Department of Economics Working Paper Series ... - American University

2 downloads 158 Views 495KB Size Report
Contact Information: American University, 4400 Massachusetts Avenue, NW, Washington .... “business proprietary informa
Department of Economics Working Paper Series Under the Cover of Antidumping: Does Administered Protection Facilitate Domestic Collusion? by Kara M. Reynolds January 2013 No. 2013-5

http://www.american.edu/cas/economics/research/papers.cfm

Copyright © 2012 by Kara M. Reynolds. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies.

Under the Cover of Antidumping: Does Administered Protection Facilitate Domestic Collusion?

Kara M. Reynolds, American University1

December 2012 ABSTRACT

Anecdotal evidence suggests that domestic firms can use the antidumping petition process to engage in collusion and increase domestic prices. In this paper, I test whether the antidumping petition process itself can help domestic firms raise prices. I propose a method to identify whether firms in the industry experience a structural break in the level of market power held by the firms at the time that they file their antidumping petition. I then use this methodology to analyze the impact of antidumping petitions on competition levels in two industries. I find little evidence that either of these industries increased their market power following the filing of petitions for trade relief, nor even from the protection that resulted from these petitions, suggesting that the widespread belief that antidumping leads to more market power may not always hold.

Keywords: Antidumping, Collusion JEL Codes: F13

1

Contact Information: American University, 4400 Massachusetts Avenue, NW, Washington, DC 20016, Telephone: (202) 885-3768, E-mail: [email protected].

1

I.

Introduction

In 1989, three U.S. producers began devising a way to create a price-fixing cartel in the ferrosilicon industry (Pierce (1999)). They decided to refuse to sell at prices below their domestic cartel price; when relatively inexpensive imports from Asia and South America began to flood the market, they filed antidumping petitions in both the United States and European Union. Under U.S. antidumping law, firms may request antidumping tariffs be imposed on products imported from specific countries at less than “normal” or “fair” value which are causing (or threatening to cause) material injury to the domestic industry. Broadly, “normal” value is defined as the price set by the foreign exporter in its domestic market, although all prices set below the firm’s average cost of production are excluded from the calculation. Thus antidumping is designed to combat both price discrimination and predatory pricing on the part of foreign exporters. In this case, the ferrosilicon producers were awarded antidumping protection in 1993. Shortly thereafter, U.S. producers offered Brazilian producers a chance to join the cartel; when the Brazilian producers refused, the U.S. producers filed another successful antidumping complaint against Brazil. Luckily, this story has a happy ending-- authorities detected and prosecuted the cartel behavior.1 Although this is certainly an extreme example, it does beg the question to what extent firms use U.S. trade laws to engage in uncompetitive behavior in the domestic market. The ferrosilicon case is an example of an overt attempt to engage in illegal, monopolizing activities. In a similar example, a number of U.S. firms in the polyester staple fiber industry were charged with illegal price fixing shortly after the industry filed for, and was awarded, antidumping protection against imports from Korea and Taiwan in April 1999.2 The cases described above suggest that the antidumping petition and investigation process reveals a significant amount of firm-specific price and cost information, which could encourage firms to engage in anticompetitive pricing activities. Note that under the Noerr-Pennington doctrine, firms are immune from prosecution under federal antitrust laws due to efforts to influence the enforcement of laws, including antidumping laws. 3 In other words, firms are allowed to collaborate on antidumping petitions without fear of being charged under antitrust statutes. The antidumping petitions must have merit, however, in order to qualify for the Noerr-

1 The U.S. Justice Department investigation resulted in a 1995 plea agreement with one firm (Elkem Metals) and a 1997 conviction of another firm and one of its senior officials (SKW Metals and Alloys). In light of these criminal activities, the U.S. International Trade Commission rescinded the antidumping order in 1999. 2 In this case, an official from Kosa pleaded guilty to price fixing charges as early as September 1999 and as late as 2001, while an official from NanYa was eventually acquitted of similar charges. The International Trade Commission did not find the evidence of price fixing to be compelling enough to review the antidumping case. 3 Antidumping is certainly not the only legal procedure that could be used to encourage anticompetitive behavior. Klein (2011) conducts a general analysis of “sham litigation,” or lawsuits brought solely to gain a competitive advantage in violation of antitrust laws.

1

Pennington exception, and several suits have been filed by foreign producers against U.S. firms claiming that their antidumping petitions were baseless and essentially “sham” litigation.4 A number of authors have studied the degree to which antidumping law may lead to increased collusion among firms, although most of these authors have focused on the use of antidumping law to facilitate collusion between domestic and foreign firms. For example, Prusa (1992) first noted that nearly 40 percent of U.S. antidumping cases filed between 1980 and 1985 were withdrawn prior to the government reaching a final determination; moreover, data suggested that these withdrawn cases restricted trade by at least as much as cases that resulted in the imposition of antidumping duties. To explain this data, Prusa (1992) developed a model in which both domestic and foreign firms benefitted from reaching an out of court settlement to raise prices in the wake of an antidumping filing. Zanardi (2004) finds empirical support for a similar bargaining model in which the ability of the domestic and foreign firms to reach an out of court settlement is hampered by coordination costs and the degree of bargaining power of the two sides. While neither Staiger and Wolak (1994) nor Taylor (2004) found empirical evidence that withdrawn petitions decreased the quantity or increased the price of imports in the United States, Rutkowski’s (2007) estimates suggested that there are collusive aspects associated with withdrawn cases in the European Union.5 Two theoretical models suggest that the existence of antidumping laws itself, including the associated revelation of information during the investigation process, could lead to increased collusion among both domestic and foreign firms. Although Hartigan (1995) develops a model in which one domestic and one foreign Bertrand competitor are unsure as to whether their rival is a high- or a low-cost producer, this model could easily be extended to one with multiple domestic firms. In this model, firms use the antidumping investigation to discover information about their rivals; while the domestic firm’s costs are revealed through the injury investigation, the foreign firm’s costs are revealed through the dumping investigation. In this model, high-cost firms will benefit from the revelation of information as prices increase, while low-cost firms will experience a loss of profits.

