literature review
bruce
684232.pdf
717
[ Journal of Law and Economics, vol. 58 (August 2015)] © 2015 by The University of Chicago. All rights reserved. 0022-2186/2015/5803-0024$10.00
“Can You Hear Me Now?” Exit, Voice, and Loyalty under Increasing Competition
T. Randolph Beard Auburn University Jeffrey T. Macher Georgetown University
John W. Mayo Georgetown University
Abstract
Competition works only if poorly performing vendors can be punished. The principal vehicle for consumers to discipline ill-performing firms is to switch to alternative providers. But switching is not the only mechanism consumers have to express disapproval. While some unhappy consumers may choose to no longer buy the good or service, other consumers express their disappointment through complaining. This article examines Albert O. Hirschman’s conjecture that as industries become more competitive, consumers’ complaints give way to switching. It offers a simple description of the theoretical relationships among market structure, quality, and complaints. It then utilizes an extensive data set to explore the empirical determinants of consumers’ complaining behavior in the local-exchange telephone industry. These data overcome the problem that complaints can depend on both competition and quality, while competition also presumably affects quality directly. The estimations accommodate this compli- cation and provide considerable support for Hirschman’s conjecture.
[E]conomists have refused to consider that the discontented con- sumer might be anything but either dumbly faithful or outright traitorous (to the firm he used to do business with). (Hirschman 1970, p. 31)
1. Introduction
While theoretical studies of market structure and quality are plentiful, they have produced few general conclusions. This dearth of interpretations has persisted for a variety of reasons. First, quality may not be apparent to buyers prior to use, so information (or the lack thereof) can alter customers’ behavior and the con-
The authors gratefully acknowledge seminar participants at the Institutions and Innovation Con- ference at Harvard Business School and at the Federal Communications Commission (FCC) for helpful comments on earlier manuscript versions. The authors also received helpful comments from Chris Borek, Matthew Demartini, Silke Forbes, J. Bradford Jensen, Nathan Miller, Stanley Nollen,
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
718 The Journal of LAW & ECONOMICS
sequent incentive for the equilibrium provision of quality. Moreover, the rela- tionship among market structure, quality, and product-information availability is itself complex (Jin and Leslie 2003). Second, quality is a multifaceted concept. Goods can vary in both horizontal and vertical quality dimensions, with different consequences for the relationship between competition and product character- istics (Waterson 1989). For instance, whether firms compete in prices or quanti- ties affects the provision of quality and the extent of differentiation (Motta 1993). Third, scale economies in the provision of goods of different qualities can cause market structure to alter cost conditions indirectly, as more concentrated mar- kets may allow greater per-firm outputs. Even the simplest quality-comparison case under monopoly and perfect competition with constant returns creates am- biguity, as the result depends on the relationship between the average willing- ness to pay for quality and the corresponding willingness of a marginal buyer (Tirole 1988). Fourth, industry conditions may alter the intertemporal behavior of firms. For instance, firms may seek to cultivate their reputations in a dynamic context, which leads to a host of quality-investment incentives that are absent in a static context (Kranton 2003). In sum, the relationship between quality and mar- ket structure is likely to vary between industries and to remain largely a matter of empirical analysis (see, for example, Goolsbee and Petrin 2004; Crawford and Shum 2007; Chu 2010).
Yet while empirical analysis may provide a fruitful path, a close consideration of the theoretical developments regarding the relationship between market struc- ture and quality reveals an underspecification that, if more fully developed, holds promising insights. In particular, existing models recognize that firms that fail to offer an acceptable level of quality to consumers need to somehow bear the con- sequences of that failure. It is primarily through consumers’ actions in response to their dissatisfaction that this disciplining is supposed to occur.1 Formal mod- els of the relationship between market structure and quality, however, provide little emphasis on this disciplinary mechanism and instead represent consumers as reacting to any alleged breach of equilibrium expectations by simply switch- ing away from the offending firm. Consumers’ discipline of firms is thought to be least effective under monopoly and presumably increases in effectiveness as market structure atomizes. The salubrious story of competitive markets thus rests largely on the ability and willingness of informed consumers to take actions that discipline ill-behaving firms. For example, if one firm’s price and/or quality is unattractive relative to that of rival firms, consumers convey their displeasure by fleeing.
———— Karok Ray, Marcia Mintz, John Rust, Dennis Quinn, Rob Shapiro, Mike Stern, Scott Wallsten, and Luc Wathieu. The authors appreciate support from the Georgetown Center for Business and Public Policy in the McDonough School of Business; the Institute for Business Innovation at the University of California, Berkeley; and the Stanford Institute for Economic Policy Research. Any errors are at- tributable solely to the authors.
1 The principal alternatives are disciplines imposed by the threat or realization of regulatory or legal actions against a firm providing low quality.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 719
While the theoretical and empirical power of such switching behavior (often termed exit) has received considerable scholarly attention (Farrell and Klemperer 2007; Farrell and Shapiro 1988), another mechanism by which consumers ex- press their dissatisfaction is much less studied. When faced with unexpectedly low quality, many consumers complain—most frequently to offending firms but also commonly to public (or private) bodies providing oversight. For example, Better Business Bureaus in the United States and Canada reported receiving over 918,000 complaints in 2013 (Better Business Bureau 2013). Just in the telecom- munications industry, the Federal Communications Commission (FCC) received more than 450,000 complaints between 2003 and 2006 (Government Account- ability Office 2008).
Despite its prevalence, customers’ complaining behavior has received consid- erably less economic consideration than switching behavior. This lack of scrutiny is not due to either the scarcity or the economic unimportance of complaining but rather is a consequence largely of a practical nature. In particular, while firms routinely receive and process consumers’ complaints, the scrutiny and manage- ment of such complaint data are almost never shared with outsiders. Firms are simply not inclined to publicize their shortcomings. Consequently, the ability of researchers to directly observe and study data on complaints is limited.
In this article, we focus on the complaint process as an essential part of the portfolio of options that consumers use to react to service failures and express their dissatisfaction. In doing so, we emphasize the role that market structure plays as a determinant of complaining behavior—an idea originally outlined by Hirschman (1970) in his construct “exit, voice, and loyalty.” Hirschman posits that there is a negative relationship between the extent of competition and the degree of observed complaining, which arises as dissatisfied consumers in more competitive markets are more apt to switch providers rather than complain to their incumbent providers. We refer to this relationship as the Hirschman con- jecture. Because complaints depend on the quality that firms offer—and quality itself may be endogenously determined by market structure—our principal chal- lenge is to separate changes in complaining behavior that arise from changes in industry structure, as conjectured by Hirschman, from changes in quality (and corollary changes in complaining) that arise endogenously with changes in mar- ket structure.
Our empirical analysis is able to overcome this challenge because of the con- fluence of several fortuitous conditions. First, while firms’ complaint data are typically unavailable, we draw on complaint data compiled by the FCC as part of its regulatory oversight of local-exchange telephone firms. Second, the data span a period in which exogenous market structure changes occurred as a conse- quence of the Telecommunications Act of 1996 (Pub. L. No. 104-104, 110 Stat. 56 [1996]), a major federal law permitting competition in a formerly monopolistic industry. Third, we utilize corollary data compiled by the FCC to control for the influence of service quality changes on complaining behavior separately from the influence of market structure changes on complaining behavior over this period.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
720 The Journal of LAW & ECONOMICS
Our results suggest that the Hirschman conjecture is indeed correct: increasing competition results in decreasing voice (that is, complaints).
The rest of this paper is organized as follows. Section 2 frames the theoreti- cal analysis by providing background and a review of the extant literature on the economic dimensions of consumers’ complaining behavior. It then offers a sim- ple descriptive model of complaining behavior, in which complaining (voice) represents a response to dissatisfaction of an intensity intermediate between suf- fering in silence (staying loyal) and exit (switching). The model highlights the in- fluence of consumers’ costs of switching providers (as will arise with changes in market structure), the costs of complaining, and the role of service quality on consumers’ propensities to complain. Section 3 presents an empirical analysis of the determinants of observed complaints in the US local-exchange telephone in- dustry over the 1999–2006 period. These data are especially useful for our pur- poses, as large market structure changes due to entry and consolidation among telephone service providers were triggered by the Telecommunications Act of 1996. Our empirical results indicate a strong effect of market structure on com- plaining behavior that is independent of any product quality effect. Section 4 of- fers concluding comments and further research suggestions.
