Journal of Competition Law and Economics

Journal of Competition Law and Economics Advance Access originally published online on August 28, 2008
Journal of Competition Law and Economics 2009 5(2):361-381; doi:10.1093/joclec/nhn025

This Article Abstract Full Text (PDF) All Versions of this Article: 5/2/361    most recent nhn025v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Request Permissions Google Scholar Articles by Bolotova, Y. Articles by Miller, D. J. Search for Related Content Related Collections K21 - Antitrust Law L10 - General Social Bookmarking          What's this?

© The Author (2008). Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org

FACTORS INFLUENCING THE MAGNITUDE OF CARTEL OVERCHARGES: AN EMPIRICAL ANALYSIS OF THE U.S. MARKET

Yuliya Bolotova ^*, John M. Connor ^** and Douglas J. Miller 

Correspondence: E-mail: yuliyab{at}uidaho.edu

JEL: K21, L10

	ABSTRACT

TOP ABSTRACT I. INTRODUCTION II. THEORETICAL AND EMPIRICAL... III. DATA DESCRIPTION AND... IV. EMPIRICAL MODELS AND... V. ESTIMATION RESULTS VI. CONCLUSION

 
Using the overcharge estimates for 333 cartel episodes, we evaluate the effect of cartel characteristics and changes in the market and legal environment on the magnitude of overcharges imposed by private cartels in the United States and other geographic markets as early as the eighteenth century. The median overcharge attained by cartels represented in our sample is 18 percent of selling price. International cartels imposed higher overcharges than domestic cartels. Longer cartel episodes generated higher overcharges. Overcharges achieved in the United States and European markets were lower than overcharges imposed in the Asian markets and in the rest of the world. Overcharges tended to decline as antitrust enforcement became stricter. Higher overcharges were associated with markets where cartels had high market shares and with markets characterized by high levels of fixed costs.

	I. INTRODUCTION

TOP ABSTRACT I. INTRODUCTION II. THEORETICAL AND EMPIRICAL... III. DATA DESCRIPTION AND... IV. EMPIRICAL MODELS AND... V. ESTIMATION RESULTS VI. CONCLUSION

 
During recent decades, antitrust authorities paid more attention to collusive conduct than ever before. Antitrust laws of many countries were modified to ensure a more effective detection and punishment of cartel participants. Despite these changes, the recent cartel cases reveal that many modern cartels were very successful in raising prices and managed to do this for a relatively long period of time. The fact that many firms are members of a number of cartel agreements and that there are firms that are repeated offenders suggests that the benefits from collusion outweigh the costs of collusion, including possible government and private sanctions. This situation encourages antitrust authorities to re-evaluate antitrust laws to ensure that they provide a proper deterrent effect. Both domestic and international antitrust policy decisions may be influenced by empirical evidence of cartel behavior and of the effectiveness of antitrust laws during the preceding decades.

The objective of our study is to conduct an empirical analysis of factors influencing the magnitude of cartel overcharges. The analysis is based on the largest data set of overcharge estimates that are associated with cartels operated during the last two centuries in different geographic and product markets. We analyze overcharges attained in the U.S. market versus overcharges achieved in other geographic markets. The study provides empirical evidence on a number of issues that are important for antitrust policy decision-making and are currently under consideration in the United States, other countries, and at the international level.

1. The average gain from price-fixing (10 percent of selling price) and a corresponding base fine (20 percent of affected commerce) established in the U.S. Federal Sentencing Guidelines (U.S. FSG) have been the focus of many controversial discussions. In its recent report, the U.S. Antitrust Modernization Commission recommended to Congress to " ... encourage the Sentencing Commission to reevaluate and explain the rationale for using 20 percent of the volume of commerce affected as a proxy for actual harm, including both the assumption of an average overcharge of 10 percent of the amount of commerce affected and the difficulty of proving the actual gain or loss."¹
   
2. Starting in 1890, antitrust laws were modified over time to ensure the effective punishment and prosecution of cartel participants in each particular period of history. Did these changes have a negative effect on the level of cartel overcharges? Does the presence of an active antitrust law in a country affect the level of overcharges attained in this country relative to the overcharges imposed in the countries without such laws?
   
3. Cartels that are international in membership represent a major concern for both domestic and international antitrust policies, especially with respect to international trade within the WTO. These cartels are more difficult to prosecute because of the bounded legal power of domestic antitrust authorities. On the other hand, international cartels face more discipline enforcement challenges than domestic cartels, which ultimately affect their success. Do international cartels bring more damages than domestic cartels?
   

In addition, we examine the effect of cartel characteristics and market environment on the magnitude of cartel overcharges. We consider the types of cartel agreement (bid-rigging or not), its legal status, cartel duration, and differences in the cartelized market structures. The paper is organized as follows. We begin with a discussion of the theoretical and empirical background, followed by a section describing the data and case selection procedure. In the following section, we present a discussion of empirical models and hypotheses. Finally, we present the results of the estimation and conclusion of the research.

	II. THEORETICAL AND EMPIRICAL BACKGROUND

TOP ABSTRACT I. INTRODUCTION II. THEORETICAL AND EMPIRICAL... III. DATA DESCRIPTION AND... IV. EMPIRICAL MODELS AND... V. ESTIMATION RESULTS VI. CONCLUSION

 
Cartel overcharge, a price increase due to collusive conduct, is a characteristic of cartel effectiveness and cartel success. Markets that are characterized by homogeneity of product and purchasing commitments, high market concentration, a small number of sellers, inelastic demand, and high barriers to entry facilitate collusive conduct and potentially contribute to its success. Driven by the objective of joint profit maximization, firms acting in these (oligopolistic) markets form collusive agreements.² If the firms manage to act like a single monopolist, theoretically they can reach a monopolistic level of price. Practically, they are not likely to achieve this price level because of the costs associated with collusion.

