ARTICLE IN PRESS
Int. J. Production Economics 92 (2004) 133–144
Price cap regulation, incentives and quality: The case of Brazilian telecommunications Luis Ota! vio Fa@anha, Marcelo Resende* Instituto de Economia, Universidade Federal do Rio de Janeiro, Av. Pasteur 250, Urca 22290-240, Rio de Janeiro-RJ, Brazil Received 18 June 2003; accepted 14 October 2003
Abstract This paper reviews the incentive properties of yardstick schemes with special reference to quality performance and to the economic foundations and practical applications of data envelopment analysis (DEA). The core of the paper is related to the construction of synthetic relative efficiency scores for quality, which were obtained for Brazilian local telephony over the period 1998–2002. The telecommunications industry in Brazil has been characterized by the recent implementation of price cap regulation. The evidence indicates substantial quality underperformance, with some improvements towards the end of the period. The efficiency scores obtained, based on the flexible DEA approach, suggest that adjustment factors should be used in the productivity offset of the price cap rule so as to provide premiums for superior quality performance in terms of lower X factors. The DEA-based yardstick procedure can provide a practical alternative to monitoring a series of specific quality indicators in terms of a set of penalties for underperformance. The composite quality indicators do not rule out the identification of optimal adjustments recommended for individual indicators. r 2003 Elsevier B.V. All rights reserved. Keywords: Price cap; Quality; Yardstick competition
1. Introduction The efficiency-inducing properties of regulatory regimes have been the object of limited investigation (see Kridel et al., 1996; Resende, 1999, 2000; Vogelsang, 2002, Sappington, 2002; Uri, 2003a). The empirical literature has mainly focused on the efficiency performance of incentive-based regulation vs. traditional rate-of-return regulation.
*Corresponding author. Tel.: +55-21-3873-5258; fax: +5521-2541-8148. E-mail addresses:
[email protected] (L.O. Fa@anha),
[email protected] (M. Resende).
Telecommunications utilities are increasingly subject to some form of incentive regulation. In particular, one observes the growing popularity of price-cap regulation (PCR) as a result of its alleged efficiency-inducing properties, offering strong incentives for cost-reduction efforts as extra rents can be appropriated. The proponents of price caps also emphasize possible decreases in prices and increases in production levels when the caps are introduced in competitive environments. A possible shortcoming of PCR, however, is that it may worsen quality performance (see Vickers and Yarrow, 1988). Laffont and Tirole (2000) point out that if quality increases are accompanied by increases in costs, the quality of
0925-5273/$ - see front matter r 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpe.2003.10.015
ARTICLE IN PRESS 134
L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
services may deteriorate. Rovizzi and Thompson (1995) stressed that the incorporation of quality sensitive factors in the price control formula leads to additional regulatory and monitoring costs (see also Lynch et al., 1994). This would jeopardize the simplicity of the price-control method under PCR. Empirical investigations of the assumptions regarding quality deterioration are also limited. Relevant references, for the US, include Curry et al. (1991), Kraushaar (1995), Roycroft and Garcia-Murrilo (2000), Ai and Sappington (2002), Banerjee (2003) and Uri (2003b). For the UK, a relevant work is Rovizzi and Thompson (1995). In any case, the establishment of the X productivity offset in PCR might eventually require quality correction factors. It is known, furthermore, that quality monitoring practices prevail in different cases.1 Quality assessments necessarily define a multidimensional problem. It is important, therefore, to go beyond the measurement and analysis of single quality indicator, as is the usual practice in the literature. Though recognized in the literature, little empirical investigation has been performed incorporating complexity of quality assessment (see e.g., Berg and Lynch, 1992). To do so, one must first identify the influential variables to be taken into account and, second, determine the relative importance of the individual factors in an overall service-quality indicator. The paper is organized as follows. Section 2 outlines the institutional and regulatory features of Brazilian telecommunications. Section 3 describes the relative efficiency measurement approach that is used in the paper and explores the related theoretical foundations. The section explores the economic foundations of DEA by means of a principal-agent framework and the possible links between yardstick compensation schemes and price cap regulation (PCR). Section 4 discusses the links between yardstick competition (YC), PCR and service-quality. Section 5 presents the data construction procedures and the empirical results that were obtained. Section 6 offers some final comments. 1 According to Littlechild’s report on privatisation and regulation of water industries (1983, apud Powell and Szymanski, 1997): ‘‘regulation would need to be permanent; covering quality as well as price’’.