2. Theory
2.1. Background and Extant Literature
Hirschman (1970) offers a seminal framework for understanding the role of exit, voice, and loyalty that has been applied in a number of contexts across the economic, political science, management, and marketing domains. At the most general level, Hirschman (1970) seeks to explain the foundational determinants of when and why some disgruntled customers exit, some customers use their voice, and some customers maintain loyalty. While the Hirschman framework has considerable intuitive and general appeal for the study of complaining, the provision of empirical insights arising from this framework has been limited for several reasons. First, capturing the voice of dissatisfaction in a systematic way is often impractical because of difficulties in securing either cross-sectional or time-series complaint data. While virtually all firms collect such data, they are understandably reluctant to share it. Second, most empirical studies examine individual consumers’ characteristics as relevant determinants of complaining behavior but neglect other factors that potentially influence this relationship.2 As a result, more is known about complainers’ characteristics than processes that generate complaints or how these processes relate to industry characteris- tics. Third, a central proposition of the Hirschman framework has hitherto been largely ignored. In particular, a fundamental implication of Hirschman’s analy- sis is that there is a relationship between the extent of marketplace competition
2 For example, marketing research has principally focused on the role of consumers’ characteris- tics on the propensity to complain. See Kolodinsky (1995) for a review.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 721
and the means that consumers use to express their dissatisfaction with a good or service. Hirschman (1970, p. 33) indicates that “[t]he voice option is the only way in which dissatisfied customers or members can react whenever the exit op- tion is unavailable. . . . In the economic sphere, the theoretical construct of pure monopoly would spell a no-exit situation, but the mixture of monopolistic and competitive elements characteristic of most real market situations should make it possible to observe the voice option in its interaction with the exit option.” It is this conjecture of Hirschman’s—namely, that moving from monopoly toward a more atomistic industry structure leads to reduced voice and increased exit—that we propose to test.
Several papers examine consumer dissatisfaction but do not invoke the Hirschman framework per se. On the theoretical side, Gans (2002) develops a model of customers’ choice and switching behavior in response to variation in suppliers’ quality. He finds that the presence of more competitors increases customers’ abilities to switch suppliers in response to poor service, which sub- sequently creates greater competitive pressures to improve quality levels. As in virtually all customer loyalty models, however, the theoretical framework is con- strained by examining only the exit and loyalty (but not the voice) options. On the empirical side, Oster (1980) provides an economic analysis of complaining behavior, examining the determinants of consumers’ complaints about different products filed with the Better Business Bureau in New Haven, Connecticut. The cross-sectional nature of the data unfortunately does not allow the relationship between market structure and complaints to be examined. Andreasen (1985) ex- amines consumers’ propensities to complain using patient survey data on physi- cian care (subjectively chosen to represent a loose monopoly) at a single point in time. Absent both cross-sectional and time-series variation in industry structure, he considers only how individual consumers’ characteristics affect complaint propensities and thus does not explore the Hirschman conjecture. Forbes (2008) uses publicly available passenger complaint data on airline service (namely, flight problems and baggage handling) from the US Department of Transportation to examine the relationships between complaints and firm quality and complaints and the level of expected firm quality. She finds that the number of complaints increases when quality decreases and that complaints are affected by consumers’ expectations about quality. Controlling for actual service levels, the higher the consumers’ expectations of quality, the greater the propensity for consumers to complain.
In summary, although complaining is ubiquitous in many markets and the characteristics of complaining customers have been analyzed in several indus- tries, the basic link between market structure and voice posited by Hirschman (1970) remains largely unexamined. To motivate such a study, we turn to a sim- ple description that characterizes complaining as an intermediate response to dissatisfaction. Even in the simplest contexts, we show that the observed relation- ship between market structure and voice will likely depend on the competitive determination of service quality.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
722 The Journal of LAW & ECONOMICS
2.2. Complaining Behavior, Market Structure, and Exogenous Quality
In the economic analysis of quality, goods or services are generally considered to be of two types. For the first type, consumers are able to observe ex ante the quality of the good or service they are purchasing (namely, search goods or ser- vices). For the second type, quality may be determined only ex post—after pur- chase of the good or service (namely, experience goods or services). Models of search goods and services and experience goods and services provide a number of insights into the incentives (or lack thereof) for firms to provide high quality, as well as the regulatory or private incentive mechanisms that may be employed to promote high-quality offerings (Laffont and Tirole 1993). But these models ignore the prospect that consumers may react to disappointments about quality by complaining. Given the prominence of customer complaints, we examine this question with a simple representation of an experience service in which consum- ers make an initial purchase decision only to potentially discover ex post that the quality of the service purchased is low. This model is perhaps the simplest possi- ble that includes complaining and illustrates the Hirschman conjecture.
Consider a market composed of a large number (N ) of price-taking consumers, each of whom makes a decision whether to purchase a given service. Consump- tion of the service provides benefits that depend both on the consumer’s value or taste for the service (given by her type t) and on her consumption experience— which depends on how well the service works and what actions or recourse she (optimally) selects in response to a service failure. For simplicity, assume that consumers have unit demands for the service (so marginal quantity choice is not analyzed), and there is a single outside composite service.
Each consumer knows her type t, which is a random variable distributed with marginal density f(t) and cumulative density F(t) on the interval [tL, tH]. We in- terpret t as the value that a consumer attaches to successfully consuming 1 unit of the service in question, for which she must pay a price P. This formulation differs from that of Shaked and Sutton (1987), for example, in that differentiated con- sumers do not consume goods of various qualities with certainty but rather at- tempt to consume goods of known characteristics that may, however, completely fail to function. The consumer has income M and utility from the composite ser- vice U(q), where q is the quantity of the composite service consumed. Assume that U is increasing in the composite service q and that the price of the composite service is $1 per unit. If a consumer purchases the service, she obtains a value of t if the service works—an outcome that occurs with a known exogenous probabil- ity θ, where 0 < θ < 1. Thus, 1 − θ represents the probability of service failure, which (at this stage) we interpret as a binary event.3
3 Our model assumes implicitly that failure has the same qualitative effect on each consumer. In particular, failure deprives consumers of some portion of the value of the service. As these values differ between types, the implications of failure are not the same for everyone. An alternative model parameterization would allow for differing degrees of failure, but this would not materially affect the conclusions if the value of functional service is held equal between customers. It would be necessary, however, to respecify the nonpurchase condition. Combining random values and random failure effects introduces substantial complexity without corresponding benefits to insight.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 723
If the service fails, then the consumer responds in one of three ways. First, the consumer may remain loyal (that is, does nothing and suffers in silence, or per- haps just complains to friends). Second, the consumer may formally complain (for example, by filing a complaint with a public regulatory body). Third, the con- sumer may exit (for example, by switching to an alternative vendor). The precise sequence of events in each of these options is not critical. What matters instead is interpreting these actions as having payoffs that are related to consumer types in a sensible way. These three options are thus better understood as shorthand representations for various consumer responses that presumably incorporate se- quential activity. For example, a loyal consumer may engage in informal (low- or no-cost) complaining to neighbors or coworkers. Similarly, a consumer who switches vendors might do so only at the end of a series of actions that begins with informal complaining followed by formal complaining to public oversight bodies, studying market information, and so on. For simplicity, we define a con- sumer’s utility V in the following simple forms:
V = U(M) if the consumer does not buy the service, V = t + U(M – P) if the consumer buys the service and the service works, V = dt + U(M – P) if the consumer buys the service, the service fails, and the
consumer remains loyal, V = bt + U(M – P) – c (where c is the cost of complaining) if the consumer
buys the service, the service fails, and the consumer selects the voice re- sponse, and
V = at + U(M – P) – s (where s is the cost of switching) if the consumer buys the service, the service fails, and the consumer selects the exit response.
We assume that 1 > a > b > d > 0 and that s > c > 0.4 We thus depart slightly from Hirschman (1970, p. 40) that “voice tends to be costly in comparison to exit” and emphasize instead the often significant costs of switching (Farrell and Klemperer 2007).5
The interpretations of these expressions are relatively straightforward. A con- sumer who buys a service that fails may respond in different ways, with the op- timal response depending on the consumer’s type (or value of service). A con- sumer can recapture part of the value attached to the service, although failure is always utility reducing: t + U(M - P) > it + U(M - P) ∀ t, i ∈ {d, b, a}. We interpret consumers’ responses ordered by their degree of aggressiveness, with loyalty considered least aggressive, complaint considered moderately aggressive, and exit considered most aggressive. While more aggressive responses are more
4 Our assumed ordering is predicated on two considerations. First, as a theoretical matter, a > b (in the presence of s > c) is required for switching behavior to be present at all. Second, the assump- tion generates the potential for an observed portfolio of consumer behaviors consistent with our empirical observations. See Section 3. However, one can easily imagine cases in which b > a. The parameterization assumed here is merely the simplest one that generates the observed portfolio of consumer behaviors.