The theory of the core is an alternative theory that explains why cartels are formed.³ This theory considers a collusive agreement as a mean to reach equilibrium rather than to extract a monopoly profit. The theory assumes that cartels are organized in the industries with high barriers to entry (that is, high fixed costs), such as ocean shipping.⁴ There are other important differences between the theory of the core and cartel theory. First, according to the theory of the core, cartels are more likely to emerge when demand and/or supply are more variable. Second, the theory of the core predicts that cartels are less likely to be organized when the entry is legally restricted. The opposite is true in the case of the cartel theory.

The size of the attained overcharge is indicative of how successful the cartel is in maximizing its benefits and minimizing the cost of collusion. A review of a number of recently conducted surveys focusing on cartel behavior suggests that many known cartels were effective in achieving positive overcharges during different periods of history.⁵ In general, the surveys differ due to the publication sources surveyed and the historical periods of cartel operation.⁶ The largest survey consists of approximately 1000 overcharge estimates.⁷

A very limited number of studies conducted econometric analysis of cartel overcharges. This type of analysis is challenging because of the problems associated with compiling a suitable data set. Griffin⁸ conducted an econometric analysis of the Lerner indices based on a sample of fifty-four cartel episodes associated with international cartels operated legally from 1888 to 1984. He finds that market concentration, cartel market share, and the complexity of cartel agreements are positively related to the changes in Lerner indices.

Using an approach similar to Griffin,⁹ Connor and Bolotova¹⁰ and Bolotova et al.¹¹ examined the relationship between the cartel overcharge and various factors influencing its magnitude using a sample of approximately 400 overcharge estimates. These estimates are associated with cartels that operated as early as in the eighteenth century in different product and geographic markets. The first study focuses on the survey and meta-analysis of cartel overcharges with a particular emphasis on the effect of different estimation methods and publication sources on the variability of overcharge estimates. The second study examines the overcharges attained in the food industry-related markets versus overcharges imposed in other industries. The authors find that the differences in the publication sources have a larger impact on the overcharge levels than the differences in the estimation methods. In addition, there is evidence that changes in antitrust laws have had some effects on the magnitude of cartel overcharges.

Although our study uses the same original data set that was used in Connor and Bolotova¹² and Bolotova et al.,¹³ it differs from these studies in: (a) overcharge case selection procedure; (b) approach used to calculate overcharge estimates; and (c) specification of empirical models. The first two differences are mentioned in a section discussing the overcharge case selection procedure. To analyze the U.S. market overcharges, this study specifies empirical models such that the estimated effects represent changes in the U.S. market overcharges. In two other studies, the estimated effects are associated with overcharges calculated for the world market.

A. Measuring Cartel Overcharges

Overcharge is the difference between the price achieved as a result of collusive actions and a more competitive (benchmark) price. The latter is a price that would exist in the market if collusion did not take place.¹⁴ Usually, cartel overcharges are represented using a relative measure, the overcharge rate. The following two relative measures are possible. The first measure (equation 1) is a ratio of the price difference due to collusion (absolute measure) to the price during collusion. The second measure (equation 2) is a ratio of the price difference to the benchmark price.

(1)

and

(2)

The difference between these two formulas is which price appears in the denominator.¹⁵ Under the assumption that a market absent collusion (a more competitive or a benchmark market) is a perfectly competitive market, the first relative measure corresponds to the Lerner Index.¹⁶ Although both approaches to calculating overcharge rates are equally acceptable, the first formula has a number of advantages over the second formula. First, overcharge estimates that are calculated using formula (1) have an upper boundary, and the possible maximum value of cartel overcharge is 100 percent. Second, these overcharge estimates can be directly compared with the overcharge proxy (gain from price-fixing) established in antitrust laws.

	III. DATA DESCRIPTION AND CASE SELECTION PROCEDURE

TOP ABSTRACT I. INTRODUCTION II. THEORETICAL AND EMPIRICAL... III. DATA DESCRIPTION AND... IV. EMPIRICAL MODELS AND... V. ESTIMATION RESULTS VI. CONCLUSION

 
A. Case Selection Procedure

To conduct an empirical analysis of cartel overcharges, we use part of the data set compiled by Connor.¹⁷ The original data set consists of approximately 1,000 overcharge estimates for approximately 280 product markets. The estimates are available for different product and geographic markets as well as for different time periods starting from the eighteenth century. The overcharges in the data set were calculated using a number of estimation methods and were published in various publication sources.

The data were collected from peer-reviewed academic journals, dissertations, court and commission decisions, Organization for Economic Cooperation and Development reports, books, government publications, working papers, and other sources. Connor's goal was to collect the largest possible number of overcharge estimates. He avoided applying any sort of quality screening.¹⁸ In the majority of cases, the surveyed sources provided the overcharge calculations. In the rest of the cases, the author derived the overcharge estimates using price data presented in the surveyed publication sources.

This large survey of cartel overcharges provides information that allows for testing a set of important hypotheses on cartel behavior that have not been tested in the existing empirical literature. A number of problems related to any data set like this have to be mentioned. This data set includes overcharges associated with cartels that: (a) have revealed themselves or have been uncovered; (b) information on which has been published; and (c) cases that have been selected by the author to be included in his data set. Therefore, the data set does not include overcharges of: (a) cartels that have never revealed themselves; (b) cartels that have become known but their information has never become publicly available; and (c) cartels that have not been selected by the author.¹⁹

We are not aware of the exact number of uncovered cartels. There are estimates suggesting that only 13–17 percent of illegal cartels are caught.²⁰ Consequently, approximately 80 percent of such cartels are not known. Many of these cartels are likely to impose relatively low overcharges, as they do not raise the concerns of their buyers and antitrust authorities. Given that cartels usually operate in highly concentrated markets, these overcharges may not be very noticeable and sometimes the price increases are very difficult to distinguish from a price fluctuation due to a demand shock or any other natural market movement.²¹

For this study, we compile a sub-set of the original data set.²² We apply a selection procedure to the original data set to eliminate some duplication in the observations. First, two types of the overcharge estimates were available in the original data set: average and peak overcharges. We analyze the average level of overcharge, as it is a more representative measure of the cartel impact over the entire period of conspiracy.²³ Second, some cartel episodes were represented by more than one overcharge estimate. This happened because the same episode was analyzed in different studies and different methods of overcharge estimation were used within a single study.