4

For example, in 1999 Cheminor Drugs filed suit against Ethyl Corporation claiming that Ethyl’s antidumping petition against Cheminor (which Ethyl withdrew prior to a final determination) was baseless, and filed solely for anticompetitive reasons in violation of antitrust laws. A court of appeals rejected Cheminor’s case, citing the fact that the International Trade Commission made a preliminary affirmative injury determination in the case. Similar allegations were made against Prestini Musical Instruments Corp. for their antidumping filings against Italian producers of musical instrument parts in 1983. The court also rejected this case. 5 Stegemann (1990) studies a similar class of antidumping cases in which the government reaches an “out of court” settlement with the foreign industry, those in which the foreign industry officially agrees to raise prices (a price undertaking) rather than face antidumping duties. Based on his sample of EU cases, Stegemann concludes that while these price undertakings are a legal substitute for illegal price fixing at some level, there is no public evidence that firms are using these undertakings to engage in collusion with their foreign counterparts in excess of that allowed for under antidumping law.

2

Veugelers and Vandenbussche (1999) develop a theoretical model to explore how antidumping law might change the incentives and profitability of domestic and/or international cartels, focusing specifically on European antidumping law.6 The authors find that antidumping legislation can be either pro-competitive or anti-competitive, depending on how the government makes its antidumping determination and the level of competition in the industry. For example, if the European Union chooses to impose antidumping duties, the model predicts that the market will always become less competitive with the tariffs encouraging the formation of cartels. Surprisingly, there are very few empirical studies that have focused on the degree to which antidumping laws change the level of competition in the domestic industry. Hartigan et al. (1989) conduct an event study methodology to evaluate to what degree financial markets believe antidumping protection will improve the competitive positions of firms. The authors find that the market only values antidumping protection when the ITC determines that protection is warranted due to the threat of injury from foreign imports, not when the ITC finds that protection is warranted because of actual injury from foreign imports. Nieberding (1999) calculates firm-level market power statistics (based on the Lerner Index of market power) for four U.S. industries that filed antidumping petitions in the 1980s: semiconductors, hydraulic cement, tapered roller bearings, and the steel industry. The author finds that firms have significantly more market power following successful antidumping petitions which result in duties (semiconductors and ball bearings), but less market power following unsuccessful petitions (hydraulic cement). Konings and Vandenbussche (2005) use a panel of 4,000 European producers who were involved in antidumping petitions in 1996 to estimate whether these investigations increased firm markups. They find that successful antidumping petitions result in an average increase in mark-ups of domestic firms of approximately 8 percentage points. They found no impact of terminated, or unsuccessful, antidumping petitions on average firm mark-ups, suggesting that it is the protection itself that increases firm’s market power rather than simply filing a petition. The results from Nieberding (1999) and Konings and Vandenbussche (2005) suggest that it is only successful antidumping petitions that result in increased market power in the domestic industry; in other words, the increased market power must be due to the fact that the antidumping duties restrict imports. In this paper, we instead explore the possibility whether the petition and investigation process itself can reduce competition in the domestic market by making it easier for firms to exchange information. To do this, I first I propose a method to identify whether 6

Although the basic tenants of U.S. and European antidumping law are the same, there are some notable differences. For example, European antidumping petitions more often result in undertakings, in which the foreign firms agree to raise their prices. European antidumping law also specifies that the government must weigh the total welfare effects of the antidumping duties, explicitly taking into account the interests of consumers as well as producers.

3

domestic firms in the industry experience a structural break in their level of market power at the time that they file their antidumping petition. I then use this methodology to analyze the impact of antidumping petitions on competition levels in two industries, semiconductors and ball bearings. I find little evidence that the domestic firms in these industries experienced a significant increase in market power from either the antidumping petition process or from the protection itself. II.

Antidumping Investigations in the United States

Under U.S. antidumping law, any interested party can file a petition on behalf of a U.S. industry requesting that antidumping tariffs be imposed on a specific product from a specific foreign country.7 As noted above, the petition must claim that products imported from specific countries are being sold at less than “normal” or “fair” value, and these imports are causing or threatening to cause material injury to the domestic industry. The petition includes a significant amount of information which may be useful to firms engaging in uncompetitive pricing behavior. For example, in other to determine whether the petitioner(s) has standing to file the petition, the petition must include the names and addresses of all petitioners, as well as all other U.S. producers. The petition must also include the total volume and value of U.S. production of the good, as well as the volume and value of each producer listed as supporting the petition. In order to determine whether there is evidence that the domestic industry has been injured by imports, the petition should include information on a things such as declining domestic prices, reduced levels of production, reduced levels of capacity utilization, declining net sales and market share, sales lost to imports, declining profitability, reduced levels of employment, and bankruptcy. Firms can indicate that certain information in the petition is “business proprietary information;” this information is not revealed in any public documents associated with the investigation. In this case, firms are requested to provide a summary of information that could be included in public documents. Virtually all industries hire law firms to prepare their antidumping petitions, thus firms would not necessarily have to reveal their business proprietary information to their competitors, even their fellow petitioners. Once accepted by the government, investigations are undertaken by two agencies. The U.S. International Trade Commission (ITC) must determine whether dumped imports are causing or threatening to cause material injury to the domestic industry. The ITC makes its determination on the basis of information included in the petition, questionnaires sent to foreign and domestic producers and U.S. importers, and information presented by interested parties in legal briefs and during public hearings. Factors taken under consideration include declines in output, sales, market shares, profits, productivity, the return on investments, and capacity utilization, as well as 7

An antidumping petition is deemed to have been filed on behalf of the domestic industry if domestic producers and workers who support the petition account for (1) 25 percent of the total production of the product and (2) 50 percent of the production of those who both support and oppose the petition.

