0,413 0,501 0,604 0,612 0,705 0,845 2007

2 Efficiency results are summarized in Table A-1, Annex I, ranked by the efficiency index of the initial year (E0 -2004).