We had to eliminate all redundant estimates to form a data set for this study. Given that overcharges characterizing the same cartel often fall in a very wide range and tend to be asymmetrically distributed, we decided to select the median overcharge estimate among the available alternatives for each episode.²⁴ As a result, the data set for this study includes 333 observations. The same product market may be represented by multiple observations when more than one overcharge estimate is collected for that market. This happens because the same product market may include overcharge estimates belonging to different geographic markets or the same cartel may experience more than one episode. Each observation represents a cartel episode, which is an uninterrupted period of collusion with a corresponding set of rules and membership.

The overcharge estimates presented in the original data set are overcharge rates calculated as the ratio of price increase from collusion to the benchmark price (formula 2). After selecting the overcharge estimates for this study, we transformed the estimates into the overcharge rates represented as the ratio of price increase from collusion to the price during collusion (formula 1).²⁵ This transformation is useful for the following reasons. First, the distribution of cartel overcharges is considerably less skewed after the transformation relative to the original data set overcharge estimates.²⁶ Second, the transformation allows us to compare the overcharge estimates in the sample with the overcharge proxy referred to in antitrust laws.²⁷

The original data set has information on the cartel characteristics associated with each overcharge estimate. We can distinguish overcharges imposed by legal and illegal cartels, bid-rigging and non-bid-rigging cartels, and cartels with domestic and international membership. Using the description of cartel episodes, it is possible to form a judgment on cartel (episode) duration and the geographic market for which the overcharge was estimated. This information is used in the empirical analysis.

To examine the effect of market structure on the overcharge level, we form a sub-set of our data set that includes only overcharges associated with modern cartels operating illegally in the U.S. market (including Canada). For this sub-set, we were able to collect data on cartel market shares and the information necessary to calculate a proxy for fixed costs. The market shares were collected from the government reports, court and antitrust authority decisions, as well as other publicly available sources discussed in Connor.²⁸

We constructed a ratio of capital expenditures to the value of shipments to be used as a proxy for fixed costs. All industries represented in the modern cartel data set are manufacturing industries. The U.S. Economic Census reports information on capital expenditures and value of shipments for all manufacturing industries.²⁹ We matched the industries represented in the data set with the industries for which the Economic Census reports the data. We used the 1997 Economic Census data, as this year is the midpoint of the historical period represented in the data set.³⁰ As a result, the modern cartels data set has thirty-six observations with twenty-seven observations associated with the U.S. market and nine observations corresponding to the Canadian market.³¹

B. Descriptive Statistics

The mean overcharge characterizing the overall sample is 20.77 percent of the market price during collusion, and the median is 18.37 percent. The minimum value of overcharge is –5.26 percent and the maximum is 81.82 percent.³² The average cartel duration is eight years and the median is five years. Of the overall sample, 55 percent of the cartels are domestic in membership, 22 percent are bid-rigging, and 72 percent are guilty. Most of the overcharge estimates are outcomes for the United States (43 percent) and Europe (36 percent) rather than for Asian markets (12 percent) and markets in the rest of the world (5 percent). As for the distribution of cartel overcharges across different antitrust regimes,³³ they are distributed relatively evenly across six periods covering 1770–2004 with 36 percent belonging to the last two decades. Of the overall sample of overcharge estimates, 23 percent correspond to the markets producing food, 15 percent to natural resources, 14 percent to metals, and 15 percent to chemicals.

The data characterizing market structure provide evidence that cartels operate in highly concentrated markets. In the case of modern cartels that were active in the U.S. market, the average cartel market share was 82.4 percent with a median of 87 percent. The minimum and maximum cartel market shares were 45 and 100 percent, respectively. The average share of the capital expenditures in the value of shipments (the fixed cost proxy) is 6.5 percent, with a minimum value of 1.5 percent and a maximum value of 11.1 percent.

The distribution of cartel overcharges across different antitrust periods is presented in Table 1. The distribution is presented for the U.S. market, other (non-U.S.) markets, and the overall sample (the U.S. and non-U.S. markets). Because the distribution of cartel overcharges is skewed, we calculate the median overcharges instead of using the average overcharges. The following features characterize the U.S. market. The median overcharge attained in this market during the period of 1770–2004 is 18.53 percent of the selling price, which is approximately the same as the median non-U.S. market overcharge. The highest overcharge is 27.38 percent (1920–1945) and the lowest overcharge is 13.04 percent (1946–1973). The median overcharge characterizing the latest antitrust law regime (1991–2004) is 19.74 percent. The lowest median overcharge in the case of both the U.S. and non-U.S. markets is associated with the period of 1946–1973. Before this period, overcharges had been increasing. After this period, there is no obvious pattern describing overcharges in the two latest regimes. In the case of the U.S. market, overcharges have been increasing. In the case of the non-U.S. markets, there was a considerable increase in the overcharge level, and then a significant decrease.

View this table: [in this window] [in a new window]   Table 1. The median overcharge estimates, by time periods and geographic markets

 
Table 2 shows the distribution of overcharge estimates across the 10 percent-points intervals of the overcharge distribution. Approximately 54 percent of the U.S. market overcharges and 50 percent of the non-U.S. market overcharges fall in the interval between 0 and 20 percent. Approximately half of these overcharges belong to the interval from 0 to 10 percent and the other half falls in the interval from 10 to 20 percent. In the case of the U.S. market overcharges, 71 percent of the analyzed estimates (101 observations) are higher than 10 percent of the market price. A similar pattern characterizes the distribution of the non-U.S. market overcharges. About 75 percent of the available estimates (143 observations) are higher than 10 percent.