4

the negative effect the imports under investigation may have had on cash flow, inventories, employment, wages, growth, the ability to raise capital, and investment. Meanwhile the International Trade Administration (ITA) of the Department of Commerce is charged with calculating the degree to which foreign firms are selling their products in the United States at less than normal value. Specifically, the ITA calculates a dumping margin equal to the difference between the normal value and the U.S. price. To do this, the ITA collects detailed price and cost information from foreign exporters in order to calculate the dumping margin. If both the ITA and ITC make affirmative determinations and an antidumping duty order is put into effect, importers must pay a cash deposit equal to the dumping margin for all imports to the United States. Between 1980 and 2009, there were 545 antidumping petitions filed in the United States.8 Although a handful of these petitions were initiated by the government or unions (1.2 percent), most were filed by domestic firms. Approximately 7 percent of petitions were filed by trade associations which declined to publicize the names of their member firms. Almost half of all petitions were filed by a single firm, and nearly three-quarters of all petitions were filed by three or fewer firms. Not all firms in the industry participate in the petition process. Reynolds (2005) finds that U.S. antidumping petitions were filed on average by 33 percent of the total number of U.S. producers of the product between 1980 and 1996. The distribution in part reflects the fact that it is more likely for firms in highly concentrated industries to file antidumping petitions. As explained in Reynolds (2005), only those firms filing the petitions have to pay the somewhat significant legal costs associated with process while all firms in the industry would benefit from the antidumping protection if awarded. Highly concentrated industries, in which the benefits of protection are also highly concentrated, are more likely to overcome this free-riding deterrent. The anecdotes discussed in the introduction suggest that antidumping investigations are more likely to result in anticompetitive behavior among firms actually filing the antidumping petition—those who were in close contact throughout the process of preparing the petition. In order to explore this topic more fully, I first propose an empirical method to identify whether there is a structural break in the level of market power held by firms benefitting from antidumping protection, whether at the time the petition is filed or at the time the protection is 8

Petitions often request that antidumping duties by imposed on more than one country, and the Department of Commerce and International Trade Commission officially launch a separate investigation for each country targeted in the petition. Thus these 545 petitions resulted in 1,200 antidumping investigations, of which 44 percent resulted in the imposition of antidumping duties and 10 percent were withdrawn prior to the government making a determination.

5

imposed. I then apply this methodology to the two industries that Nieberding (1999) identified as having benefitted from antidumping protection: semiconductors and tapered roller bearings. III.

Empirical Methodology and Data

I follow Nieberding (1999) in looking for evidence of collusion by examining whether there is a change in the degree of market power exhibited by U.S. firms at the time that the firms filed antidumping petitions. Like Nieberding (1999), I use the empirical model initially derived in Martin (1988), who proposed that an oligopolistic industry faces a constant elasticity demand curve: (1) where qi is the output of firm i, Q-i is the output of the industry less that of firm i, and ε is the elasticity of demand. Firms form expectations about the behavior of their competitors according to a conjectural elasticity term: .9

(2)

Profit maximization using equations (1) and (2) results in the following expression for the firm’s marginal revenue (MR): (3) where MSi is the market share of firm i. The firm does not exercise market power if marginal revenue is equal to price, or if: (4) Martin (1988) develops an empirical model to test for firm market power; his final specification, which is derived from the firm’s profit maximization problem assuming that the firm has a constant returns to scale technology is equivalent to: (5) where PCM is the firm’s price-cost margin as measured by its profit to sales ratio, KSR is the firm’s capital stock to sales ratio, ISR is the firm’s interest expense to sales ratio, and NSR is the ratio of the firm’s receipts from new stock issue to sales. Comparing equation (5) to equation (4), one can use the parameters from the empirical model to test for firm market power. Specifically, if: 9 This is admittedly a non-trivial assumption. Corts (1999) demonstrates that conduct parameter methods can result in severe mis-measurement if the underlying firm behavior is not the result of a conjectural variations equilibrium.

6

(6) then the firm exhibits some market power, and the larger this combination of coefficients and market share, the greater the market power of the firm. Structural Breaks Nieberding (1990) specifically tests whether firms exhibited evidence of holding more market power after the resolution of their antidumping petitions than in the period prior to the resolution. I will try to find evidence of whether there was a clear structural break in the degree of market power of these firms, as measured by equation (6), and whether this structural break occurred prior to the resolution of the antidumping petition, which may indicate that the firms were able to use the information exchanged during the petition process to increase their market power. Alternatively, if there is either an initial or an additional increase in market power after the imposition of protection, then the protection itself changed the market position of the firms. Prior to formally testing the model for structural breaks, it is useful to visualize how the measure of market power has changed over the sample period in each firm. To do this, I first run rolling regressions of equation (5) for each firm and estimating the market power statistic given by equation (6) at time period t using a window of observations around period t. In the figures presented here, I use a rolling regression window of 30 quarters. To more formally test for the presence of a structural break in the coefficients of the model, I conduct the test suggested in Bai and Perron (1998). The primary benefit of this procedure is that it allows one to consider linear models in which there may be more than one structural break in one or more of the coefficient values. Specifically, the model is expressed as: (6) Where PCM is the vector of the price cost margins for firm i, X is the matrix of explanatory variables, and Z is a block diagonal matrix that divides the explanatory variables into m periods, where m is the number of structural breaks found by the model. The estimation procedure first estimates the parameters of the model β and δ using least squares for each possible partition or break at time Tj in the model. Putting these parameter estimates into the model, the estimated break points are defined as: (7) where ST is the sum of square residuals, and the minimization is taken over all possible break points. In other words, the estimated break points are the global minimizers of the objective function. Bai and Perron (1998) more fully describes how the model can be estimated sequentially, the properties of the model, and the distribution of the statistic that allows one to test for the possibility of l versus l+1 breaks in the model.