5 We note that in his initial discussion of voice, Hirschman references not competition but mo- nopoly, in which the cost of exit is prohibitive. In other words, the costs of voice and switching depend on market structure. Alternately, one could assume b > a, in which case c > s would be not only allowed but also necessary to generate all three observed behaviors.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
724 The Journal of LAW & ECONOMICS
expensive (s > c > 0), they provide consumers with greater expected recovery (a > b > d) of service value t. Consumers who value a service more may thus find it optimal to respond more aggressively to service failure than consumers who value a service less. This assumption is the most reasonable if consumers’ responses are to be determined solely by a single parameter. It is this sorting that makes our conceptualization informative empirically. In our context, consumers who buy the service and are satisfied—or do not buy the service—neither com- plain nor switch. For consumers who buy the service and experience a failure, complaining is a means to an end. That is, we exclude the case of the rare individ- ual who enjoys complaining for its own sake.6 We instead argue that complaining is an action taken to discipline firms—absent a service failure, no complaining occurs. Similarly, no independent utility arises from switching. We instead argue that switching (and complaining) takes time and subsequently presents opportu- nity costs that reduce the value of the service obtained. In short, complaining and switching both use up time and/or resources that could otherwise be utilized to obtain value from the service. Further, consumers may experience psychological costs from complaining or switching. We incorporate these factors into the cost parameters c and s.
From this basic setup, a consumer of type t will not buy the service whenever
U M U M P t dt bt c at s( ) ( ) ( )max( , , ),> - + + - - -q q1 (1)
so it is sufficient that U(M) > t + U(M - P). For a consumer who buys a ser- vice that subsequently fails, her response is governed by her type and the values of the parameters a, b, d, c, and s. A number of outcomes are possible, although we are guided in our specification by the simple observation that firms generally have some consumers who are satisfied, others who are quietly dissatisfied, others who complain, and still others who switch. Because of the linearity of consumer utilities in consumers’ types, if a consumer of type t prefers to utilize voice in- stead of loyalty or exit instead of voice, then any consumer of higher type would agree with this bilateral judgment. In other words, preferences satisfy the single- crossing property.
We thus complete our specification of restrictions on the value and cost pa- rameters by assuming that there exist values t0, t1, and t2 such that tL < t0 < t1 < t2 < tH and that consumers with types below t0 do not buy, consumers with types between t0 and t1 buy and do not complain (that is, remain loyal) even when ser- vice failure occurs, consumers with types between t1 and t2 buy and complain if service failure occurs, and consumers with types above t2 buy and exit if service failure occurs.
Of particular interest in the parameterization is the restriction that exit requires a higher type than voice. This restriction is equivalent to the requirement that (a - b)/(b - d) > s/c, which has a fairly natural interpretation. The expression a - b measures the additional proportion of value captured by exit compared with
6 We choose not to explore the possibility of a pure consumption value of complaining. We also assume that the benefits of complaining accrue solely to the complaining party.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 725
voice, while the expression b - d makes the same comparison for voice and loy- alty. Their ratio measures the relative additional gain from switching and must be compared with (and smaller than) the ratio of the cost of switching (s) to the cost of complaining (c) (since we take the cost of remaining loyal to be 0). In other words, if switching is cheap relative to complaining, then complaining will not be observed, as consumers will immediately transition from remaining loyal to switching. Only for certain values of the cost and benefit parameters would one observe the pattern of customer responses that we suggest.
Figure 1 provides an illustration of our argument. We include three curves cor- responding to the utility obtained by different responses to failure, along with the cutoff minimal utility below which no service is purchased. Since the utilities are straight lines, their upper envelope is always convex. We interpret complaining as a response to service failure lying between loyalty and exit. The particular inter- section points for t0, t1, and t2 are defined by the equations
t U M U M P
d0 =
- -( ) ( ) , (2)
t c
b d1 =
- (3)
and
t s c a b2
= - -
. (4)
Given these considerations, a simple representation of the extent of complaining and/or switching obtains whereby magnitudes relate to levels of dissatisfaction.
Figure 1. Theoretical model
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
726 The Journal of LAW & ECONOMICS
In any given market, the numbers of customers falling into different categories by the theoretical frequencies are as follows:
total customers = -N F t[ ( )],1 0 (5)
satisfied customers = -qN F t[ ],( )1 0 (6)
dissatisfied customers = - -( ) ( ) ,[ ]1 1 0q N F t (7)
loyal customers = - æ
è ççç
ö
ø ÷÷÷÷-
é
ë ê ê
ù
û ú ú
( ) ( ) ,1 0q N F c d
F t (8)
vocal customers = - - -
æ
è ççç
ö
ø ÷÷÷÷-
æ
è ççç
ö
ø ÷÷÷÷
é
ë ê ê
( )1 q N F s c a b
F c d
ùù
û ú ú , (9)
and
exiting customers = - - - -
æ
è ççç
ö
ø ÷÷÷÷
é
ë ê ê
ù
û ú ú
( ) .1 1q N F s c a b
(10)
The effects of changes in underlying cost parameters, or in product quality θ, on the theoretical frequencies are obtained directly. Our interest focuses mainly on complaints. It is easy to see that the number of complaints rises as s rises, falls as c rises, and rises as (1 - θ) rises. In particular, if we denote the number of dis- satisfied customers that adopt the voice response by π, then we obtain
¶ ¶
= - - -
æ
è ççç
ö
ø ÷÷÷÷ - >
-p q s
Nf s c a b
a b( ) ( ) ,1 01 (11)
¶ ¶
= - - - -
æ
è ççç
ö
ø ÷÷÷÷ - +
æ
è ççç
ö
ø ÷÷÷÷
é
ë ê ê
ù
û
-p q c
N f s c a b
a b f c d
( ) ( )1 1 úú ú < 0, (12)
and
¶ ¶
= - - -
æ
è ççç
ö
ø ÷÷÷÷-
æ
è ççç
ö
ø ÷÷÷÷
é
ë ê ê
ù
û ú ú <
p q
N F s c a b
F c d
0. (13)
Similar expressions are available for all categories of consumers. We are now able to examine the probable effects of changes in competition on consumers’ complaining behavior. For reasons of practicality and correspondence with Hirschman’s conjecture, we focus primarily on the consequences of changes in market structure on the costs of switching providers (s) and on firms’ service quality (θ).
Equation (11) represents the pure Hirschman conjecture in our framework. As a general matter, the cost of switching to another provider is plausibly declining in the degree of market competition for several reasons. First, switching costs are plainly infinite in pure monopoly. Second, greater competition increases the like- lihood that a disappointed buyer is located close to an alternative seller, either
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 727
geographically or metaphorically. (An exception to this general supposition could arise if vendor-specific equipment or operating-system incompatibilities are im- portant.) Third, firms in competitive markets offering subscription services often take steps to make switching cheaper for (rival firms’) customers by, for exam- ple, paying some switching costs or handling the porting of accounts to the new provider. Incumbents can of course also make it more costly for customers to switch to competitive vendors, so the net effects are unclear. Fourth, it is presum- ably easier to leave one seller for another when the first has a small market share and all others collectively have a large market share. While all of these effects are plausible and suggest that more competition can reduce switching costs, it is clear that switching costs may not generally be monotonic in market structure every- where. For example, while an oligopoly must offer lower switching costs than a monopoly, the comparison between oligopoly and atomistic competition is less clear. Hirschman (1970) repeatedly refers to markets exhibiting a combination of monopolistic and competitive elements, so the limiting case of perfect competi- tion may solely be of theoretical interest.
The effect represented in equation (11), however, generally cannot be inde- pendently observed. This is because of the confounding effect represented in equation (13), which illustrates the impact of service quality on complaints. In the simple representation outlined here—and in the commonsense view—com- plaints are less likely when service quality is high, ceteris paribus. However, the relationship between competition and product quality—even of the vertical sort—is theoretically ambiguous. Models in the literature can be parameterized to produce either a positive or a negative relationship between competition and vertical product quality. This ambiguity implies that, for any given industry or market, the nature of the relationship can only be resolved empirically. While it is generally assumed that competition encourages quality and that firms lacking competition often provide mediocre or poor service, this issue remains one that requires practical study.