2. Brazilian telecommunications: Institutional and regulatory background Only a few years ago, utilities in Brazil were publicly owned. In the case of telecommunications, the system was organized around the holding company Telebra! s, which was comprised of a local municipal operator, a long-distance operator (Embratel) and state telephone companies. The system experienced rapid growth in the 1970s, but afterwards began to show signs of exhaustion due to reduced investment capacity. The necessity of expanding and modernizing the sector prompted privatization initiatives. In fact, the privatization of the system occurred in July 1998 with a sale that reached the amount of US$ 19 billion (see Pires, 1999; Mattos, 2002; Resende, 2003, for discussions of the evolution of the Brazilian telecommunications system and related regulatory issues). In the post-privatization period, one could observe the emergence of a new institutional and regulatory framework centered around an independent regulatory agency [the Ag#encia Nacional * de Telecomunica@oes—ANATEL]. The newly created system was intended to stimulate competition. In the case of fixed telephony, a duopoly structure was conceived for both local and long distance telephony. For local telephony, the country was divided into three regions (with each regional market divided between two companies), while for long distance authorities determined that competition would take place on a nationwide basis. The duopoly would be formed by the regionally (regrouped) privatized firms (the socalled concessionaires) and a second firm created to compete with the first (the so-called mirror companies).2 Moreover, concessionaires are in 2 Introducing competition in the different regional markets did not render the expected results, as recognized by ANATEL officials in print media [see O Globo, 09/21/2001, p. 26]. Part of the model failure can be attributed to the large dimension and heterogeneity of Brazil, where few high-traffic segments are likely to attract a second competitor. In the case of longdistance, one has a duopoly structure gradually subject to entry from local operators. In that segment, competition appears to be strong as indicated by intense price and non-price competition.
ARTICLE IN PRESS L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
135
Table 1 Productivity offsets—Brazil (1998/2005) Tariff components
1998
1999
2000
2001
2002
2003
2004
2005
Local calls Long-distance calls International calls Local access Long-distance access
0 2 5 0 2
0 2 5 0 2
0 2 15 0 2
1 4 15 5 4
1 4 15 10 4
1 4 15 15 4
1 5 15 20 5
1 5 15 20 5
Source: ANATEL.
principle forced to grant universal access and meet quality targets whereas the mirror companies do not have to meet similar requirements. In the case of cellular telephony, a duopoly was initially considered, but in terms of 10 different regions. In this case, there is significant potential for competition.3 The most significant achievement of the new system appears to be the rapid expansion and modernization of the telecommunications network. In fact, the number of telephone access lines increased enormously over the last few years. In particular, the penetration of cellular telephony is evident in all segments of Brazilian society, a result of the widespread use of pre-paid cell phone cards. It is worth mentioning, however, some negative aspects. First, even though access to telephone facilities became cheaper and faster, tariff levels are still high for low-income consumers. Second, telephone companies often failed to meet quality targets, as demonstrated by information provided by ANATEL and by the huge volume of complaints from consumer defense organizations. Concerning regulation, the Telebra! s system was characterized by traditional rate-of-return regulation with the associated poor efficiency-inducing features. The new regulatory framework put forward by ANATEL included price cap schemes where the productivity offset (the X factor) is
gradually increased over time. The programmed evolution of that factor is summarized in Table 1. Table 1 shows that the price cap is gradually becoming significant in terms of larger X factors. In Brazil, all concessionaires are subject to the same form of regulation, whereas in the US local operators are subject to different regulatory regimes, depending on the state they are located in. This allows for interesting regulatory comparisons (see Resende, 2000). In Brazil, regime comparisons are only meaningful over time and for changes in specific aspects of the PCR scheme. The present paper will address service-quality measurements for local operators in Brazil. In Section 3, we deal with the related methodological aspects.
3. Methodology 3.1. Data envelopment analysis (DEA): Basic technical aspects An influential approach to efficiency measurement was proposed by Charnes et al. (1978). The model can be briefly described as follows:4 There are m inputs (indexed by the subscript i), s outputs (indexed by subscript r) and n decisionmaking units (DMUs, indexed by subscript j); additionally one assumes that xij > 0 and eyrj > 0 4
3
The focus of the present paper is on fixed telephony, for more details on the cellular segment see Resende (2003). We should also point out that cellular telephony does not seem to exert significant competitive pressure on fixed telephony, as the tariff levels of the former are much higher than those of the latter.
Though DEA models constitute a mathematical programming technique, they possess economic foundations. In fact, Bogetoft (1997, 2000) derives its properties from a principal– (multi) agent model and explores incentive properties in the context of yardstick competition. Sections 3.2 and 4 will briefly explore those foundations. For more details on DEA models see Cooper et al. (2000).
ARTICLE IN PRESS L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
136
denote non-negative inputs and outputs, respectively. CCR consider the following optimization problem: s m X X maxu;v hk ¼ ur yrk = vi xik ð1Þ r¼1
i¼1
subject to: ur X0 ðfor r ¼ 1; y; sÞ
vi X0
ðfor i ¼ 1; y; mÞ; ð2Þ
s X r¼1
ur yrj =
m X
vi xij p1 ðfor j ¼ 1; 2; y; k; y; nÞ:
i¼1
ð3Þ The problem must be solved for each DMU. The goal is maximize virtual output relative to virtual input,5 subject to the constraint that no DMU can operate beyond the efficiency frontier (constraint (3)) and the constraint relating to nonnegative weights (constraint (2)). In this paper, we will focus on the behavior of total efficiency as obtained from the CCR model.6 Inputs and outputs weights are endogenously determined as the solution to the problem for each DMU. It can be shown that the previous model can be represented in terms of a linear programming problem to be solved for each DMU. The resulting scores are relative efficiency measures where a score of 1 indicates an efficient unit and scores less than 1 indicate inefficient units. Inefficient units are identified by comparison with reference units (peers). These peers are identified by a data-driven procedure where inefficient units are compared to the most similar efficient units. In other words, if there is an efficient DMU that has similar levels of inputs as an inefficient DMU under analysis, then the latter could increase levels of outputs with the current levels of inputs. The former DMU would be considered a reference peer for the inefficient DMU. If a unit often appears as a peer then its 5
This corresponds to the output orientation whereas the input orientation would consider the minimization of the virtual input relative to a given virtual output. The two notions are equivalent only in the constant returns to scale case. 6 Banker et al. (1984) and Banker (1984) extend the CCR model from the constant returns to scale to the variable returns to scale case.