7

All of the firm-specific variables used in this analysis were obtained from Compustat (quarterly observations) for the years between 1975 and 2011. As a result, the analysis is limited to firms that were public at the time that the industry filed its antidumping petition. Appendix A provides a more detailed description of the calculation of each variable. IV.

Results

In this section, I apply this technique to firms within the two industries that Nieberding (1999) found benefitted from successful antidumping petitions, the semiconductor industry in 1985 and the tapered roller bearing industry in 1986. 1. The U.S. Semiconductor Industry The semiconductor industry, both in the United States and abroad, has benefitted from a wide variety of government policies over the years, including government support for science and R&D. Perhaps because of the importance of government funding to the semiconductor industry, five of the leading U.S. semiconductor firms (including Fairchild Semiconductor, Intel, AMD, National Semiconductor and Motorola) founded the Semiconductor Industry Association (SIA) in 1977. According to Tyson (1999), although the members of SIA were fierce competitors, the SIA allowed them to address their shared interests, including the increasing competition from Japanese manufacturers. To address this increasing competition, the SIA filed a Section 301 petition with the U.S. government in June of 1985 in order to gain access to the Japanese semiconductor market, charging that Japanese producers were engaging in unfair trade practices. Section 301 of the U.S. Trade Act of 1974 authorizes the President to take action (including potential retaliation) against any trading partner that violates an international trade agreement or unfairly restricts U.S. commerce. Industries can file Section 301 petitions with the U.S. Trade Representative’s office requesting that action be taken against a specific foreign country. Although today Section 301 complaints are brought to the Dispute Settlement Body of the World Trade Organization, throughout the 1980s these complaints typically resulted in unilateral action on the part of the United States or bilateral side agreements between the U.S. and the trading partner in question. Members of the SIA were hesitant to file antidumping petitions against Japanese producers at this time, fearing that an antidumping petition would weaken support among U.S. end-users of semiconductors for the industry’s Section 301 petition (Tyson, 1999). Nevertheless, on July 3, 1985, Micron Technology broke ranks from the SIA, filing an antidumping petition (Case 731TA-270) against Japanese producers of 64K Dynamic Random Access Memory (DRAM) chips. At the time of the petition, there were 11 firms producing DRAMS in the United States.10 Shortly thereafter, Intel Corp., AMD, and National Semiconductor filed an antidumping petition (Case 731-TA-288) against Japanese producers of erasable programmable read only memory

10

U.S. DRAM producers included Advanced Micro Devices (AMD), AT&T Technology, Fujitsu, Hitachi, IBM, Mitsubishi, Mostek, NEC Electronics, Texas Instruments, and Toshiba. Three additional firms had just recently ceased production of DRAMS, including Intel, Motorola, and National Semiconductor.

8

chips.11 On December 17, 1985, the Department of Commerce self-initiated an antidumping petition (Case 731-TA-300) against Japanese producers of 256K DRAMS. A complete timeline of these cases is presented in Table 1. Although final antidumping duties were initially imposed against Japanese 64K Drams on June 16, 1986, these duties were lifted on August 7, 1986 when the United States and Japan reached a suspension agreement known as the 1986 Semiconductor Trade Agreement. The agreement specified, among other things, that the Department of Commerce could establish foreign market values based on Japanese cost information. If Japan exported products to the United States at below these foreign market values, then Japan would be presumed to be dumping and the dumping duties could be reinstituted.12 The Agreement had fairly mixed results on firms within the U.S. industry. As reported in Tyson (1999), the prices of DRAMS began to increase in the second quarter of 1987 and throughout the end of 1988, at which point they began to decline again. Nevertheless, DRAM production continued to decline in the United States, and by 1990 only Texas Instruments and Micron (which specialized in DRAM production) remained in production. Although prices of EPROMs did not significantly increase over this same time period, National Semiconductor, SGS-Thomas and Texas Instruments all entered or re-entered the market following the Agreement. On July 31, 1991, the original Semiconductor Trade Agreement was extended for an additional 17 months, but it eliminated the dumping provisions of the original agreement. Instead, Japanese firms were required to collect cost information in case the United States decided to pursue a “fast track” antidumping investigation.13 It is likely that U.S. producers of DRAMS and EPROMS benefitted from the dumping provisions of the Semiconductor Trade Agreement. Of interest in this study, however, is whether they benefitted from the act of simply preparing the Section 301 petition and the 1985 antidumping petition filed by Intel Corp., AMD, and National Semiconductor. Specifically, did the process of preparing these petitions allow the firms involved in them to exchange information, thus encouraging at least implicit collusion in the industry? 14, 15 In order to answer this question, I estimate the model proposed in Section 3 on five of the leading public companies producing semiconductors in the United States, including AMD, Intel Corp., Micron, National Semiconductor and Texas Instruments. 11