If an increase in competition increases (average) quality in the market, then the analysis implies that the numbers of complaints will decline. This outcome is because increased competition lowers switching costs (so more consumers switch at the expense of complainers), while higher product quality results in fewer dis- satisfied buyers in every category. A potential difficulty arises, however, when de- creases in market concentration trigger reduced product quality. In this unlikely but not theoretically impossible case, the number of complaints might increase if the effects of reduced quality overwhelm the reduced switching costs. This out- come implies that any conclusion regarding the validity of Hirschman’s conjec- ture on voice and market structure must be based on an analysis capable of iden- tifying and separating these effects.
Finally, it is clearly possible that changes in market structure might also be thought to affect other parameters in the simple analysis given here. We have largely ignored the potential effect of a change in market structure on the costs of using voice, for example. Changes that increase this cost will discourage com-
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
728 The Journal of LAW & ECONOMICS
plaints, which will lead to more switching and more silent suffering. It is not clear what effect market structure might have on these costs, but it is plausible that the existence of some sort of scale economies in customer service could give the mo- nopoly a cost advantage in fielding service calls. Such an effect, if it existed, would work against any finding in favor of Hirschman’s conjecture.
It is therefore apparent that the competitive process adds complexity and idio- syncrasy to the role of customers’ complaints. Changes in the level of competi- tion might alter the costs of voicing complaints and/or switching but at the same time may also increase or decrease product quality, which leads to changes in complaint flows. The empirical challenge then is to tease apart these potentially distinct economic phenomena. It is to this task that we now turn.
3. Empirics
3.1. Empirical Setting
Our empirical setting is the telecommunications industry—a sector whose in- dustrial organization has evolved significantly over time. Along with the 1984 AT&T divestiture, the Telecommunications Act of 1996 represents a significant watershed in the transformation of the industry from a monopoly environment to the more competitive industry of today. In particular, the act represented the first full-throated endorsement of industry competition. The act provided not only a rhetorical embrace of competition but also language that established the means by which new entrants (competitive local-exchange carriers; CLECs) could directly compete against incumbents (local-exchange carriers; LECs) for the patronage of residential and business customers.
The 1996 Telecommunications Act envisioned that new entrants would com- pete as resellers of local-exchange services, purchase unbundled network ele- ments, or become full facilities-based providers of local-exchange telephone service. Given the substantial expense associated with full facilities-based entry, CLECs predominantly entered either as resellers or as purchasers of unbundled network elements in the wake of the act and, consequently, required the use of LEC-owned facilities in order to compete. While general principles to guide new entrants’ access to incumbent LECs’ facilities were part of federal law, the de- tailed implementation of access was left predominantly to individual states. In the first instance, CLECs were required to negotiate with incumbent LECs over the terms and conditions of access. In the event that the parties could not come to terms, individual state public-utility commissions (PUCs) were charged with establishing appropriate rates, terms, and conditions. As CLECs largely failed to reach terms with incumbent LECs in the wake of the act, state PUCs arbitrated access terms on their behalf. Consequently, over the 1996–98 period CLECs were immersed in administrative proceedings instead of marketplace competition, and their combined market share remained under 1 percent. Significant entry into the local-exchange marketplace began in earnest post-1999, however, with CLECs capturing market share vis-à-vis incumbent LECs over the 1999–2006 period.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 729
Three policy elements precipitated by the Telecommunications Act provide a useful setting in which to test the relationship between consumers’ complain- ing behavior and marketplace competition. First, the act represents a relatively clean shift in local-exchange telephone service from a monopoly environment to a competitive environment. This policy change thus offers a quasi-natural- experiment setting in which the impact of industry structure changes on com- plaining behavior can be tested.7 Second, competition that emerged in the wake of the act was far from uniform, as considerable geographic variation in both the number of new entrants and the extent of market share captured by new entrants resulted. We take advantage of not only intertemporal changes in the intensity of competition but also geographic variations in the intensity of competition to isolate the effects of market structure changes on complaining behavior. Third, an enduring feature of the telecommunications industry is its distinct customer types. As business customers are often considerably larger than residential cus- tomers, we are able to test the effects of market structure changes for these two consumer types using separate econometric models.
Finally, two fortuitous characteristics of the data on the telecommunications industry facilitate our empirical analysis. First, the study of complaining behav- ior has been limited by the highly proprietary nature of the data on complaints. Because LECs are overseen by state and federal regulatory agencies, however, the FCC collects data on complaints. Second, and as described in Section 2, the emer- gence of competition can theoretically affect numbers of complaints either di- rectly by altering consumers’ costs for switching providers or indirectly by alter- ing marketplace quality levels. The FCC also collects data on perceptions of LECs’ customer-service quality. These data allow us to identify the effects of changes in quality on complaining behavior separate from the effects of industry structure on complaining behavior as conjectured by Hirschman.
3.2. Data
Our data are drawn from the FCC’s electronic Automated Reporting Manage- ment Information System (ARMIS) filing system.8 We utilize the FCC’s Service Quality Report (Report 43-05) and Customer Satisfaction Report (Report 43-06) in our empirical analysis. Each report spans the years 1996–2006 and contains data on LECs. We utilize the years after 1998 for both conceptual and practical reasons. As described above, new entrants were largely engaged in administrative proceedings rather than marketplace competition during 1996–98. As a practi- cal matter, moreover, the de minimis presence of new entrants resulted in the FCC withholding market-share data in a large number of states, which severely
7 The “quasi” modifier is necessitated by the prospect that changes in market structure and com- petition may generate changes in the quality—in turn, altering changes in propensities to complain. We account for this indirect effect in the empirical analysis.
8 For more information and data available via the Automated Reporting Management Informa- tion System (ARMIS), see FCC, ARMIS Data Descriptions (http://www.fcc.gov/encyclopedia/armis -data-descriptions-1).
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
730 The Journal of LAW & ECONOMICS
truncates the available data over this 2-year period. Finally, while the industry structure has continued to evolve since 2006, the FCC altered the categorization of CLECs in 2007, thereby creating an unfortunate incompatibility with the ear- lier data collected. We accordingly focus our empirical analysis on 1999–2006, the period in which the industry experienced its most prominent changes as competition emerged. The data include observations on all reported incumbent local- exchange telephone carriers in the FCC’s database. Because of corporate re- organizations, mergers, and spinoffs, the data constitute an unbalanced panel of between 172 and 196 companies (observations), depending on the year.
The FCC’s Service Quality Report provides data on customers’ complaints made to state PUCs.9 Once state PUCs receive these complaints, they inform the relevant LECs so that service issues may be resolved. The LECs are required to report these complaints to the FCC on or before April 1 of each year, which tallies them in its ARMIS database. Our focus is on service-quality complaints, which pertain to service, installation, and repair but not to billing, operator service providers, or 900 and 976 numbers (see FCC 2007). The Service Quality Report separates complaints by residential and business customers. Figure 2 displays average reported service-quality complaints across LECs by customer type over 1999–2006. Following an initial spike, average residential customer and business customer complaints decline over this time period. Figure 2 also indicates that average reported residential customer complaints considerably outpace average reported business customer complaints—a relatively unsurprising result given marked size differences in residential customers versus business customers.
The FCC’s Customer Satisfaction Report provides data on customers’ satisfaction levels. The LECs are required to report to the FCC annual customer- satisfaction survey results for residential and business customers based on customer- service and business procedures related to installations, repairs, and business of- fices. We incorporate these data as a proxy for service quality. Figure 3 displays average reported satisfaction levels across LECs by customer type for 1999–2006. Residential customers’ perceived quality of local-exchange telephone service un- dulates, while business customers’ perceived quality of local-exchange telephone service consistently improves after an initial drop over the observed time period.
Annual versions of the FCC’s Local Competition Reports provide data on CLEC competition. While no ideal measure of the extent of competition exists, we utilize two commonly used proxies. First, the reports provide the number of CLECs operating in each state.10 Prior to 2005, the FCC collected data only from those CLECs with at least 10,000 switched access lines in a particular state. Be- ginning in 2005, all CLECs regardless of size were required to report these data.11 Second, the reports provide the aggregate market share held by CLECs in each
9 While complaints are also made directly to the FCC, they represent a small fraction of the com- plaints received. We therefore utilize the more granular data afforded in the complaints made to local regulatory agencies.
10 See FCC, Local Telephone Competition and Broadband Deployment (http://www.fcc.gov/wcb/ iatd/comp.html).
11 We account for this change in our empirical estimation.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 731
state.12 Figure 4 displays the average number of CLECs and the average CLEC market share across states over 1999–2006. The average number of CLECs in- creased from five to nearly 40, and the average CLEC market share increased from 4 to 18 percent over this time period.