View this table: [in this window] [in a new window]   Table 2. Distribution of overcharge estimates

 

	IV. EMPIRICAL MODELS AND HYPOTHESES

TOP ABSTRACT I. INTRODUCTION II. THEORETICAL AND EMPIRICAL... III. DATA DESCRIPTION AND... IV. EMPIRICAL MODELS AND... V. ESTIMATION RESULTS VI. CONCLUSION

 
A. Empirical Models

Two empirical models are developed to evaluate the impact of cartel characteristics and the market and legal environment of cartel operation on the magnitude of cartel overcharges. The first model is applied to the overall sample (333 observations), a sample of U.S. market overcharges (142 observations), and a sample of non-U.S. (other) market overcharges (191 observations). The second model is applied to a sample of modern cartels operated in the U.S. and Canadian markets (thirty-six observations).³⁴

(3)

(4)

The dependent variable in both models is the overcharge rate (OVRATE) achieved during cartel episode i. The overcharge rate is calculated as a ratio of price increase from collusive conduct to price during collusion (formula 1). The explanatory variables in the first model (equation 3) are: is an intercept; DURATIONi is a continuous variable representing cartel (episode) duration; Ci is a binary variable representing cartel characteristics; Fi is a binary variable representing product markets; Gi is a binary variable representing geographic markets; Pi is a binary variable representing antitrust law regimes; and i is an error term.

The explanatory variables in the second model (equation 4) are an intercept (); three continuous variables representing cartel (episode) duration (DURATIONi), cartel market share (CMSi), and cartelized industry fixed costs (FCOSTi); a set of binary variables associated with various product markets (Ii); two binary variables representing (a) Canadian market (CAi) and (b) domestic in membership cartels (DOMESTICi); and an error term (i). A detailed description of the explanatory variables and their expected signs is presented in Table 3, and hypotheses are discussed in the following section.

View this table: [in this window] [in a new window]   Table 3. Definition of explanatory variables

 
B. Hypotheses

The size of cartel overcharge is influenced by numerous factors characterizing the internal and external (economic and legal) environment of cartel operation. Given the available information, we quantify the effect of the following factors on the magnitude of cartel overcharges. The market concentration, barriers to entry, market structure, and the type of market institution are the market environment factors. The changes in antirust regulations over time and across geographic markets of cartel operation represent the legal environment. Finally, cartel duration, cartel membership structure, and cartel legal status are cartel characteristics that are hypothesized to influence the magnitude of cartel overcharges.

As predicted by the theory of oligopoly,³⁵ market concentration is positively related to the degree of market power, which is measured by the overcharge rate in our study. Highly concentrated markets are usually characterized by high barriers to entry and exit. Firms acting in these markets incur high fixed costs. To recover these fixed costs, the firms have incentive to collude.³⁶ The level of fixed costs is expected to be positively correlated with the size of overcharge. Therefore, the estimated coefficients for cartel market share and fixed costs (FCOST) are expected to be positive.³⁷

Given that industries differ due to market structure, including demand and supply conditions, barriers to entry and exit, product homogeneity, technology, structures of the upstream and downstream industries, and other factors, overcharge levels are likely to vary significantly across different industries. We introduce a set of binary (dummy) variables characterizing twelve groups of industries represented in our cartel sample to capture these differences. At this stage, we do not make any predictions about the signs of the estimated coefficients for the industry-specific variables. However, we expect many of them to be statistically significant from the reference group represented by the industries producing chemicals (CHEM).

As noted by Stigler,³⁸ collusion is more effective against buyers who report prices correctly and fully. An example is sealed-bid auctions organized by governments. This type of market institution allows cartel participants to use the reported information to monitor their agreements and to detect cheating. Cartels operating in this type of market institution are referred to as bid-rigging cartels. The participants of these cartels are involved in submitting non-competitive bids.³⁹ The estimated coefficient for bid-rigging cartels (BIDRIG) is expected to be positive.

Cartel duration is a characteristic of cartel effectiveness and cartel success and is expected to have a significant effect on the overcharge level. In particular, longer cartel duration may lead to a higher overcharge attained by the cartel. If a cartel is successful in maintaining its discipline, it can have a larger effect on the price movement by the mean of a direct price increase and the price variance control. Therefore, this cartel can operate longer than a less successful cartel whose members do not follow the established price discipline. The latter situation results in a lower overcharge level and a faster termination of the cartel. The estimated coefficient for cartel duration (DURATION) is expected to be positive.

Cartels that were found or pled guilty are likely to impose lower overcharges than those not prosecuted (legal). Legal cartels do not have to incur the costs to mask their price-fixing activity from antitrust authorities. Conversely, the overcharges attained by legal cartels may be the same or even lower than the overcharges imposed by illegal price-fixing agreements. Legal cartels are legal because they predate antitrust laws or because they are required to register with a government authority. As illegal cartels, legal cartels are private self-enforced agreements with the internal mechanism of discipline supervision. If the members of legal cartels do not follow their discipline, they are not able to raise prices to higher levels despite their legal status. Therefore, opportunistic behavior of legal cartel members may result in lower overcharges than those attained by illegal cartels with strongly enforced discipline.⁴⁰ It is also important to note that overcharges reported in the cases of guilty cartels may be underestimated.⁴¹ In addition, different methods of overcharge calculation may introduce either upward or downward bias.⁴² In summary, we are uncertain about the sign of the estimated coefficient for illegal cartels (GUILTY). However, it is likely that there is no statistically significant difference between the overcharges achieved by legal and guilty cartels.

Cartels that are international in membership (and generally operate in a number of countries) are expected to achieve higher overcharges relative to cartels that are domestic in membership. International cartels are more difficult to prosecute because of the bounded legal power of domestic antitrust enforcement. Therefore, international cartels have more favorable conditions to exercise their price-fixing activity than domestic cartels. Because they usually operate in a number of countries, national borders make geographic price (overcharge) discrimination more likely. Also, international cartels do not have the import competition that domestic cartels may face. On the other hand, international cartels may face more communication problems due to cultural differences, and this may hamper cartel success. In summary, the estimated coefficient for domestic in membership cartels (DOMESTIC) is expected to be negative.