Other U.S. EPROM producers at this time included Fujitsu, Mostek (a division of United Technologies until 1985, at which time it was acquired by Thomson SA), Motorola, Rockwell International, SEEQ Technology, and Texas Instruments. 12 The agreement also included provisions for Japan to provide sales assistance to U.S. companies trying to export to Japan, with the goal of U.S. producers achieving a 20 percent market share within five years (Tyson, 1999). 13 Micron again asked for antidumping protection from foreign producers of semiconductors: antidumping duties were imposed against Korean producers of DRAMS on May 10, 1993 and against Taiwanese producers of SRAMS on April 16, 1998. These duties were revoked on October 5, 2000 and January 14, 2002, respectively. 14 Because the semiconductor industry already had a long history of cooperation to prior to 1985, one may not expect to see a structural break in their level of competition due solely to their coordination on this trade relief petition. 15 Tyson (1999) reports that there is some evidence, including pricing rules and a restraint on investment, that the Agreement may have led to cooperative behavior (or collusion) among Japanese firms in 1988 and 1989.

9

The results rolling regressions are pictured in Figures 1 (a)-(e). The results from the rolling regression provide little evidence that the firms in the semiconductor industry benefitted simply from the act of filing the Section 301 and antidumping petition. Although the graphs do suggest some change in market power during this time period, visually it appears as if these changes occurred due to the protection itself, which would be consistent with the results found in Nieberding (1999). For example, both ADM and Intel appear to exhibit an increase in market power in the third quarter of 1986, the same time that the Semiconductor Agreement went into effect. Micron (who filed the DRAM petition on its own) exhibits some increase in market power starting in the fourth quarter of 1985, which is when preliminary antidumping duties were imposed on Japanese producers of DRAMS. Visually, National Semiconductor appears to exhibit some increase in the degree of market power starting in the fourth quarter of 1984—the same time at which it, AMD and Intel were preparing their antidumping petition against Japanese producers of EPROMS. One must take these visualizations with a grain of salt, however, as the changes in the coefficients exhibited in the rolling regressions may be due to increased sampling error. The results from the Bai and Perron (1998) test for structural breaks are summarized in Tables 2(a) and (b). The table reports both the estimated number of structural breaks in the parameters of the model, along with the estimated break dates and the 95 percent confidence interval associated with these break dates. The table also reports the market power statistic as calculated from equation (6) associated with each firm and time period. These market power statistics were calculated from the regression coefficients reported in Tables 3(a) – 3(b). The Bai and Perron (1998) test found evidence of two structural breaks in the parameters of the model for ADM, Micron, National Semiconductor, and Texas Instruments; the test found evidence of three structural breaks in the parameter estimates associated with Intel. However, none of these structural breaks occurred at the same time across firms, making it unlikely that the firms were engaging in any coordinated effort to either increase prices or, alternatively, decrease prices in an effort to ensure approval of the antidumping protection. The strongest evidence that firms may benefit simply from the act of filing an antidumping petition comes from Intel, which exhibit a significant increase in market power (the market power statistic increased from 0.513 to 0.796) in the first quarter of 1985, the quarter before the Semiconductor Industry Association filed the Section 301 petition. The boost in Intel’s market power may be unrelated to the antidumping petition, and more related to changes in the firm’s business strategy. According to Intel’s 1986 annual report, in 1985 the firm phased out their DRAM line entirely in order to focus entirely on EPROMs, a product in which they had historically been a market leader and technological innovator. None of the other firms in the industry, including the two other firms that filed the EPROM petition in conjunction with Intel (ADM and National Semiconductor) exhibit a similar increase in market power at or around the time that the antidumping petition was filed. Indeed, ADM did not exhibit an increase in market power until the first quarter of 1990, well after the

10

Semiconductor Agreement was signed, and National Semiconductor did not exhibit any significant change in its market power statistic during the sample period. Although Micron, the instigator of all of the U.S. antidumping petitions against foreign producers of DRAMS, exhibited structural breaks in its parameter estimates in the first quarter of 1997 (the quarter it filed its cases against Korean and Taiwanese producers) and in the third quarter of 2000 (the quarter in which duties on Korean DRAMS were eliminated), the firm did not exhibit a statistically significant change in its market power statistic over the course of the sample period. Although the parameter estimates associated with Texas Instruments suggest that this firm experienced an increase in market power during the middle period (second segment) of the sample, this break occurred long after both the filing of the antidumping petition and the signing of the Semiconductor Trade Agreement in the third quarter of 1995. 2. Ball Bearings A single, dominant firm in the industry, Timkin Co., filed the second petition analyzed in this paper, tapered roller bearings. However, nine other U.S. firms manufactured the product in question at the time of the petition, including three that were publically listed at the time: Brenco Inc., Ingersoll-Rand (the parent company of Torrington), and SKF. Note that the ball bearing industry had benefitted from antidumping protection against tapered roller bearings from Japan since 1975. In 1984, the U.S. chose not to impose antidumping protection against similar products from Germany, Italy and Japan and in 1986 the Department of Commerce recommended rescinding the original 1975 antidumping order against one of the leading Japanese firms. At approximately the same time, Timkin Co. filed an antidumping petition against imports of tapered roller bearings from Hungary, Italy, Japan, China, Romania and Yugoslavia. As noted in Table 4, preliminary antidumping duties were imposed on products from all countries on February 6, 1987. Between June and October of 1987, final antidumping duties were imposed against all of the targeted countries. While the antidumping orders against Yugoslavia and Italy in 1995 and 1996, respectively, other orders were in place until 2000. The industry later benefitted from another series of successful antidumping petitions filed in 1988. Because of missing variables, I was only able to run the complete model on Timkin and Ingersoll-Rand. Graphs using the rolling regression methodology are presented in Figure 2, while the results from the Bai Perron structural break model are presented in Tables 5 and 6. Results from the rolling regressions do not indicate that either firm experienced a significant increase in market power at the time of the 1986 imposition of antidumping protection. Indeed, if anything the results associated with Ingersoll Rand suggest a gradual decrease in the market power of the firm from the late 1970s until about 2001, when it shoots up. Note that in 2001 the United States passed the Continued Dumping and Subsidy Offset Act (CDSOA), which distributed the antidumping duties back to the domestic firms that supported the original petitions. The ball bearing industry, including Torrington and Timkin, was the leading 11