3.3. Empirical Strategy and Variables
According to theory, increasing competition may affect complaints both di- rectly by altering the costs for consumers to switch providers (the Hirschman conjecture) and indirectly via changes in service quality that may accompany changes in the level of competition. Our empirical analysis seeks to separately identify these channels. In particular, we seek to determine if, after controlling for changes in quality that may accompany changes in market structure, the separate influence on complaining behavior conjectured by Hirschman exists. To do so, we exploit geographic and temporal differences in competition in the wake of the Telecommunications Act.
12 Because individual competitive local-exchange carrier (CLEC) market shares are competitively sensitive, the FCC reports only aggregate CLEC market shares in each state-year. If only a small number of CLECs are present in a state-year (especially in 1999) the FCC withholds CLEC market- share data but provides data on the number of CLECs. In these instances, we first calculate average CLEC market share in a state-year once aggregate market-share data are reported (say, in 2000) using the data on number of CLECs. We then use this average CLEC market share to backcast ag- gregate CLEC market share in the unreported year using the average CLEC market share and the number of CLECs. This approach allows for a more complete data panel. As a robustness check, we estimated models with the smaller set of raw data provided by the FCC and confirm that no substan- tive differences obtain.
Figure 2. Average yearly number of complaints
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
732 The Journal of LAW & ECONOMICS
We use several variables to conduct our analysis. The term Complaintsijt rep- resents the natural logarithm of the number of complaints received by incumbent LEC i in state j during year t.13 We separately identify and measure complaints for residential (REZ) and small-business (BIZ) customers, and we estimate models separately for these customer types.
We use two measures of competition. The term CLEC Countijt is the natural logarithm of the number of CLECs competing in the area served by incumbent i in state j during year t. Because the FCC reports the number of CLECs at the state (rather than service-territory) level, we derive our competition count measure as the number of CLECs facing an incumbent LEC relative to the largest incumbent LEC operating in the state-year. We thus assume that the number of competitors facing an incumbent LEC is proportional to its share of lines relative to the larg- est incumbent LEC.14 The term CLEC SoMijt represents the market share held by CLECs competing in the area served by incumbent i in state j during year t.
13 We experimented with several alternative specifications of the dependent variable, but no sub- stantive changes to the empirical results reported below obtain.
14 In our measure, CLEC Countijt equals CLEC Countjt × (Linesijt /Max Linesjt), where Linesijt is the number of (residential or business) lines held by incumbent local-exchange carrier (LEC) i in state j at time t and Max Linesjt is the (residential or business) line count of the largest LEC in state j at time t. We also examined two alternative competition-count measures as robustness tests. We first substituted Sum Linesjt (the summation of lines in state j at time t) for Max Linesjt, which effectively proxies for the number of competitors an incumbent LEC faces by its share of total lines in that state-year. We then used CLEC Countjt (the logged statewide count of competitors). Neither of these alternative competition-count measures substantively alters the results or conclusions. For similar measurement approaches to competitive entry, see Abel (2002) and Greenstein and Mazzeo (2006),
Figure 3. Average yearly percentage of satisfied customers
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 733
We account for service quality using three customer-satisfaction measures, which provide the percentage of either residential or business customers sur- veyed that report satisfaction in installations, repairs, and business- office ser- vices for incumbent LEC i in state j in year t. We create a composite residen- tial customer- satisfaction measure (REZ Pct Satijt) and a composite business customer- satisfaction measure (BIZ Pct Satijt) by averaging across the three mea- sures.15
We control for incumbent LEC size using the natural logarithm of the number of residential lines (REZ Linesijt) or business lines (BIZ Linesijt) in service. This approach permits the data to flexibly reveal the relationship of complaint levels to firm size, instead of having it imposed as would occur by using the number of complaints per line as the dependent variable. To account for the FCC’s 2005 change in reporting on competitors, we include a dummy variable (Post 2004t) set equal to one for years 2005 and 2006 and zero otherwise.
Our analysis also includes time-varying state-level measures of population (Populationit), per capita income (Per Capita Incomeit), and the percentage of
which use new-competitor counts, and Economides, Seim, and Viard (2008), which uses new-entrant market share.
15 Slightly less than 3 percent of the residential customer-satisfaction observations (19 of 668) are at 100 percent, while slightly more than 3 percent of the small-business customer-satisfaction obser- vations (22 of 668) are at 100 percent. Accordingly, ceiling effects are not binding for the subsequent empirical analysis. We also experimented with different permutations (for example, the total of the aggregate satisfaction measure) as robustness checks. The results are substantially invariant to those reported.
Figure 4. Average yearly number and market share of competitive local-exchange carriers
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
734 The Journal of LAW & ECONOMICS
Democrat PUCs (Pct Dem PUCit). As described below, these variables serve as first-stage instruments in an analysis that permits the level of competition and the level of consumer satisfaction to be determined endogenously.
3.4. Sample Statistics
Table 1 provides summary statistics for the unlogged dependent and inde- pendent variables. The average annual number of residential customers’ and business customers’ complaints per company, respectively, over the entire sam- ple are roughly 131 and 18. These numbers indicate that residential customers’ complaints are far more prevalent than business customers’ complaints, but sub- stantial heterogeneity is nevertheless observed for each variable. The number of CLECs and CLEC market share also demonstrate significant heterogeneity. Some states have no CLECs or are characterized by limited CLEC market share, while other states have up to 70 CLECs in operation. Residential and business custom- ers similarly demonstrate significant variation in their satisfaction with installa- tions, repairs, and business-office operations, via the aggregate measures used in the baseline empirical estimations.
Table 2 provides correlation statistics for the variables. We find significant pos- itive correlations between numbers of residential and business complaints, be- tween numbers of residential and business lines, and between numbers of (res- idential and business) complaints and (residential and business) lines. There are also moderate negative correlations between numbers of (residential and busi- ness) complaints and (residential and business) aggregate customer-satisfaction levels related to installations, repairs, and business-office operations. The number of CLEC competitors is negatively correlated with the number of both residen- tial and business complaints (although pairwise significance is achieved only for the latter), while the market share held by CLEC competitors is negatively and significantly correlated with the number of residential and business complaints.
Table 1 Summary Statistics
Variable Mean SD Min Max REZ Complaints 130.93 401.83 .00 4,982.00 BIZ Complaints 17.84 60.86 .00 1,096.00 REZ Lines 577,471.50 1,594,666.00 .00 40,200,000 BIZ Lines 282,657.50 706,988.70 .00 6,745,436 CLEC Count 19.75 16.68 .00 70.00 CLEC SoM 12.76 7.13 .00 46.00 REZ Pct Sat .92 .03 .77 1.00 BIZ Pct Sat .91 .03 .77 1.00 Per Capita Income 30,860.02 4,675.17 20,053 57,746 Population 8,449,827.00 7,812,824.00 479,602 36,000,000 Pct Dem PUC .436 .288 .000 1.000
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Ta bl
e 2
C or
re la
ti on
S ta
ti st
ic s
V ar
ia bl
e (1
) (2
) (3
) (4
) (5
) (6
) (7
) (8
) (9
) (1
0) R
EZ C
om pl
ai nt
s 1.
00 B
IZ C
om pl
ai nt
s .8
0* 1.
00 R
EZ L
in es
.4 6*
.4 4*
1. 00
B IZ
L in
es .4
8* .4
8* .7
9* 1.
00 C
LE C
C ou
nt −.
04 −.
06 *
.0 4
.0 7*
1. 00
C LE
C S
oM −.
06 *
−. 06
* .0
1 .0
3 .5
4* 1.
00 R
EZ P
ct S
at −.
36 *
−. 28
* −.
25 *
−. 26
* −.
22 *
−. 12
* 1.
00 B
IZ P
ct S
at −.
33 *
−. 33
* −.
16 *
−. 19
* −.
02 .0
1 .6
4* Pe
r C
ap ita
In co
m e
.0 2
.0 0
.1 1*
.1 4*
.5 8*
.6 1*
−. 21
* 1.
00 Po
pu la
tio n
.0 3
.0 4
.2 0*
.2 7*
.4 1*
.2 4*
−. 20
* .3
5* 1.
00 Pc
t D em
P U
C −.
05 −.
03 .0
1 .0
2 −.
01 −.