The geographic location of cartel operation is likely to affect the magnitude of cartel overcharges. The output allocation and price-fixing schemes are considered to be illegal under the antitrust laws of many countries. It should be expected that overcharges imposed in the countries without antitrust laws or without rigorously enforced antitrust laws (Latin American, African, and most Asian countries) are higher than overcharges attained in the countries with a long history of antitrust law enforcement (the United States, Canada, EU, and European countries).⁴³ Relative to the reference group represented by overcharge estimates for the U.S. market, the estimated coefficients for regional binary variables for Canada, EU, and European nations are expected to be either positive or negative, but not marginally and statistically significant from the reference group. In contrast, the estimated coefficients for regional binary variables for Asia (ASIA) and the rest of the world (ROW) are expected to be positive.

Connor⁴⁴ distinguishes six antitrust law regimes that describe the evolution of antitrust law during the last two centuries. Changes in both the legal and economic environments were taken into account when this classification was developed. The first break point is 1890, which is chosen because of the enactment of the Sherman Act in the United States and the Anti-Combines Act in Canada. The second break point is 1919; it represents the end of World War I. The third break point is 1945, the beginning of the post-World War II era. Most international cartels were not active during the world wars. During the period of 1945–1973, many countries outside of North America started developing and enforcing antitrust laws. The U.S. penalties became more severe in comparison with the previous periods of history. The fourth breakpoint is 1974, the year the U.S. antitrust laws made price-fixing activity a felony. This resulted in an increase in corporate sanctions and in individual prison sentences. It is also about the time when the EU began fining cartels. The last breakpoint is 1990. The U.S. antitrust laws increased the penalties from $1 million to $10 million. In addition, the leniency program was greatly improved in 1993. Therefore, under the assumption that each following regime is more effective, it is expected that the magnitude of the overcharge becomes smaller in each subsequent period. Relative to the reference antitrust law regime represented by the period of 1919–1945 (P3), the estimated coefficients for the latest antitrust law regimes P4 (1946–1973), P5 (1974–1990) and P6 (1991–2004) are expected to be negative.

	V. ESTIMATION RESULTS

TOP ABSTRACT I. INTRODUCTION II. THEORETICAL AND EMPIRICAL... III. DATA DESCRIPTION AND... IV. EMPIRICAL MODELS AND... V. ESTIMATION RESULTS VI. CONCLUSION

 
Given the survey nature of our data set, we do not make any strong assumptions about the error distribution, and we estimate the models with the ordinary least squares (OLS) estimator as semiparametric linear regression models.⁴⁵ We do not conduct any formal tests for the presence of autocorrelation. Because the data come from different periods of time and the overcharges are estimated for cartel episodes with different length, we cannot organize the data to easily capture dynamics in the time dimension. We examined the data for the presence of influential observations (outliers) because the distribution of cartel overcharges is slightly skewed. Using the interquartile approach to distinguish between mild and extreme outliers, we find that there are no extreme outliers, and there are ten mild outliers.⁴⁶

To determine whether we can pool the U.S. market and non-U.S. market data, we introduce a set of interaction variables in the regression model based on the product of each explanatory variable with the binary variable U.S. We then use a Wald test to determine whether the interaction effects are jointly significant, which would indicate that we should estimate a separate model based on the U.S. market overcharges. We fail to reject the pooled model and can combine the U.S. and non-U.S. market data in a single regression model.⁴⁷ However, given the focus of the paper and the importance of the topic for antitrust policy decision-making, we estimate two separate regressions based on the U.S. and non-U.S. samples. The OLS estimation results and diagnostic test statistics for the overall sample of cartel overcharges and the U.S. and non-U.S. market overcharges are presented in Table 4. The magnitude of the estimated coefficients and their statistical significance are consistent across the three models.

View this table: [in this window] [in a new window]   Table 4. Cartel overcharges: the ordinary least squares estimation results

 
Most of the estimated coefficients have expected signs and several of them are statistically significant. The independent variables included in the overall regression explain approximately 16 percent of variation in the dependent variable, the overcharge rate. We conducted a series of Wald tests to examine whether the individual groups of independent variables had the joint statistically significant effect on the overcharge rate. Cartel characteristics, the latest antitrust law regimes, and differences in the product markets tend to have a stronger impact on the magnitude of cartel overcharges than differences in the geographic markets (Table 4).

As the majority of independent variables is represented by binary variables, it is important to remember that the intercept (constant) in the all-cartels regression is the median overcharge attained in the U.S. market by cartels that are international in membership, non-bid-rigging, and legal that operated during the period of 1920–1945 in the industries producing chemicals. All the estimated coefficients are interpreted relative to this reference group.

Each additional year of cartel operation increases the overcharge level by 0.2 percent-points, on average. Domestic-in-membership cartels attain overcharges that are 6 percent-points lower than overcharges imposed by international-in-membership cartels. Although bid-rigging cartels achieve higher overcharges than non-bid-rigging cartels, this difference is not statistically significant. A similar situation characterizes overcharges imposed by guilty cartels versus overcharges achieved by legal cartels.

The estimated coefficients for regional binary variables for Canada (CA), Asia (ASIA), and the rest of the world (ROW) are positive (that is, overcharges in these markets are higher than the reference group overcharges). However, they are not statistically different from the reference (the U.S. market) overcharges. Conversely, the estimated coefficients for regional binary variables for European Union (EU) and European Nations (EN) are negative and also not statistically significant from the reference group.

The estimated coefficients for the antitrust law regimes variables characterize the changes in the U.S. market overcharges over time (relative to the reference group, the 1920–1945 U.S. market overcharges). Consistent with the descriptive statistical analysis, the highest overcharges were attained during the period of 1920–1945. Overcharges associated with all other regimes are lower than the reference regime overcharges and are statistically different from the reference group (except for the period of 1974–2004). For example, overcharges corresponding to the period of 1946–1973 are almost 8 percent-points lower than the reference group overcharges. Overcharges associated with the latest antitrust period (1991–2004) are 6 percent-points lower than the reference group overcharges.

Differences in the market structures tend to affect the magnitude of overcharges attained in the analyzed industries. For example, overcharges achieved in the markets of forest products (FOREST), equipment (EQUIPM), and fuel (FUEL) are 7–8 percent-points lower than overcharges corresponding to the reference group markets producing chemicals. Conversely, overcharges associated with services and other industries are 12.3 and 6.1 percent-points higher than the reference group overcharges.