beneficiary of this law, collecting $74.6 million dollars in 2001 alone. Timkin appears to experience a decrease in market power around the time the antidumping duties were imposed. These results are confirmed using the Bai-Perron methodology. Results indicate that Torrington had a single structural break, in the third quarter of 2006, at which point they experienced a slight decrease in market power. Interestingly, this is approximately the same time that the CDSOA was repealed by Congress. Timken exhibited evidence of two structural breaks, one at the time of the antidumping petition (the third quarter of 1986), and a second in the third quarter of 1993. My analysis shows that Timken experienced a decrease in market power in the third quarter of 1986, which could have prompted them to file the antidumping petition in the first place. Importantly, there is no evidence that Timken experienced a subsequent increase in market power once the antidumping duties were put into place in February of 1987.16 V.

Conclusion

Anecdotal evidence suggests that domestic firms can use the antidumping petition process to engage in collusion and increase the domestic price. Indeed, antidumping petitions require domestic firms to collect a significant amount of data about the firms in their industry which could lead, whether intentionally or unintentionally, to greater collusion among domestic firms. Previous literature has confirmed that antidumping petitions, and particularly successful antidumping petitions, raise the market power of domestic firms, but this increase in market power is typically attributed to reduced competition from foreign firms. In this paper, I test whether the antidumping petition process itself can help domestic firms raise prices. Using quarterly data from public firms in the U.S. semiconductor and tapered roller bearing industries between 1975 and 2011, I analyze the data for structural breaks using the tests suggested in Bai and Perron (1998). Although the results of this analysis do suggest that firms within both industries exhibit structural breaks in the determinants of the price/cost margin over the sample period, there is little evidence that these breaks occurred at the time that the firms were preparing their antidumping petitions or during the course of the antidumping investigation. I conclude that the antidumping petition process did not increase the likelihood of collusion among domestic firms, at least among the firms in these two samples.

16

This result contradicts the results in Nieberding (1999), who found that Timkin experienced an increase in the market power statistic in the period after antidumping protection (between January 1988 and December 1992). I was unable to replicate Nieberding’s (1999) results even after restricting my sample to the same time period and including a single break in January 1988. This may be due to the vastly different market share values used in the analysis. As noted in the Appendix, I calculated market share using two different denominators: the sum of sales reported in Compustat by the five public firms at the time of the antidumping petition who produced the product under investigation and the average quarterly sales in the ball bearing industry as calculated using the data from the U.S. Census Bureau’s Annual Survey of Manufacturers were qualitatively similar to those presented here. Although the results presented here use the former definition, results from the later were qualitatively similar. Neither market share definition replicates those reported in Nieberding’s (1999) summary statistics, and the author did not report how he calculated the denominator of his market share statistic.

12

Perhaps more surprisingly, I also find little evidence that the domestic firms experienced an increase in market power following the imposition of protection itself (or in the case of the semiconductor industry, the suspension agreement reached by the United States and Japan). Indeed, although most firms in the sample exhibited evidence of an increase in market power later in the sample, these structural breaks came long after the imposition of protection, and occurred at different times for different firms. The results suggest that perhaps earlier studies which have found evidence of an increase in market power should be reevaluated at exactly when this increase occurred, and whether it can be attributed specifically to the imposition of antidumping protection.

13

References Bai, Jushan and Pierre Perron (1998), “Estimating and Testing Linear Models with Multiple Structural Changes,” Econometrica 66(1): 47-78. Corts, Kenneth S. (1999), “Conduct Parameters and the Measurement of Market Power,” Journal of Econometrics 88: 227-250. Hartigan, James C. (1995), “Collusive Aspects of Cost Revelation Through Antidumping Complaints,” Journal of Institutional and Theoretical Economics 151 (3): 478-489. Hartigan, James C., Kamma, Sreenivas, and Perry, Philip P. (1989), “The Injury Determination Category and the Value of Relief from Dumping.” Review of Economics and Statistics 71(1): 183-186. Hansen, Bruce E. (2001). “The New Econometrics of Structural Change: Dating Breaks in U.S. Labor Productivity.” Journal of Economic Perspectives 15(4): 117-128. Klein, Christopher (2011), “Deregulation and the Demise of Sham Litigation: A Structural Time Series Test.” Working Paper. Konings, Jozef and Hylke Vandenbussche (2005), “Antidumping Protection and markups of domestic firms,” Journal of International Economics 65: 151-165. Martin, Stephan (1988). “The Measurement of Profitability and the Diagnosis of Market Power,” International Journal of Industrial Organization 6: 301-322. Nieberding, James F. (1999), “The Effect of U.S. Antidumping Law on Firms’ Market Power: An Empirical Test,” Review of Industrial Organization 14: 65-84. Pierce, Richard J. (1999), “Antidumping Law as a Means of Facilitating Cartelization,” Antitrust Law Journal 67: 725-743. Prusa, Thomas J. (1992), “Why are so many antidumping petitions withdrawn,” Journal of International Economics 33: 1-20. Reynolds, Kara (2005), “Free Riders Among the Rent Seekers: A Model of Firm Participation in Antidumping Petitions.” Working Paper.