22 *
−. 04
.0 2
−. 05
1. 00
* Pa
ir w
is e
si gn
ifi ca
nc e
at p
= .0
5.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
736 The Journal of LAW & ECONOMICS
3.5. Empirical Results
In parallel with the development of the theoretical treatment of the complaint- generation process discussed above, we examine the empirical relationship be- tween industry structure and complaint propensities. Complaints are modeled as a function of the degree of competition, the level of firms’ service quality, and a set of controls. Our initial theoretical specification of consumers’ complaint pro- pensities is conditional on an exogenously determined service-failure process. The basic form of the model estimated is
Complaints CLEC Comp Pct Satijt ijt ijt ijt i ijt= + + + + +b b b d g e0 1 2 X , (14)
where CLEC Compijt is measured as either CLEC Countijt or CLEC SoMijt, Pct Satijt is calculated using an aggregate measure of customer satisfaction (REZ Pct Satijt or BIZ Pct Satijt), Xijt is a vector of controls, β and δ are parameters to be es- timated, γi is the set of state fixed effects, and εijt is a random disturbance term.
Table 3 provides the initial ordinary least squares estimation results separated into residential and business customers.16 Model 1 utilizes CLEC Countijt as the competition measure; model 2 utilizes CLEC SoMijt as the competition measure. The R2-statistics indicate that considerable explanatory power obtains in all esti- mations. Likelihood ratio tests confirm statistically significant explanatory power from the inclusion of state fixed effects in the econometric models.
Several noteworthy insights emerge from the estimations. First, after con- trolling for other variables including the observed levels of quality, the results provide clear support for the Hirschman conjecture. Across both residential and business customers the results indicate that increases in competition nega- tively and statistically significantly (p < .01) influence observed complaint lev- els. This result robustly holds not only for residential and business consumers but also with respect to the measures of competition. Second, while our principal goal is to identify or refute the Hirschman conjecture, the analysis also neces- sarily raises the question of whether changes in competition following the Tele- communications Act led to changes in quality that may, in turn, have affected the observed level of complaints. The results in Table 3 provide some initial in- sights on this issue. In particular, we see that both residential and business cus- tomers’ complaints are negatively and highly statistically significantly (p < .01) correlated with consumers’ perceived levels of quality.17 The inclusion of either REZ Pct Satijt or BIZ Pct Satijt controls for the possible confound in our identifi- cation of the Hirschman effect (were we to not control for observed quality lev- els) and identifies a separate influence on complaining behavior attributable to post-competition changes in quality. Third, we unsurprisingly observe that the
16 Given the potential for correlations in the errors across our residential and business equations, we also estimated the residential and business models via seemingly unrelated regression as a ro- bustness check. The results are inconsequentially different than those reported. We also explored whether nonlinearities from interactions or squaring terms added significantly to the power or in- sights of the model, but these alternatives were found to add little to the results reported here.
17 See Forbes (2008) for a similar result for the US airline industry.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 737
level of complaints is related to firm size but more interestingly find that the elas- ticity of observed complaints with respect to size is consistently less than unity. As the estimations control for consumers’ perceptions of quality, this declining propensity to complain may be due to a sense that larger firms are less responsive to complaints, and, consequently, consumers are less likely to complain for any given level of service failure.
Although the results in Table 3 provide insight into consumers’ complaining behavior, several considerations potentially cloud the inferences from these es- timations. Of particular concern is the possiblity that both the observed level of quality and the level of competition are endogenously determined. As shown in Hörner (2002), Kranton (2003), and Levhari and Peles (1973), service quality may be endogenously driven by changes in competition. Similarly, observed levels of competition may themselves be endogenous to prevailing market conditions. To account for the possible confounds that may arise with such endogeneity, we em- ploy an instrumental variables (IV) approach for the customer-satisfaction and competition variables.18 Our search for satisfactory instruments draws on three prior findings. First, as market size has been shown to be a determinant of com-
18 Durbin-Wu-Hausman tests confirm endogeneity in the residential customer and business cus- tomer estimations, which suggests that an instrumental variables approach is warranted.
Table 3 Ordinary Least Squares Results
Model 1 Model 2
REZ Complaints
BIZ Complaints
REZ Complaints
BIZ Complaints
CLEC Count −.337** (.086)
−.228** (.074)
CLEC SoM −.363** (.077)
−.344** (.070)
REZ Pct Sat −11.612** (1.410)
−12.415** (1.371)
BIZ Pct Sat −8.128** (1.126)
−8.131** (1.250)
REZ Lines .960** (.055)
.769** (.028)
BIZ Lines .673** (.045)
.555** (.022)
Post 2004 .338* (.134)
.058 (.113)
.118 (.090)
−.007 (.079)
Constant 3.766* (1.566)
2.650+ (.365)
6.845** (1.492)
4.128** (1.277)
F-test 68.48** 59.20** 79.51** 62.26** R2 .825 .774 .827 .752 Note. Standard errors (in parentheses) are robust and clustered by firm. All regressions include state fixed effects. N = 668.
+ p < .10. * p < .05. ** p < .01.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
738 The Journal of LAW & ECONOMICS
petitive entry (Abel 2002; Burton, Kaserman, and Mayo 1999; Bresnahan and Re- iss 1991), we expect state-level population and per capita income to correlate with CLEC entry. These variables are not obviously related to variations in complaints, however, which suggests their suitability as instruments. Second, as market size has been shown to be a determinant of product quality (Berry and Waldfogel 2010), we expect state-level population and per capita income to affect endog- enous investments in service quality and, subsequently, customer-satisfaction levels. Finally, especially in regulated industries such as telecommunications, lo- cal regulatory policies as determined in part by the political composition of reg- ulatory commissions may affect competition and customer-satisfaction levels (Smart 1994; Fremeth, Holburn, and Spiller 2012). With these considerations in mind, Populationit, Per Capita Incomeit, and Pct Dem PUCit serve as first-stage instruments.
Tables 4 and 5 provide the two-stage least squares results for the residential customer and business customer models. Table 4 provides the second-stage re- sults that are of primary interest and are very similar to those presented in Table 3. In particular, we find the presence of both direct effects of market structure on complaining behavior and indirect effects on complaints via quality changes. For both competition measures and across residential and business lines, increases in competition result in statistically significant (p < .01) reductions in the level of observed complaining. These findings provide support for the Hirschman con- jecture. Estimations that employ the number of CLEC entrants indicate that the elasticity of complaints with respect to changes in competition is −.70 for resi- dential customers and −.57 for business customers. Holding all other variables at their respective means, we find that a 1-standard-deviation increase in CLEC entry decreases residential complaints nearly 55 percent and decreases business complaints nearly 48 percent. Estimations that employ the market share of CLEC entrants indicate that the elasticity of complaints with respect to changes in com- petition is −.51 for residential customers and −.54 for business customers. A 1-standard-deviation increase in CLEC market share decreases residential com- plaints more than 27 percent and decreases business complaints more than 28 percent, when all other variables are held at their respective means.
Table 5 reports the first-stage results and provides support for the chosen in- struments. Overall, F-tests of the joint significance of the instruments is signifi- cant at p < .01. Subsequent empirical tests provide confidence that the variables employed to correct for endogeneity represent appropriate excluded instruments. Sargan-Hansen tests of overidentifying restrictions are not rejected (Hayashi 2000), which indicates that the excluded instruments are valid. Kleibergen-Paap underidentification tests are rejected (Kleibergen and Paap 2006), which indicates that the instruments are relevant. And redundancy tests are rejected (Breusch et al. 1999), which indicates that the instruments are not redundant.19
19 The specifics of these first-stage results, while not of primary interest, nonetheless are informa- tive. For instance, we find that competition is negatively correlated with population for our mea- sures of competition that are based on the number of firms in the market while being positively
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 739
3.6. Discussion
The consistent and strong evidence of a negative relationship between the level of competition and the propensity to complain provides empirical evidence con- sistent with the conjectures described in Section 2. While the estimations do not distinguish whether reductions in the propensity to complain result in more cus- tomers moving into the loyal or the exit category, the logic of our theory suggests that the emergence of competitive alternatives reduces the costs associated with switching and thereby shrinks the category of complainers and expands the cat- egory of switchers. Telephone industry data also support this general proposi- tion, as the number of incumbent LECs’ residential and business lines declined by over 43 million during the 2000–2006 period. This decline is widely attributed to a combination of customers switching to newly emergent CLECs, switching to wireless carriers, and reducing their number of lines (FCC 2008).
We also find that competition-induced changes in quality exert an indepen-
related to competition measures based on CLEC market share. This is consistent with the earlier en- try by the largest CLECs (AT&T and MCI) that resulted in the capture of significant market shares, while smaller and more numerous CLECs targeted less populous states. For the count measure of competition, we also find, consistent with Abel (2002), that the political orientation of public-utility commissions also impacted the number of entrants.