The U.S. market regression estimation results follow the general pattern characterizing the overall sample. We will observe some differences. First, the effect of cartel duration on the overcharge magnitude is almost two times stronger in the U.S. regression. Second, the estimated coefficient for bid-rigging cartels is negative, which contradicts our expectations; however, this effect is not statistically significant. Third, the estimated coefficients for antitrust law regimes exhibit slightly higher magnitudes in the U.S. market regression compared with the overall regression.

The interpretation of the non-U.S. market regression results is similar to the overall regression. It is important to remember that the reference group in this regression is the overcharges attained in Europe (EU and EN) by cartels that are international in membership, non-bid-rigging, and legal that operated during the period of 1920–1945 in the industries producing chemicals. All estimated coefficients have to be interpreted relative to this group.

The estimation results for the regression based on the modern U.S. and Canadian market overcharges are presented below (standard errors appear in parentheses).

R2 = 0.18; BPG LM Statistic P-value = 0.37; N = 36 observations. The superscripts denote statistical significance as it is described in a footnote to Table 4.

The magnitude of the estimated coefficients for cartel duration and cartel membership structure is similar to the coefficients characterizing the overall sample regression and the U.S. market regression. However, in the case of modern U.S. and Canadian cartels, these effects are not statistically significant, perhaps, due to a small sample size. The estimated coefficients for CMS and FCOST have reasonable magnitude, expected signs and are statistically significant. A 1 percent-point increase in a cartel market share results in a 0.16 percent-point increase in the overcharge level. Consequently, a 20 percent-points increase in a cartel market share leads to a 3.2 percent-points increase in the overcharge. A higher level of fixed costs results in a higher level of overcharges. A 1 percent-point increase in the share of capital expenditures in the value of shipments results in a 1.46 percent-points increase in the overcharge level. As in the overall sample regression, differences in the market structures that are captured by the industry-specific variables affect the magnitude of cartel overcharges. The Canadian market overcharges are not statistically significant from the U.S. market overcharges.

	VI. CONCLUSION

TOP ABSTRACT I. INTRODUCTION II. THEORETICAL AND EMPIRICAL... III. DATA DESCRIPTION AND... IV. EMPIRICAL MODELS AND... V. ESTIMATION RESULTS VI. CONCLUSION

 
The analysis presented in the paper provides evidence suggesting that cartels are successful in raising market prices and many of them manage to do this for a relatively long period of time. The results suggest that the average gain from price-fixing is approximately 20 percent of selling price for the most recent antitrust regime. This estimate is twice as high as the average gain from price-fixing proxy (10 percent of selling price) and is equal to the base fine, 20 percent of affected sales, established in the U.S. Federal Sentencing Guidelines. The U.S. Federal Sentencing Guidelines are based on the principle of deterrence and assumes that the base fine has to be two times as high as the average gain from price-fixing. The current situation, where the average gain from price-fixing exceeds the sanctions imposed on cartel participants, results in a sub-optimal deterrence. The implication of this finding is that the presumption on the average gain from price-fixing has to be carefully re-examined and, perhaps, should be raised. Consequently, the base fine has to be adjusted upward as well.

The differences in antirust law systems across various countries influence the overcharge level. Overcharges tend to be lower in the countries with more effective antitrust systems (the United States, EU, and individual European countries) than in the countries without active or effectively enforced antitrust laws (Asia and ROW). The evidence on the effect of changes in antirust law over time on the overcharge level is mixed. The most effective antitrust law regime was the period of 1946–1973.

One of the problems dealing with the effectiveness of the latest antitrust regime (1991–2004) is a growing number of international cartels that are more difficult to prosecute than domestic cartels. In addition, some global cartels do not have any competition at all, and are likely to exercise geographic price (overcharge) discrimination by imposing higher overcharges in the countries with weak antitrust regulations and lower overcharges in the countries with strong antitrust systems. Our results suggest that international-in-membership cartels impose higher overcharges than domestic-in-membership cartels, and the difference in the overcharge level on average is 6 percentage points. Therefore, more efforts have to be taken at both a domestic and international level to deter this type of conduct effectively.

Another issue to consider is that illegal cartels manage to attain at least the same level of overcharges as legal cartels and even higher. This may imply that illegal cartel agreements become more sophisticated and cartel participants manage to enforce them very effectively taking into account both the economic and legal environment. It should be pointed out that in the case of some of the modern cartels, overcharges reported in plea agreements are likely to be understated.

The current U.S. antitrust regulation assumes that the overcharge level (rate) does not depend on cartel duration; in other words, it is the same during all years of cartel operation. The empirical findings suggest that cartel duration is positively related to the overcharge level. Longer cartel episodes generate higher overcharges, and each additional year of cartel operation raises the attained overcharge level. This occurs because of the effective discipline supervision and rigorous enforcement of the price (output) control strategy. As a result, cartel actions affect not only the price level but also price variance as well, which becomes lower. The current antitrust law does not target cartel duration effectively and focus only on the overcharge level.

The study provides empirical evidence that cartels are organized in highly concentrated markets with high level of fixed costs. The median cartel market share in the case of a sample of modern cartels operated in the U.S. market is 87 percent. Overcharges tend to be higher in the markets where cartels have larger market shares and the markets characterized by higher levels of fixed costs.

Finally, differences in the market structures influence the overcharge levels attained in the analyzed industries. Some industries provide more favorable environments for cartel actions. For example, the highest overcharges are imposed in markets providing services (including shipping services), producing metals, chemicals, and in other industries. The lowest overcharges are in the markets producing forest products, equipment, and fuel. One of the most important factors explaining this outcome is the level of demand elasticity. Markets with inelastic demand and the absence of substitution possibilities are those with the highest overcharges, and services are an example of this type of market.

	FOOTNOTES

 
^* Assistant Professor, Department of Agricultural Economics and Rural Sociology, University of Idaho, Moscow, ID 83844-2337, USA.