condition as an efficiency reference is reinforced. Therefore, it is desirable that the involved units be comparable. On the other hand, it is worth mentioning that the efficiency frontier might include very different units in terms of input and output levels. It is also important to emphasize that the model imposes few restrictions on the production set. One convenience of this is that one does not need to assume a direct transformation of the postulated inputs into the chosen outputs.7 The present paper will consider efficiency frontiers for quality indicators in the context of Brazilian local telephony. The use of DEA models for efficiency assessments is somewhat scarce in telecommunications. Examples include Majumdar (1995), Resende (2000), Uri (2001) and Resende and Fa@anha (2002). These authors consider conventional inputs and outputs and regulatory regimes comparisons in some cases. The flexibility of the DEA approach can be particularly useful for quality assessments as it allows us to simultaneously consider multiple dimensions of servicequality. As we shall see in Section 5, it is possible to interpret some quality indicators as inputs and some as outputs in the context analysis. That possibility was explored by Resende and Fa@anha (2004) for the case of US local telephony with the purpose of comparing regulatory regimes. Evidence was found demonstrating the worsening of quality performance under PCR as reflected by the relative efficiency quality scores. The objective of the referred approach is to obtain a synthetic quality indicator that incorporates the various dimensions of the problem with a sensible (and endogenously determined) weighting scheme upon the individual service-quality indicators. In fact, according to Berg and Lynch (1992): ‘‘Regulatory agencies monitor numerous dimensions of quality, but collapsing these into a single index represents a useful complement to current utilization of passfail indicators’’. The proposed approach contrasts with the previous literature on the topic. This literature 7 DEA attempts to capture non-allocative inefficiencies that are possibly associated with managerial wastes. The implied xinefficiency can be properly assessed by means of DEA models as defended by Frantz (1988, 1992).
ARTICLE IN PRESS L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
can be broadly divided into three groups: (i) aggregate descriptive analysis for a few selected quality indicators (Curry et al., 1991; Kraushaar, 1995; Rovizzi and Thompson, 1995); (ii) methodological works that attempt to justify weighting schemes for specific quality indicators taking as reference expert evaluations in order to summarize the indicators into a single quality index (Berg and Lynch, 1992; Lynch et al., 1994); and (iii) econometric studies to explain individual quality indicators separately (Roycroft and Garcia-Murrilo, 2000; Ai and Sappington, 2002; Banerjee, 2003; Uri, 2003b). The aforementioned literature recognizes the need for comparative quality evaluations that motivate the adoption of some variant of yardstick practice to improve regulatory procedures. Thus, before proceeding further, it is important to briefly explore the economic foundation of the DEA approach that will help in the construction of relative quality performance indicators. These considerations will be addressed in the next few sections. 3.2. Yardstick competition: Some basic aspects It is well known that traditional rate-of-return regulation does not provide incentives for cost reduction. An intuitive alternative for bypassing that shortcoming can be provided by yardstick-YC procedures where comparable units are regulated with reference to more efficient units.8 Early . (1982) treatments of YC appear in Holmstrom and Shleifer (1985).9 YC is based on relative efficiency assessments and it is therefore important to be confident on the optimality of best practice norms. Fortunately, cost norms based on data envelopment analysis (DEA) can be shown to be second-best in an asymmetric information setting. Specifically, one can justify the following reimbur8
In principle, one can control for different types of heterogeneities by considering adjustments by means of econometric methods. As long as there are spatially separated non-colluding units, yardstick methods can be feasible. Moreover, it is important that spillovers between cost reducing activities across firms are not present, see e.g. Bogetoft (1997), Powell and Szymanski (1997) and Dalen (1998). 9 For further theoretical developments on yardstick competition see Dalen (1998) and Sobel (1999).
137
sement rule (see Bogetoft, 1997, 2000): bi ðX ; Y Þ ¼ ci þ ri ðcDEA ðYi Þ ci Þ;
ð4Þ
where ci denotes the actual cost of unit i; Yi represents the related demand (multiple outputs are allowed) and ri the value of slack relative to profit. In other words, this last parameter shows the extent to which the unit will be rewarded for reducing costs below the DEA cost norm, which would indicate costs savings. This cost norm can be shown to be equal to the product of the DEA efficiency score and the actual cost of unit i:10 The underlying motivation for this reimbursement rule can be summarized as follows. From the perspective of the DMU, one intends to maximize the utility accruing from the reimbursement minus the disutility of effort. An information asymmetry arises because the DMU has private information on its true cost structure that depends on an intrinsic efficiency (adverse selection) parameter as well as on an effort parameter. The regulator, on the other hand, seeks to minimize the expected value of reimbursement subject to participation and incentive compatibility constraints. The former constraint assures that DMUs utility exceeds a minimum (reservation) value. The latter restriction guarantees that the reimbursement rule is attractive when the DMU chooses the most efficient production plan with the implied minimum costs. More generally, as already pointed out, such a condition is imposed to limit informational rents obtained by the DMU when it knows the true parameters of the cost function and the regulator only knows their probability distributions (examples include the intrinsic efficiency, service-quality and effort parameters). The informational rent arises when the contract allows the DMU to declare a lower efficiency parameter and a higher cost structure than it actually has. In the DEA-based yardstick scheme, it is assumed that cost is observable and the regulator can take advantage of DMU’s preference for profit. That is ‘‘1$ of slack costs the regulator $1, but it only gives the agent r $ of profit. The same utility can be 10
Bowlin et al. (1989) had suggested a similar DEA scheme in the context of budgeting without considering more refined theoretical foundations as done by Bogetoft (1997).