14

Rutkowski, Alexsander (2007), “Withdrawals of Antidumping: Complains in the EU: A Sign of Collusion,” World Economy 470-503. Staiger, Robert W. and Wolak, F.A. (1994), “Measuring Industry Specific Protection: Antidumping in the United States.” In M.N. Bailey, P.C. Winston (ed). Brookings Papers on Economic Activity: Microeconomics. The Brookings Institution, Washington, DC, pp. 51-118. Stegemann, Klaus (1990), “EC Antiduping Policy: Are Price Undertakings a Legal Substitute for Illegal Price Fixing,” Weltwirtschaftliches Archiv 126(2): 268-298. Taylor, Christopher (2004), “The economic effects of withdrawn antidumping investigations: is there evidence of collusive settlements?” Journal of International Economics 62: 295-312. Tyson, Laura D’Andrea (1992), “Managing Trade and Competition in the Semiconductor Industry,” in Who’s Bashing Whom? Trade Conflict in High-Technology Industries. Washington, DC: Institute for International Economics. Veugelers, Reinhilde and Hylke Vandenbussche (1999), “European anti-dumping policy and the profitability of national and international collusion,” European Economic Review 43: 1-28. Zanardi, Marizio (2004), “Antidumping Law as a Collusive Device,” Canadian Journal of Economics 37(1): 95-122.

15

Appendix A As noted in the text, all variables were obtained from Compustat’s quarterly data. Specific definitions of each variable are included below. Price-Cost Margin (PCM): Pre-Tax Income (piq) / Total Revenue (revtq) Market Share (MS): Firm Sales (saleq) / Total Sales. Although the Compustat variable includes the firm’s global sales, I calculate the market share in the semiconductor industry by dividing by U.S. domestic semiconductor sales (i.e. those sales made by all firms (foreign and domestic) in the U.S. market, as reported by the Semiconductor Industry Association, in order to make sure I capture the sales of all competitors, not just the public company competitors. The results presented for the ball bearing industry calculates total sales by adding the sales reported in Compustat of the five public firms at the time of the antidumping petition who produced the product under investigation. Results using average quarterly sales in the ball bearing industry as calculated using the data from the U.S. Census Bureau’s Annual Survey of Manufacturers were qualitatively similar to those presented here. Interest Expense to Sales Ratio (ISR): Interest and Related Expenses (xintq) / Firm Sales (saleq) Capital Stock to Sales Ratio (KSR): Total Property, Plant and Equipment (ppentq) / Firm Sales (saleq). Receipts from new Stock Issue to Sales Ratio (NSR): After 1983, I calculate this variable as the difference between the Sales of New Common Stock (sstky) and Purchase of New Common Stock (prstkcy) divided by firm sales (saleq). Prior to 1983, I had to calculate this variable using the change in outstanding shares (shares) multiplied by share price at the end of the quarter. For each firm, I took into account stock splits in calculating this variable.

16

Figure 1: Rolling Regression of Market Power Statistics (Semiconductors) (a)

Advanced Micro Devices (AMD)

Advanced Micro Devices (AMD) 1

1998q4

1987q2

-.5

0

Power

.5

1986q3

1975q1

1980q1

1985q1

1990q1 Year/Quarter

(b)

1995q1

2000q1

2005q1

1995q1

2000q1

2005q1

Intel Corp.

.6

1982q2

.4

1986q3

1984q1

.2

Power

.8

1

Intel Corp.

1975q1

1980q1

1989q3

1985q1

1990q1 Year/Quarter

17

(c)

National Semiconductor

1991q3

.5

Power

1

1.5

2

National Semiconductor

0

1983q1

-.5

1984Q4

1975q1

1980q1

1985q1

(d)

1990q1 Year/Quarter

1995q1

2000q1

2005q1

Micron Technology

.6 .2

.4

1994q2

1985q4

0

Power

.8

1

Micron Technology

1982q3

1996q3

1988q1

1993q3 Year/Quarter

18

1999q1

2004q3

(e)

Texas Instruments

.5

Power

1

1.5

Texas Instruments

0

1992Q1

1975q1

1980q1

1985q1

1990q1 Year/Quarter

19

1995q1

2000q1

2005q1

Figure 2: Rolling Regression of Market Power Statistics (Ball Bearings) (a) Torrington

.4

1984q4

0

.2

Power

.6

.8

Ingersoll-Rand (Torrington)

1975q1

1980q1

1985q1

1990q1 Year/Quarter

1995q1

2000q1

2005q1

1995q1

2000q1

2005q1

(b) Timkin

.4 .3 .2

Power

.5

.6

Timkin Co.