^** Professor, Department of Agricultural Economics, Purdue University, West Lafayette, IN 47907-2056, USA. E-mail: jconnor{at}purdue.edu. Purdue ARP number: 2008-18367.

Associate Professor, Department of Economics, University of Missouri, Columbia, MO 65211-6040, USA. E-mail: millerdou{at}missouri.edu. Research for this paper was supported in part by the Edmund S. Muskie Ph.D. Fellowship Program, a program of the Bureau of Educational and Cultural Affairs (ECA), U.S. Department of State, administered by the American Councils for International Education (ACTR/ACCELS). The opinions expressed in the paper are those of the authors and do not necessarily express the views of either ECA or the American Councils.

¹ U.S. ANTITRUST MODERNIZATION COMMISSION REPORT AND RECOMMENDATIONS, Chapter 3, CIVIL AND CRIMINAL REMEDIES 295, (2007), http://www.govinfo.library.unt.edu/amc/report_recommendation/chapter3.pdf.

² George J. Stigler, A Theory of Oligopoly, 72 J. POL. ECON. 44 (1964).

³ William Sjostrom, Collusion in Ocean Shipping: A Test of Monopoly and Empty Core Models, 97 J. POL. ECON. 1160 (1989); Lester G. Telser, A Theory of Self-Enforcing Agreements, 53 J. BUS. 27 (1980).

⁴ To reach the equilibrium, the firms have to be assigned quantities to produce. Unless the quantities are assigned, the equilibrium does not exist, as the number of potential producers exceeds the number of acting producers. The former are making attempts to enter the industry, as they have already incurred high fixed costs. See Sjostrom, supra note 3.

⁵ Gregory J. Werden, The Effect of Antitrust Policy on Consumer Welfare: What Crandall and Winston Overlook, U.S. Department of Justice Antitrust Division Discussion Paper No. EAG 03-2, (2003), http://papers.ssrn.com/sol3/papers.cfm?abstract_id=384100; Organization for Economic Cooperation and Development (OECD) Report, Fighting Hard-Core Cartels: Harm, Effective Sanctions and Leniency Programs, (2002); OECD Report, Hard Core Cartels: Recent Progress and Challenges Ahead (2003); Margaret C. Levenstein & Valerie Y. Suslow, What Determines Cartel Success? 44 J. ECON. LIT. 43 (2006); John M. Connor, Price-Fixing Overcharges: Legal and Economic Evidence, Purdue University Department of Agricultural Economics Staff Paper No. 04-17, (2005), http://www.agecon.purdue.edu/staff/connor/papers/PRICE_FIXING_OVERCHARGES_APPENDIX_TABLES_8-05.pdf; John M. Connor, Price-fixing Overcharges: Legal and Economic Evidence, 22 RES. LAW ECON. (2007).

⁶ The publication sources surveyed by the authors include peer-review journals, chapters in edited books and monographs, antitrust authority and court decisions, government reports, working papers, business newspapers, trade magazines, and news services.

⁷ See Connor, supra note 5.

⁸ James Griffin, Previous Cartel Experience: Any Lessons for OPEC?, in ECONOMICS IN THEORY AND PRACTICE: AN ECLECTIC APPROACH 179 (Lawrence R. Klein & Jamie Marquez eds., Boston, Kluwer Academic Publishers 1989).

⁹ Id.

¹⁰ John M. Connor & Yuliya Bolotova, Cartel Overcharges: Survey and Meta-analysis, 24 INT'L. J. INDUS. ORG. 1109 (2006).

¹¹ Yuliya Bolotova, John M. Connor & Douglas J. Miller, Factors Influencing the Magnitude of Cartel Overcharges: An Empirical Analysis of Food-Industry Cartels, 23 GRIBUSINESS: INT'L J. 17 (2007).

¹² Supra note 10.

¹³ Supra note 11.

¹⁴ Price before collusion and price after collusion are some examples of the benchmark prices.

¹⁵ There is a simple relation between these formulas. If an overcharge is calculated using one of them, it can be converted to the alternative representation. L = [O/(100 + O)] x 100%, where L[=OvRate (1)] is an overcharge calculated as a ratio of the price increase to the price during collusion, and O[=OvRate (2)] is an overcharge calculated as a ratio of the price increase to the benchmark price.

¹⁶ The role of the benchmark market definition in antirust settings is discussed in John M. Connor, "Our Customers Are Our Enemies:" The Lysine Cartel of 1992–1995, 18 REV. INDUS. ORG. 5 (2001) and in Lawrence J. White, Lysine and Price-fixing: How Long? How Severe? 18 REV. INDUS. ORG. 23 (2001).

¹⁷ Supra note 5.

¹⁸ Overcharges appeared in newspapers, magazines, and newsletters were not included because these were usually anonymous sources.

¹⁹ A part of the latter group is represented by the anonymous overcharge estimates that intentionally were avoided by the author.

²⁰ Peter G. Bryant & E. Woodrow Eckard, Price Fixing: The Probability of Getting Caught, 73 REV. ECON. STAT. 531 (1991).

²¹ There is an empirical study that provides evidence that cartels that impose relatively low overcharges tend to have longer duration than cartels that impose higher overcharges. See Yuliya V. Bolotova, Three Essays on the Effectiveness of Overt Collusion: Cartel Overcharges, Cartel Stability and Cartel Success (2006) (Ph.D. dissertation, Department of Agricultural Economics, Purdue University), available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=939078. This type of conduct (that is, low overcharge and long cartel duration) has a lower probability of being detected by either buyers or antitrust authorities.

²² The data set for this study was formed using information presented in Appendix Table 1 and Table 2 of Connor, supra note 5.

²³ In contrast, the peak overcharges are usually those achieved during a relatively short (the most successful) period of cartel history.

²⁴ Connor & Bolotova, supra note 10, and Bolotova et al., supra note 11, used the minimum overcharge case selection procedure.

²⁵ We follow the approach used in Bolotova et al., supra note 11. Connor & Bolotova, supra note 10, used the overcharge expressed as a ratio of price increase due to collusion to the benchmark price (formula 2).