ARTICLE IN PRESS 138
L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
given to the DMU if the regulator spends r $ paid out in profits’’. The reimbursement rule is in fact the solution to the previously outlined problem. This illustrates that the flexible mathematical programming approach of DEA has sound economic foundations and can constitute the basis for a yardstick scheme, as is evident in the DEA-based cost norm that appears in the expression referred to above. Yardstick schemes-YC can be incorporated within PCR as recognized in the literature (see Sappington, 2002). In fact, YC can provide guidance to the choice of the productivity offset X ; as the expected efficiency growth can be approximated in terms of an external norm. From the point of view of the present paper, an important challenge is to establish the relation between DEA-based reimbursement rules and price cap schemes.
4. Yardstick, regulation and quality 4.1. Yardstick schemes and price cap regulation The DEA-based reimbursement rule can be generalized to a multi-period setting, as shown by Agrell et al. (2002). The authors use the following expression: bit ðYt Þ ¼ ct þ R0 c0 þ rðcot Et cot Þ;
ð5Þ
where ct denotes total cost for period t; ðR0 c0 Þ indicates the gross margin (R0 is the base year revenue), r defines the power of the reimbursement rule (as previously mentioned), cot denotes operational cost at period t and Et refers to the DEA efficiency score. As previously, cot Et represents the DEA-based cost norm.11 11
The model proposes that one should consider superefficiency when calculating Et : This means that efficiency is computed for all the units excluding the unit under analysis. This procedure might imply efficiency scores superior to 1 and therefore the term rðcot Et cot Þ should be substituted by gðR0 c0 Þ; with g indicating a profit margin. It should be mentioned that the determination of both r and g is still a controversial issue depending on the bargaining between the regulator and the regulated firm.
The expression above defines a revenue cap for the regulated firm. In regulatory practice, similar DEA-based schemes are used in Scandinavian countries.12 Nonetheless, the existing literature does not provide much guidance with respect to the choice of the productivity offset X : Some guidelines for determining that factor are provided by Bernstein and Sappington (1999, 2000, 2001). The bulk of the analysis simply decomposes profit variation into its components and imposes a longrun zero profits condition. The resulting expression shows that price cap mimics the competitive behavior in the sense that the difference between the growth of input and output prices should reflect total factor productivity growth. It is essential, however, to guarantee the incentive properties of PCR by considering a forward looking X factor. Otherwise, PCR would resemble traditional rate-of-return regulation with the associated shortcomings. Moreover, the authors have already stressed the ideal conditions for a proper functioning of PCR, that may serve to delineate guidelines for determining the appropriate value for the X factor: (i) all of the regulated firm’s services are subject to PCR; (ii) no major structural changes are expected in the industry under PCR; (iii) the rate of price inflation outside the regulated factor is not affected by the pricing decisions of the regulated firm; and (iv) the economy outside the regulated sector is competitive. In the Brazilian case, which is the focus of the present paper, no special concerns arise with respect to conditions (i) and (ii). In fact, the whole service basket is subject to PCR and no important structural changes have occurred under PCR in terms of strengthening competition. With respect to condition (iii), the large weight of utilities prices in the general price index as well as the strong influence of regulated prices over the prices of final goods imply a substantial endogeneity in the economy-wide rate of inflation. Corrective measures would constrain price increases in terms of a tougher X factor. Finally, condition (iv) can be a special concern as oligopolistic factors are pervasive in the Brazilian economy and only recently have competition 12
See Agrell et al. (2002, p. 15).
ARTICLE IN PRESS L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
policies become more visible and active. This would provide, therefore, an additional motivation for a higher X factor. The choice of the X factor in actual regulatory practice is always an elusive issue. On the other hand, yardstick schemes can potentially provide a useful guidance to the determination of such productivity offset and the previous discussion tried to make salient the sound theoretical foundations of DEA-based yardstick comparisons. Actually, we can conceive the establishment of the X factor taking as reference relative efficiency scores obtained from the DEA approach. In the case of multiple periods, F.are et al. (1994) proposed Malmquist productivity indexes based on DEA scores. One powerful result of this technique is that total factor productivity growth can be decomposed in terms of two components: the first reflecting changes in relative efficiency (catch-up component) and the second reflecting technical change.13 The second component can be relevant for orienting the choice of X ; as long as the pace of future technical progress is somewhat similar to past behavior. In any case, this technical change component should be a better proxy for the X factor than the catch-up component, which tends to be more backward-looking. In Section 4.2, we take advantage of the previous relative efficiency measurement for service-quality and the conceptual discussion of DEA-based yardstick schemes to study the role of quality adjustment factors in the price cap rule. 4.2. Quality and regulation The issue of quality in the context of specific market structures can be traced back to the seminal contributions of Levhari and Peles (1973) and Spence (1975, 1976). The former points out that market structure does not exert any effect on quality, whereas Spence compares quality under monopoly and perfect competition from 13
Loosely speaking, the Malmquist index for a given DMU considers a geometric average of distances between output and input vectors (in the base year and for the current year) relative to the two periods best-practices norms. See Coelli et al. (1998) for technical details.