Table 4 Second-Stage Two-Stage Least Squares Results
Model 1 Model 2
REZ Complaints
BIZ Complaints
REZ Complaints
BIZ Complaints
CLEC Count −.700** (.161)
−.573* (.272)
CLEC SoM −.507** (.090)
−.538** (.203)
REZ Pct Sat −26.679** (10.076)
−19.090* (8.768)
BIZ Pct Sat −20.124* (10.095)
−5.989 (12.428)
REZ Lines 1.089** (.104)
.737** (.050)
BIZ Lines .790** (.188)
.566** (.065)
Post 2004 .721** (.243)
.614* (.264)
.167 (.120)
.091 (.081)
Constant 17.103 (10.301)
12.983 (11.191)
13.800 (8.915)
2.209 (12.199)
F-test 37.99** 36.74** 64.22** 61.65** R2 .766 .711 .817 .776 Note. Standard errors (in parentheses) are robust and clustered by firm. All regressions include state fixed effects. N = 668.
* p < .05. ** p < .01.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Ta bl
e 5
Fi rs
t- St
ag e
T w
o- St
ag e
L ea
st S
qu ar
es R
es ul
ts
M od
el 1
M od
el 2
C LE
C
C ou
nt R
EZ
Pc t S
at C
LE C
C
ou nt
B IZ
Pc
t S at
C LE
C
So M
R EZ
Pc
t S at
C LE
C
So M
B IZ
Pc
t S at
Po pu
la tio
n −2
.5 07
** (.6
89 )
.0 18
(.0 68
) −1
.8 76
** (.7
23 )
.0 36
(.0 90
) 1.
64 7*
(.6 81
) .0
18 (.0
68 )
1. 64
8* (.6
81 )
.0 36
(.0 90
) Pe
r C
ap ita
In co
m e
1. 20
8+ (.6
32 )
−. 15
8* *
(.0 50
) 1.
01 1
(.6 98
) −.
12 9*
* (.0
47 )
−. 39
1 (.6
31 )
−. 15
8* *
(.0 50
) −.
39 2
(.6 31
) −.
12 9*
* (.0
47 )
A vg
D em
P U
C −.
23 4*
(.1 06
) −.
00 5
(.0 08
) −.
26 1*
(.1 11
) −.
00 4
(.0 07
) −.
03 4
(.0 75
) −.
00 5
(.0 08
) −.
03 4
(.0 75
) −.
00 4
(.0 07
) T
im e
T re
nd .1
33 **
(.0 23
) .0
05 **
(.0 02
) .1
24 **
(.0 26
) .0
07 **
(.0 02
) .2
31 **
(.0 20
) .0
05 **
(.0 02
) .2
31 **
(.0 20
) .0
07 **
(.0 02
) R
EZ L
in es
.5 54
** (.0
15 )
−. 00
5* *
(.0 01
) .0
00 (.0
06 )
−. 00
5* *
(.0 02
) B
IZ L
in es
.5 15
** (.0
14 )
−. 00
5* *
(.0 01
) .0
00 (.0
06 )
−. 00
5* *
(.0 01
) Po
st 2
00 4
.7 11
** (.0
51 )
−. 00
4 (.0
04 )
.6 70
** (.0
57 )
.0 01
(.0 04
) −.
31 4*
* (.0
41 )
−. 00
4 (.0
04 )
−. 31
4* *
(.0 41
) .0
01 (.0
04 )
C on
st an
t −2
50 .9
76 **
(3 8.
04 9)
−8 .1
76 *
(3 .1
87 )
−2 39
.5 83
** (4
3. 31
2) −1
2. 18
7* *
(3 .2
03 )
−4 79
.6 95
** (3
3. 86
1) −8
.1 76
* (3
.1 87
) −4
79 .7
50 **
(3 3.
80 6)
−1 2.
18 7*
* (3
.2 03
) F-
te st
18 6.
21 **
17 .7
0* *
13 2.
74 **
21 .1
1* *
69 .8
0* *
17 .7
0* *
69 .9
1* *
21 .1
1* *
R 2
.9 06
.3 99
.8 97
.3 60
.8 36
.3 99
.8 36
.3 60
N ot
e. S
ta nd
ar d
er ro
rs (i
n pa
re nt
he se
s) a
re r
ob us
t a nd
c lu
st er
ed b
y fir
m .
+ p
< .1
0. *
p <
.0 5.
** p
< .0
1.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 741
dent effect on complaint levels.20 In particular, we find that observed levels of changes in quality as captured by variations in REZ Pct Satijt or BIZ Pct Satijt are negatively and statistically significantly related to complaint levels.21 Consistent with previous research (Forbes 2008), our results indicate that higher levels of consumer satisfaction are associated with lower complaint levels.
4. Conclusion
Competition works only if poorly performing vendors can be punished. In most markets, the principal vehicle for consumers to discipline ill-performing firms is to switch to alternative providers of the good or service. Considerable and appropriate attention has accordingly been given to the magnitude of switching costs that customers face. Switching is by no means the only mechanism consum- ers have to express disapproval. While some unhappy customers may choose to no longer buy the good or service, other consumers express their disappointment through complaining or, as Hirschman (1970) terms it, using voice. Although Hirschman describes the many roles that voice plays, his analysis of its applica- tion incorporates a famous conjecture concerning its relationship with market structure. Noting that under monopoly complaining is the sole means for buyers to express dissatisfaction, Hirschman suggests that the use of voice declines as markets become more competitive. Although entirely plausible and inherently interesting, Hirschman’s conjecture has not previously been rigorously tested.
As a matter of practice, large-scale studies of complaining behavior have been limited by the fact that firms do not readily provide data on consumers’ com- plaints. For regulated industries, however, it is often the case that one key func- tion of regulatory oversight bodies is to receive and process customers’ com- plaints and to adopt appropriate public-policy responses to them. By drawing on a large-scale database of complaints recorded by the FCC regarding local- exchange service in the United States, we have been able to explore the relation- ship between complaint levels and industry structure. The empirical analysis pro- vides strong support for the Hirschman conjecture.
While data on complaints are necessary to test Hirschman’s conjecture, there 20 As a robustness check, we also created quality variables representing the summation of the per-
centage of satisfied consumers across the installation, repairs, and business-office categories. Using these variables in place of the average consumer-satisfaction variable produces similarly negative and statistically significant coefficient estimates in the residential and business models.
21 In separate regressions, we also sought to identify any direct influence of competition on our two quality measures. These regressions used our quality measures as a dependent variable, with both competition (number or market share) and a set of controls as independent variables. The re- sulting estimations provide considerable (but not unanimous) support for the intuitive proposition that increases in quality led to higher levels of satisfaction. In particular, the estimations indicate a positive and statistically significant effect from CLEC entry (CLEC CNTijt) on the percentage of res- idential customers reporting satisfaction (p < .01) and a positive and statistically significant effect from CLEC entry on the percentage of business customers reporting satisfaction (p < .01). Results for the estimations are more mixed when competition is measured by the market share of CLECs, with a statistically insignificant effect from CLEC market share on the percentage of residential cus- tomers reporting satisfaction but a positive and statistically significant effect from CLEC market share on the percentage of business customers reporting satisfaction (p < .05).
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
742 The Journal of LAW & ECONOMICS
is an additional and important complication. A large literature in industrial eco- nomics suggests several ways in which market structure and competition could be linked to endogenous product quality. Product quality, however, is known to significantly affect complaining behavior in the expected way. Changes in market structure may therefore affect complaint levels through the quality channel but also through changes in the use of voice, as suggested by Hirschman. Our analy- sis incorporates this complication—an approach made possible by the availability of standardized indices of consumer satisfaction—so that quality can, in an ap- proximate sense, be measured directly.
Our findings are strongly supportive of Hirschman’s insight. Using both single- equation and IV techniques on a large unbalanced panel of telephone-service complaint data, we find evidence that increasing competition reduces recourse to voice, holding the quality of the underlying service constant. These results ob- tain for both residential and small-business customers. Further, as documented in myriad other related literature, service quality matters, with higher quality re- sulting in fewer complaints, ceteris paribus.
While our results provide supportive preliminary evidence on the nature of the relationship between competition and voice, they also suggest the possibil- ity of additional explorations. Several potential refinements of the theoretical models are readily apparent and may yield additional insights. For example, it seems plausible that as the intensity of competition increases, the extent of inter- nal complaint mechanisms utilized by firms may evolve (Fornell and Wernerfelt 1988). This effect might be best captured by making the effectiveness of informal complaints (that is, complaints to the firm rather than a public oversight body) a positive function of industry fragmentation. In this case, a more sophisticated model that permits firms to optimize across public and private complaints may provide insights into both complaints and complaint-management processes that are not considered here.