²⁶ Before transformation the mean, median, minimum and maximum overcharge values were 26.2, 22.5, –5.0, and 450.1 percent, respectively. After transformation similar statistics are 20.8, 18.4, –5.3, and 81.8 percent, respectively.

²⁷ Overcharge calculated as a percentage of affected sales is the same as overcharge calculated as a ratio of price increase due to collusion to the price during collusion (formula 1).

²⁸ Supra, note 5.

²⁹ U.S. Census Bureau, 1997 Economic Census. Manufacturing: Industry Series: Industry Statistics by Industry and Primary Product Class Specialization, http://factfinder.census.gov/servlet/IBQTable?_bm=y&-geo_id=&-fds_name=E9700A1&-ds_name=E9731I5&-_lang=en.

³⁰ US Economic Census surveys are conducted and published every five years. The modern cartels data set included cartels that operated from 1990 to 2004.

³¹ The number of the U.S. market overcharges is ten observations less than the number reported in Table 2 for the last antitrust law regime. This is because the concentration ratio data were available only for twenty-seven cartels out of thirty-six cartels.

³² There is one negative overcharge (undercharge) estimate (–5.26%) and 13 zero overcharge estimates in the sample of 333 overcharges.

³³ Antitrust law regimes are discussed in a section presenting hypotheses.

³⁴ Given a very limited number of the U.S. market overcharges in the case of modern cartels (twenty-nine observations), we estimate the model using the U.S. and Canadian market overcharges. We controlled for the differences between these two markets using a binary variable.

³⁵ See Stigler, supra note 2.

³⁶ See Sjostrom, supra note 3.

³⁷ Griffin, supra note 8, found a positive relationship between the cartel market share and Lerner index in a sample of international cartels operating legally.

³⁸ Supra note 2.

³⁹ There are various strategies used by this type of cartels. They vary from restricting a number of participating firms in each particular auction to agreeing on the levels of bids. The US Federal Sentencing Guidelines establish one additional offence level beyond the base offence level for firms involved in this type of conduct. This is mostly because it is difficult to estimate the size of affected sales in these cases. See Federal Sentencing Guideline Manual 2R1.1: Bid-rigging, price-fixing, or market-allocation agreements among competitors (2006), http://www.ussc.gov/2006guid/CHAP2-4.html.

⁴⁰ An example of legal cartels is the voluntary Webb-Pomerene cartels that operated under the antitrust exemption of the Webb-Pomerene Export Trade Act and were self-enforced agreements. See Andrew R. Dick, Identifying Contracts, Combinations and Conspiracies in Restrain of Trade, 17 MANAGERIAL DEC. ECON. 203 (1996) and Andrew R. Dick, When are Cartels Stable Contracts? 39 J. L. ECON. 241 (1996). Approximately 45 percent of these cartels were involved in price-fixing, and they could have the same problems with the internal discipline enforcement as illegal cartels, because the power of government was not applied.

⁴¹ For example, in the rare instances of overcharges appearing in the plea agreements, they are likely to be understated.

⁴² An empirical analysis of the impact of different estimation methods on the variability of overcharge estimates is presented in Connor & Bolotova, supra note 10. For example, using price before conspiracy as a benchmark price in calculation of overcharges results in higher overcharge estimates than using the price after conspiracy method.

⁴³ There is empirical evidence suggesting that strong anti-cartel enforcement regimes had a deterrent effect in the case of the global vitamin cartel. Overcharges tended to be lower in the countries with strong antitrust law enforcement, even in the cases when cartels were not prosecuted. See Julian L. Clarke & Simon J. Evenett, The Deterrent Effects of National Anticartel Laws: Evidence from the International Vitamins Cartel, 38 ANTITRUST BULL. 689 (2003).

⁴⁴ Supra note 5.

⁴⁵ A semiparametric regression model does not require an explicit assumption about the probability distribution of the error component. The semiparametric model assumptions only require that the errors are mean zero and uncorrelated. In our case, we do not make further strong assumptions about the error distribution because we are not aware of any specific distributional properties of the error process. The semiparametric model is appropriate when the error distribution may be skewed, as in our case. For further details, see RON C. MITTELHAMMER, GEORGE G. JUDGE & DOUGLES J. MILLER, ECONOMIC FOUNDATIONS (Cambridge University Press 2000). We also considered alternative estimation methods (Maximum Likelihood Tobit estimation procedure) to account for the potential discrete mass point at zero, but this adjustment has little impact on the estimation results. The ML Tobit estimation results are presented in Bolotova, supra note 21.

⁴⁶ An observation is classified as a mild outlier if it belongs to the interval between 57.8% and 87%, representing the upper inner and upper outer thresholds. Also, we examined the data using a plot to locate these outliers. We find that although there is a group of observations that have higher values, these observations are concentrated very close to all other observations. We kept the mild outliers in the sample.

⁴⁷ The Wald statistic P-value is 0.9075. For this test, we estimated the pooled model with the reference group represented by the European market overcharges. In this case, we could introduce a set of interaction effects with binary variable U.S.

CiteULike    Connotea    Del.icio.us    What's this?

This Article Abstract Full Text (PDF) All Versions of this Article: 5/2/361    most recent nhn025v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Request Permissions Google Scholar Articles by Bolotova, Y. Articles by Miller, D. J. Search for Related Content Related Collections K21 - Antitrust Law L10 - General Social Bookmarking          What's this?

Online ISSN 1744-6422 - Print ISSN 1744-6414

Copyright © 2010 Oxford University Press

Oxford Journals Oxford University Press

• Site Map
• Privacy Policy
• Frequently Asked Questions

Other Oxford University Press sites: Oxford University Press Oxford Journals China Oxford Journals Japan American National Biography Booksellers' Information Service Children's Fiction and Poetry Children's Reference Corporate & Special Sales Dictionaries Dictionary of National Biography Digital Reference English Language Teaching Higher Education Textbooks Humanities International Education Unit Law Medicine Music Online Products Oxford English Dictionary Reference Rights and Permissions Science School Books Social Sciences Very Short Introductions World's Classics