139
the standpoint of welfare evaluations. A main result that emerges is the tendency of quality undersupplying under monopoly relative to a competitive environment (assuming both have the same cost structure and output level). In fact, the social planner would select quality based on its effect on all consumers (average quality) and the monopolist considers the impact of an increase in quality taking as reference the marginal consumer’s willingness to pay for quality.14 Another strand of the literature identifies quality with durability through the analysis of durable goods monopolies. Examples include Kihlstrom and Levhari (1977) and Liebowitz (1982), though the issue of service-quality in the context of utilities that produce non-durable goods is not the focus of the analysis. Despite the strong concerns with quality degradation under PCR, quality provision in specific regulatory regimes has not been intensely investigated. A recent theoretical contribution on regulation of quality under YC is provided by Tangera( s (2002). In this model, the firms (hospitals, universities) supply unpriced and non-regulated quantities with total production cost that depends positively on the efficiency parameter and the quality investment and negatively on effort. The regulator observes cost but does not know the referred parameters. Nonetheless, he is able to obtain information (that later becomes common knowledge) on a synthetic quality indicator that will stochastically operate as a quality signal for consumer perceptions. That is to say, expected demand is entirely governed by perceived quality, suggesting that is important that the regulator improves quality measurement procedures. The regulator designs an optimal incentive mechanism based on yardstick comparisons and one of the main results is that optimal effort increases as investments in quality increase. The causal effects run from quality improvements to greater cost savings and returns to scale and effort. Thus there is a link among investments in quality, declines in marginal production costs and increase in productivity. 14 A more detailed discussion of these models appears in Tirole (1988).
ARTICLE IN PRESS 140
L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
Analyses of quality under PCR appears in Weisman (2002), who studied a risk-neutral regulated firm subject to a price cap and with a service-quality level depending stochastically on the investment in quality. The optimal quality investment that emerges from the expected profit maximization generates some salient comparative static results: (i) weaker price caps lead to higher investment in quality; (ii) the larger the extent to which the firm participates in competitive complementary markets the greater the incentive for quality improvement; (iii) penalties for quality underperformance lead to a reduction of investment in quality when they are based on revenues; (iv) the greater the information dissemination on quality performance, the greater the incentive to increase investment in quality. Having briefly discussed the theoretical literature, it is important to summarize the main results obtained from the empirical literature. Roycroft and Garcia-Murrilo (2000) considered the impact of state-level competition and regulation on quality as measured by aggregate trouble reports indicators defined for holding companies. Among other results, the authors found no support for the assumption of quality degradation under PCR. Ai and Sappington (2002) and Banerjee (2003) investigated the impact of incentive regulation at a state-level on different quality indicators (considered one by one) and found no evidence of quality deterioration. Uri (2003b) studied the effect of the introduction of PCR in interstate access services (in 1991) on several individual measures of service quality. The evidence suggested quality deterioration. To mitigate quality underperformance, the author suggests a scheme that resembles the mechanisms discussed in Section 4.1. He proposes the creation of an ‘‘aggregate measure of service-quality–ASQ’’ and a price cap formula with a specification: PCIt ¼ GDP2PIt Xt þ pASQt : This is a standard price cap formula with a quality adjustment term. In the present paper, we argue that DEA-based overall quality indicators can provide useful ingredients for quality adjustments in price cap schemes. For this purpose, relative efficiency measurements for service-quality and related yardstick schemes can be useful.
The yardstick schemes outlined before rely on DEA-based cost norms. It is difficult, however, to conceive a specific norm that pertains to quality since disaggregating common costs is a demanding task. A peculiar trait of quality assessment is the existence of limits to quality improvements. These limits render the analysis somewhat distinct from the case of general efficiency improvements (as discussed in the context of the determination of factor X ). Actual regulatory practice, to the best of our knowledge, does not usually make quality adjustments in price cap rules, but establishes targets for specific quality indicators and related penalties for underperformance.15 In the Brazilian case, quality premiums on the X factor were considered for a specific electricity distribution utility.16 Particularly, a quality adjustment factor scales down the X factor in the sense that higher quality performance implies a smaller X factor. A serious shortcoming, however, is that quality performance is measured in terms of an arbitrary weighted average of three specific quality indicators. A natural extension of our approach would be to establish ranges for the relative efficiency scores for quality and assign premiums that increase the X factor for the better performing units. This suggestion still has an exploratory status and we hope that it can motivate further developments on the topic. Section 5 provides an empirical application for Brazilian local telephony in the 1998–2002 period by proposing a DEA-based relative efficiency measurement for quality indicators. 5. Empirical analysis 5.1. Data construction ANATEL, Brazil’s regulatory agency for telecommunications, was created after the privatization process in July 1998. That agency provides 15 In the Brazilian case, the regulatory agencies recently had difficulty in actually receiving values related to fines, as those were repeatedly challenged in court. This can provide an additional motivation for developing synthetic quality indicators that offer guidance to quality adjustments in the price cap rule. 16 For a discussion of this case and related prospects for yardstick schemes, see Resende (2002).