While we have focused on the relationship between complaints and market structure, our empirical results convey a relationship between customers’ satis- faction and complaining that is worthy of additional consideration. While firms are ultimately interested in the level of customers’ satisfaction with their goods or services, the level of satisfaction is often not directly observable—firms most typi- cally simply observe that customers do or do not complain. Our analysis suggests, however, that a slip of considerable and varying size may exist between the cup of satisfaction and the lip of complaints. The simultaneous presence of satisfaction and complaint data may consequently afford a more detailed investigation into this relationship than has heretofore been possible.
Our empirical results similarly raise several managerial and public-policy con- siderations. Consider how individual firms assess data that they receive from complaining customers. While firms may be tempted to draw inferences regard- ing improved customer service or quality from shorter queues of complaining customers, such inferences may prove unwarranted. For any given level of sat- isfaction, our results indicate customers’ reduced propensities to complain as
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 743
competition grows. Customers are instead apt to move more quickly from loy- alty to exit, bypassing voice completely, as the number of competitors increases. Firms with internal complaint mechanisms in place—but without sophisticated customer-retention metrics—simply cannot conclude that they are doing better as the number of customer complaints falls. In terms of public policy, we sug- gest that while monopolistic industry structures gave rise to the establishment of public complaint mechanisms in many regulated industries, the emergence of competition in the telecommunications industry increasingly gives customers the ability to express discontent quite apart from that which they can express to reg- ulators. In particular, more competition increasingly allows customers to avoid the burdens of making complaints—instead allowing them to rely on the market- place alternative of exit to punish ill-behaving firms. In the face of such ultimate punishment, the merits of public complaint mechanisms are likely to diminish.
Finally, while our theoretical results and corresponding empirical analysis seek to advance understanding of relationships among market structure, quality, and complaints, the applicability of our results in other contexts is worthy of addi- tional exploration. Our theoretical results stem in part from various simplifying assumptions, while our empirical results are set in a single industry. Additional theoretical and empirical research may reinforce or provide additional insights into the robustness of our results.
References
Abel, Jaison R. 2002. Entry into Regulated Monopoly Markets: The Development of a Competitive Fringe in the Local Telephone Industry. Journal of Law and Economics 45:289–316.
Andreasen, Alan R. 1985. Consumer Responses to Dissatisfaction in Loose Monopolies. Journal of Consumer Research 12:135–41.
Berry, Steven, and Joel Waldfogel. 2010. Product Quality and Market Size. Journal of In- dustrial Economics 53:1–31.
Better Business Bureau Serving Metropolitan New York, Long Island, and the Mid-Hudson Region. 2013. 2013 Annual Report. BBB Serving Metro NY, New York. http://www.bbb .org/globalassets/local-bbbs/new-york-ny-24/media/documents/annual-report-2014 -final.pdf.
Bresnahan, Timothy F., and Peter C. Reiss. 1991. Entry and Competition in Concentrated Markets. Journal of Political Economy 99:977–1009.
Breusch, Trevor, Qian Hailong, Peter Schmidt, and Donald Wyhowski. 1999. Redundancy of Moment Conditions. Journal of Econometrics 91:89–111.
Burton, Mark L., David L. Kaserman, and John W. Mayo. 1999. Modeling Entry and Barri- ers to Entry: A Test of Alternative Specifications. Antitrust Bulletin 44:387–420.
Chu, Chenghuan Sean. 2010. The Effect of Satellite Entry on Cable Television Prices and Product Quality. RAND Journal of Economics 41:730–64.
Crawford, Gregory S., and Matthew Shum. 2007. Monopoly Quality Degradation and Regulation in Cable TV. Journal of Law and Economics 50:181–219.
Economides, Nicholas, Katja Seim, and V. Brian Viard. 2008. Quantifying the Benefits of Entry into Local Phone Service. RAND Journal of Economics 39:699–730.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
744 The Journal of LAW & ECONOMICS
Farrell, Joseph, and Paul Klemperer. 2007. Coordination and Lock-In: Competition with Switching Costs and Network Effects. Pp. 1967–2072 in vol. 3 of Handbook of Indus- trial Organization, edited by Mark Armstrong and Robert H. Porter. Amsterdam: North-Holland.
Farrell, Joseph, and Carl Shapiro. 1988. Dynamic Competition with Switching Costs. RAND Journal of Economics 19:123–37.
FCC (Federal Communications Commission). 2007. FCC Report 43-05—Report Defini- tion. http://www.fcc.gov/wcb/armis/documents/2007PDFs/4305c07.pdf.
———. 2008. Trends in Telephone Service. Washington, DC: FCC Wireline Competition Bureau, Industry Analysis and Technology Division. https://apps.fcc.gov/edocs_public/ attachmatch/DOC-284932A1.pdf.
Forbes, Silke J. 2008. The Effect of Service Quality and Expectations on Consumer Com- plaints. Journal of Industrial Economics 56:190–213.
Fornell, Claes, and Birger Wernerfelt. 1988. A Model for Customer Complaint Manage- ment. Marketing Science 7:287–98.
Fremeth, Adam R., Guy L. F. Holburn, and Pablo T. Spiller. 2012. The Impact of Consumer Advocates on Regulatory Policy in the Electric Utility Sector. Public Choice 161:157–81.
Gans, Noah. 2002. Customer Loyalty and Supplier Quality Competition. Management Sci- ence 48:207–21.
Goolsbee, Austan, and Amil Petrin. 2004. The Consumer Gains from Direct Broadcast Satellites and the Competition with Cable TV. Econometrica 72:351–81.
Government Accountability Office. 2008. FCC Has Made Some Progress in the Manage- ment of Its Enforcement Program but Faces Limitations, and Additional Actions Are Needed. Report to the House of Representatives Committee on Energy and Commerce Subcommittee on Telecommunications and the Internet Chairman. Washington, DC: Government Accountability Office.
Greenstein, Shane, and Michael Mazzeo. 2006. The Role of Differentiation Strategy in Lo- cal Telecommunication Entry and Market Evolution: 1999–2002. Journal of Industrial Economics 54:323–50.
Hayashi, Fumio. 2000. Econometrics. Princeton, NJ: Princeton University Press. Hirschman, Albert O. 1970. Exit, Voice, and Loyalty: Responses to Decline in Firms, Orga-
nizations, and States. Cambridge, MA: Harvard University Press. Hörner, Johannes. 2002. Reputation and Competition. American Economic Review
92:644–63. Jin, Ginger Zhe, and Phillip Leslie. 2003. The Effect of Information on Product Qual-
ity: Evidence from Restaurant Hygiene Grade Cards. Quarterly Journal of Economics 118:409–51.
Kleibergen, Frank, and Richard Paap. 2006. Generalized Reduced Rank Tests Using the Singular Value Decomposition. Journal of Econometrics 133:97–126.
Kolodinsky, Jane. 1995. Usefulness of Economics in Explaining Consumer Complaints. Journal of Consumer Affairs 29:29–54.
Kranton, Rachael E. 2003. Competition and the Incentive to Produce High Quality. Eco- nomica 70:385–404.
Laffont, Jean-Jacques, and Jean Tirole. 1993. A Theory of Incentives in Procurement and Regulation. Cambridge, MA: MIT Press.
Levhari, David, and Yoram Peles. 1973. Market Structure, Quality, and Durability. Bell Journal of Economics 4:235–48.
Motta, Massimo. 1993. Endogenous Quality Choice: Price vs. Quantity Competition. Jour-
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Competition and Consumers’ Complaints 745
nal of Industrial Economics 41:113–31. Oster, Sharon. 1980. The Determinants of Consumer Complaints. Review of Economics
and Statistics 62:603–9. Shaked, Avner, and John Sutton. 1987. Product Differentiation and Industrial Structure.
Journal of Industrial Economics 36:131–46. Smart, Susan R. 1994. The Consequences of Appointment Methods and Party Control for
Telecommunications Pricing. Journal of Economics and Management Strategy 3:301–23. Tirole, Jean. 1988. The Theory of Industrial Organization. Cambridge, MA: MIT Press. Waterson, Michael. 1989. Models of Product Differentiation. Bulletin of Economic Re-
search 41:1–28.
This content downloaded from 134.053.121.170 on October 29, 2018 12:10:01 PM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