ARTICLE IN PRESS L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
monthly indicators for different aspects of service– quality in fixed telephony. We considered quality indicators generated by the General Plan for Quality Targets (Plano Geral de Metas de Qualidade (PGMQ)/ANATEL). The referred indicators are similar to those provided by the Federal Communications Commission for the US, but due to omissions we had to focus on a few selected indicators for the period July 1998–March 2002. We list these indicators below:17 Outputs: *
*
REDS: rate of effective dialing signal with minimum waiting time of 3 seconds; RCLC: rate of completed calls referring to local calls-morning. Inputs:
*
*
*
NRR: number of repair requests per 100 telephone lines; NPUB: number of repair requests of public telephones per 100 telephone lines; NCOM: number of complaints per 1000 issued telephone bills.
The logic underlying the previous list of variables is that ‘‘outputs’’ should reflect specific quality dimensions that one would be willing to increase whereas ‘‘inputs’’ refer to negative quality aspects that one is seeking to reduce. This logic for assigning inputs and outputs in terms of ‘‘negatives’’ and ‘‘positives’’ is used in the DEA literature by Retzlaff-Roberts (1997). As mentioned above, the usefulness of the DEA approach in this context is that it does not require a transformation of traditional inputs into outputs. 5.2. Empirical results The relative efficiency scores for quality indicators are presented in Table 2. The heterogeneity 17 In the case of actual practical implementation, the set of relevant quality indicators could be defined by a panel of experts. Typical quality variables in recent studies by Roycroft and Garcia-Murrilo (2000), Resende and Fa@anha (2002, 2004) and Uri (2003b) include number of initial and repeated trouble reports, average installation time, percent of installation commitment met, average repair time for access services and several types of complaint intensity variables.
141
Table 2 Efficiency scores for quality indicators—local telephony in Brazil Firms
July 1998
March 2002
TELERJ TELEMIG TELECOM (MG) TELEST TELERGIPE TELEASA TELPE TELPA TELRN TELECEARA TELEPISA TELMA TELEPARA TELEAMAPA TELEAMAZON TELEAIMA SERCOMTEL TELECOM (MS) TELEMAT TELEGOIAS TELECOM (GO) TELEBRASILIA TELERON CRT CTMR TELESP CETERP TELECOM (SP) CBTCAMPO Mean Coefficient of variation
0.478 0.712 0.626 0.558 0.448 0.944 0.543 0.716 0.507 0.462 0.607 0.658 0.447 0.268 0.636 0.833 0.632 1.000 0.474 0.728 0.749 1.000 0.341 0.222 0.489 0.582 1.000 1.000 0.861 0.639 0.344
0.398 0.638 0.729 0.596 0.580 0.742 0.528 0.822 0.761 0.646 0.712 0.472 0.355 0.412 0.593 1.000 0.644 1.000 1.000 0.852 1.000 0.837 0.651 0.513 1.000 0.672 1.000 1.000 0.614 0.716 0.283
Note: firm that acted as reference peer the largest number of cases.
among different concessionaires is striking (for example, the coefficient of variation is large) and many units are well below the efficient level (only 4 units in 1998 and 7 units in 2002 appeared as efficient). The results suggest that service-quality in Brazilian telecommunications is an important issue and therefore cannot be ignored in the context of regulatory practices. Another aspect worth exploring concerns the inter-temporal comparison of efficiencies. In fact, as stressed before, the magnitude of the productivity offset was increased over the considered period and therefore we should investigate the impact of this increase on quality performance. An
ARTICLE IN PRESS 142
L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
interesting related result from Table 2 is the increase in the mean efficiency level from 1998 to 2002.18 This result is somewhat striking since it contrasts with the traditional view that quality would decrease under PCR. Moreover, as already mentioned, there is little competition in Brazilian telecommunications. Furthermore, the sector was regulated in terms of a uniform price cap scheme that became tighter over time, as shown in Table 1. It is worth mentioning that the DEA approach indirectly captures improvements in organizational practices that are associated with regulatory changes and could be responsible for quality improvements. Though the assessment of quality changes following regulatory shifts is not entirely conclusive, this is not the focus of the present paper. Our concern, rather, is with raising the possibility of conceiving regulatory schemes that reward superior quality performance.19 The efficiency scores for quality obtained in this work could form the basis for a quality-adjusted price cap formula, as suggested in the previous section.20 The DEA approach generated a synthetic quality indicator for each DMU. If the regulator is interested in monitoring performance according to specific quality indicators, the referred approach enables us to identify critical quality indicators determining overall quality efficiency. To elaborate on the results of Table 2, we present in Table 3 the average adjustment of inputs and outputs (by concessionaires) required to lead units to an efficient performance. For example, RCLC is a particularly important indicator of overall performance. On average, a large positive adjustment is needed in that variable (simultaneously with the required adjustments in the other
Table 3 Average input and output adjustments (%) necessary to achieve the efficiency frontier Quality indicators
July 1998
March 2002
NRR (input) NPUB (input) NCOM (input) REDS (output) RCLC (output)
19.62 2.09 3.60 80.52 106.38
8.93 1.05 0.67 52.31 62.30
indicators) to reach overall efficiency (106.38% in July 1998 and 62.30% in March 2002). It is worth mentioning that the adjustments in inputs and outputs are generated for each DMU. The aggregate analysis presented here has an illustrative purpose. In all cases, the required adjustments decline from 1998 to 2002. Additionally, the necessary adjustments for the quality indicators that were employed as inputs are small or moderate. This contrasts with the analogous adjustments for the quality indicators that served as outputs. Although one can observe declining necessary adjustments, they are still extremely high in March 2002. These results reinforce the perception that qualityperformance remains poor in Brazilian telecommunications. Moreover, they show that it is possible to identify the sources of inefficiency with the DEA approach. As explained in the previous section, DEA-based efficiency scores can serve as a starting point for devising adjustment factors in the productivity offset of a price cap formula.
6. Final comments 18
The privatization process took place in July 1998. Though price cap regulation was introduced, residual effects of previous rate-of-return regulation could have had an impact on the beginning of the sample period. 19 A comparison of relative efficiency performance in quality under different regulatory regimes in US telephony is considered by Resende and Fa@anha (2002, 2004). 20 In order to further explore differences in quality-performance across the two periods, we undertook the nonparametric test for two related samples (Wilcoxon signed-rank test). The resulting statistic was z ¼ 1:944 (p-value=0.046) indicating that there are significant differences in the efficiency patterns across periods.
In this article we reviewed the foundations for yardstick schemes as an alternative orientation for incentive regulatory mechanisms. In particular, we emphasized the importance of constructing relative efficiency measures based on the flexible approach of DEA. The empirical analysis for Brazilian local telephony showed that although some overall improvement could be detected between 1998 and 2002, quality underperformance is pervasive. That period was associated with
ARTICLE IN PRESS L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
stable competition and an increasing productivity offset. Prima facie, this quality improvement sounds counterintuitive if compared to the recent theoretical literature. In fact, investment in quality is believed to be positively affected by a more lenient cap. We should recognize, however, that investment in quality has a stochastic impact on actual quality performance. The paper also explored the possibility of defining quality corrections in the X factor, taking as a reference DEA-based quality relative efficiency scores. This would benefit the firms with better quality performance. However, our proposal is mainly exploratory and is intended to stimulate further research on the topic. One possible avenue for research is an econometric investigation of the determinants of service-quality taking into account the role of heterogeneities (for example, regional heterogeneities). A second would involve additional work on the qualityperformance of different regulatory regimes, building on Resende and Fa@anha (2004). Finally, we tried to integrate different strands of the literature on YC. Nevertheless, further developments are necessary in order to provide regulators with tools that can be used in practice.
Acknowledgements The authors would like to thank Luis Henrique Garcia and Bruno Caselli for their research assistance and also for the comments received from two anonymous referees, but the usual caveats apply. Financial support was generously provided by CNPq.
References Agrell, P.J., Bogetoft, P., Tind, J., 2002. DEA and dynamic yardstick competition in Scandinavian electricity distribution. Mimeo. Ai, C., Sappington, D.E.M., 2002. The impact of state incentive regulation on the US telecommunications industry. Journal of Regulatory Economics 22, 133–160. Banerjee, A., 2003. Does incentive regulation ‘‘cause’’ degradation of retail telephone service quality? Information Economics and Policy 15, 243–269.
143
Banker, R.D., 1984. Estimating the most productive scale size using data envelopment analysis. European Journal of Operational Research 17, 35–44. Banker, R.D., Charnes, A., Cooper, W.W., 1984. Some models for estimating technical and scale inefficiencies and data envelopment analysis. Management Science 32, 30–44. Berg, S.V., Lynch, J., 1992. The measurement and encouragement of telephone service quality. Telecommunications Policy 16, 210–224. Bernstein, J.I., Sappington, D.E.M., 1999. Setting the X factor in price-cap regulation plans. Journal of Regulatory Economics 16, 5–25. Bernstein, J.I., Sappington, D.E.M., 2000. How to determine the X in RPI-X regulation: A user’s guide. Telecommunications Policy. 24, 63–68. Bernstein, J.I., Sappington, D.E.M., 2001. Corrigendum to ‘How to determine the X in RPI-X regulation: A user’s guide’. Telecommunications Policy 25, 537. Bogetoft, P., 1997. DEA-based yardstick competition: The optimality of best practice regulation. Annals of Operations Research 73, 277–298. Bogetoft, P., 2000. DEA and activity planning under asymmetric information. Journal of Productivity Analysis 13, 7–48. Bowlin, W.F., Wallace, J.R., Murphy, R.L., 1989. Efficiencybased budgeting. Journal of Cost Analysis 8, 33–54. Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efficiency of decision making units. European Journal Operational Research 2, 429–444. Coelli, T., Rao, D.S.P., Battese, G.E., 1998. An Introduction to Efficiency and Productivity Analysis. Kluwer Academic Publishers, Boston. Cooper, W.W., Seiford, L.N., Tone, K., 2000. Data Envelopment Analysis: A Comprehensive Text with Models, Applications, References and DEA-Solver Software. Kluwer Academic Publisher, Boston. Curry, R.L., Kraushaar, J.M., Gryb, R.M., Ake, J.R., Buzas, T.E., Berg, S.V., Lynch, J.G., Cole, L.P., 1991. Service quality. In: Cole, L.P. (Ed.), After the Breakup: Assessing the New Post-AT&T Divestiture Era. Columbia University Press, New York, pp. 235–284. Dalen, D.M., 1998. Yardstick competition and investment incentives. Journal of Economics & Management Strategy 7, 105–126. F.are, R., Grosskopf, S., Norris, M., Zhang, Z., 1994. Productivity growth, technical and efficiency change in industrialized countries. American Economic Review 84, 66–83. Frantz, R.S., 1988. X-Efficiency: Theory, Evidence and Applications. Kluwer Academic Publishers, Boston. Frantz, R.S., 1992. X-efficiency and allocative efficiency: What have we learned. American Economic Review 82, 434–438. . Holmstrom, B., 1982. Moral hazard in teams. Bell Journal of Economics 13, 324–340. Kraushaar, J., 1995. Quality-of-service measurement and the federal communications commission. In: Lehr, W. (Ed.), Quality and Reliability of Telecommunications
ARTICLE IN PRESS 144
L.O. Fa@anha, M. Resende / Int. J. Production Economics 92 (2004) 133–144
Infrastructure. Lawrence Erlbaum Associates Publishers, Mahwah, NJ, pp. 187–211. Laffont, J.J., Tirole, J., 2000. Competition in Telecommunications. MIT Press, Cambridge, MA. Levhari, D., Peles, Y., 1973. Market structure, quality and durability. Bell Journal of Economics and Management Science 4, 235–248. Liebowitz, S., 1982. Durability, market structure and new-used goods model. American Economic Review 72, 816–824. Lynch Jr., J.G, Buzas, T.E., Berg, S.V., 1994. Regulatory measurement and evaluation of telephone service quality. Management Science 40, 169–194. Mattos, C.C., 2002. O Modelo Brasileiro de Reforma das * Telecomunica@oes. D.Sc. thesis, Universidade de Bras!ılia. Majumdar, S.K., 1995. X-efficiency in emerging competitive markets: The case of US telecommunications. Journal of Economic Behaviour and Organization 26, 129–144. Pires, J.C.L., 1999. A reestrurura@*ao do setor de * telecomunica@oes no Brasil. Revista do BNDES 6, 187–214. Powell, K., Szymanski, S., 1997. Regulation through comparative performance evaluation. Utilities Policy 6, 293–301. Resende, M., 2000. Regulatory regimes and efficiency in US local telephony. Oxford Economic Papers 52, 447–470. Resende, M., 2002. Relative efficiency measurement and prospects for yardstick competition in Brazilian electricity distribution. Energy Policy 30, 637–647. Resende, M., 2003. Brazilian telecommunications: Experience and future perspectives. In: Madden, G. (Ed.), World Telecommunications Markets: International Handbook of Telecommunications Economics, Vol. 3. Edward Elgar, Cheltenham, pp. 319–346. Resende, M., Fa@anha, L.O., 2002. Privatization and efficiency in Brazilian telecommunications: An empirical study. Applied Economics Letters 9, 823–826. Resende, M., Fa@anha, L.O., 2004. Price cap regulation and service–quality in telecommunications: An empirical study. Information Economics and Policy 16, forthcoming. Retzlaff-Roberts, D., 1997. A data envelopment analysis approach to discriminant analysis. Annals of Operations Research. 73, 299–321.
Rovizzi, L., Thompson, D., 1995. The regulation of product quality in the public utilities. In: Bishop, M., Kay, J., Mayer, C. (Eds.), The Regulatory Challenge. Oxford University Press, Oxford, pp. 336–357. Roycroft, T.R., Garcia-Murrilo, M., 2000. Trouble reports as an indicator of service quality: The influence of competition, technology and regulation. Telecommunications Policy 24, 947–967. Sappington, D.E.M., 2002. Price regulation and incentives. In: Cave, M., Majumdar, S., Vogelsang, I. (Eds.), Handbook of Telecommunications Economics. Elsevier Science, Amsterdam. Shleifer, A., 1985. A theory of yardstick competition. RAND Journal of Economics 16, 319–327. Sobel, J., 1999. A re-examination of yardstick competition. Journal of Economics & Management Strategy 8, 33–60. Spence, M., 1975. Monopoly, quality and regulation. Bell Journal of Economics 6, 417–429. Spence, M., 1976. Product differentiation and welfare. American Economic Review 66, 407–414. Tanger(as, P.P., 2002. Regulation of cost and quality under yardstick competition. Working Paper No. 573, Research Institute of Industrial Economics. Tirole, J., 1988. The Theory of Industrial Organization. MIT Press, Cambridge, MA. Uri, N.D., 2001. The effect of incentive regulation on productive efficiency in telecommunications. Journal of Policy Modeling 23, 825–846. Uri, N.D., 2003a. The adoption of incentive regulation and its effect on technical efficiency in telecommunications in the United States. International Journal of Production Economics 86, 21–34. Uri, N.D., 2003b. Service quality effects of incentive regulation on access service in telecommunications in the United States. European Journal of Law and Economics 16, 369–390. Vogelsang, I., 2002. Incentive regulation and competition in public utilities markets: A 20-year perspective. Journal of Regulatory Economics 22, 5–27. Weisman, D.L., 2002. Price Regulation and Quality. Kansas State University, Mimeo.