1987q1

1975q1

1980q1

1985q1

1990q1 Year/Quarter

20

Table 1 Timeline of Protection: U.S. Semiconductor Industry (1985-1992) Date Action June 1985  SIA files Section 301 Petition July 1985  Micron Files AD Petition Against Japanese producers of 64K Drams (731-TA-270) October 1985  Intel, AMD, and National Semiconductor file AD Petition Against Japanese producers of EPROMS (731-TA-288) December 1985  Preliminary AD Duties imposed in 731-TA-270  Department of Commerce Self-Initiates Case against Japanese producers of 256K DRAMS (731-TA-300) March 1986  Preliminary AD Duties imposed in 731-TA-288 and 731-TA-300 June 1986  Final AD Duties Imposed in 731-TA-270 August 1986  Suspension Agreement reached in 731-TA-300, 731-TA-288, and 731TA-270  U.S. and Japan enact the Semiconductor Trade Agreement April 1987  Sanctions imposed by the United States for violations of the Semiconductor Trade Agreement. August 1991  Suspension Agreement lifted in 731-TA-300  A limited semiconductor Trade Agreement extended for 17 months, but it excluded all dumping provisions

21

Table 2(a) Market Power Statistics: ADM, Intel and National Semiconductor1 National ADM Intel Semiconductor No. of Breaks 2 3 2 Initial Market Power 0.357** 0.513** 0.281** (0.051) (0.025) (0.056) Break Date 1 Market Power After Difference

Break Date 2 Market Power After Difference

1990Q1 (1987Q4-1990Q2) 0.489** (0.065) 0.132** (0.047)

1985Q1 (1984Q4-1985Q2) 0.796** (0.032) 0.282** (0.028)

1994Q2 (1990Q4-1994Q3) 0.421 (0.276) 0.139 (0.263)

2005Q2 (2003Q2-2005Q3) 0.046 (0.075) -0.443** (0.087)

1994Q4 (1994Q3-1995Q1) 0.344** (0.084) -0.451** (0.094)

1997Q4 (1997Q3-2000Q1) 0.352 (0.329) -0.068 (0.586)

Break Date 3

2003Q2 (2002Q3-2003Q4) Market Power After 0.425** (0.065) Difference 0.081* (0.048) 1 Standard errors in parentheses. ** indicate those market power statistics significant at the 95 percent level.

22

Table 2(b) Market Power Statistics: Motorola, National and Texas Instruments1 Micron Texas Instruments No. of Breaks 2 2 Initial Market Power 0.646** 0.308** (0.047) (0.027) Break Date 1 Market Power After Difference

1997Q1 (1996Q4-1997Q2) 0.612** (0.081) -0.034 (0.080)

Break Date 2

1995Q3 1995Q1-1995Q4 0.622** (0.052) 0.314** (0.035)

2000Q3 2001Q2 (2000Q2-2000Q4) 2001Q1-2002Q2 Market Power After 0.716** 0.204** (0.109) (0.046) Difference 0.104 -0.418** (0.169) (0.070) 1 Standard errors in parentheses. ** indicate those market power statistics significant at the 95 percent level.

23

Table 3 Regression Results1

Constant MS KSR NSR BREAK1 BREAK1*MKTSH BREAK2 BREAK2*MKTSH BREAK3 BREAK3*MKTSH

ADM 0.215 (0.151) 4.614 (4.491) -0.163** (0.020) 0.001 (0.079) 0.170 (0.200) 0.136 (7.893) -0.747** (0.331) 20.383 (14.284)

Intel 0.535** (0.074) -0.252 (0.807) -0.231** (0.014) 0.001 (0.016) -0.204** (0.079) 3.671** (0.913) 0.398** (0.070) -3.245** (0.561) -0.311 (0.123) 1.523** (0.753)

National Micron Semiconductor 0.642** 0.321** (0.079) (0.085) 0.401 -0.516 (4.739) (0.585) -0.161** -0.221** (0.009) (0.043) -0.271** -0.027 (0.083) (0.043) -0.431 0.042 (0.407) (0.641) 16.075 5.307 (15.344) (32.554) 0.287 0.176 (0.426) (0.650) -9.408 -14.535 (15.705) (34.302)

Texas Instruments 0.341** (0.037) -0.118* (0.069) -0.246** (0.024) 0.011 (0.082) 0.781** (0.139) -8.392** (1.935) -0.874** (0.195) 13.784** (3.196)

Adj. R2 0.451 0.799 0.814 0.291 0.514 1 Standard Errors in parentheses. ** indicate those coefficients significant at the 95 percent level.

24

Table 4 Timeline of Protection: U.S. Ball Bearing Industry (1986-1987) Date Action August 1986  Timkin Co. files antidumping petition against tapered roller bearings produced by Hungary, Italy, Japan, China, Romania and Yugoslavia February 1987  Preliminary antidumping duties imposed against all targeted countries June 1987  Final antidumping duties imposed against China, Romania and Hungary August 1987  Final antidumping duties imposed against Italy and Yugoslavia October 1987  Final antidumping duties imposed against Japan

25

Table 5 Market Power Statistics: Ball Bearings1 Ingersoll-Rand (Torrington) Timken No. of Breaks 1 2 Initial Market Power 0.218** 0.669** (0.045) (0.073) Break Date 1 Market Power After Difference

2006Q3 (2000Q2-2006Q4) 0.128** (0.038) -0.091** (0.024)

Break Date 2

1986Q3 (1984Q1-1986Q4) 0.498** (0.075) -0.171** (0.044)

1993Q3 (1992Q2-1994Q4) Market Power After 0.554** (0.072) Difference 0.056 (0.063) 1 Standard errors in parentheses. ** indicate those market power statistics significant at the 95 percent level.

26

Constant MS KSR NSR ISR BREAK1 BREAK1*MKTSH

Table 6 Regression Results1 Ingersoll-Rand (Torrington) 0.186** (0.044) 0.061 (0.077) -0.143** (0.055) -0.010 (0.013) -1.739 (1.426) 0.497** (0.120) -1.213** (0.243)

BREAK2 BREAK2*MKTSH

Timkin 0.693** (0.077) -0.085 (0.083) -0.225** (0.017) 0.038 (0.092) -1.586 (1.561) -0.275** (0.049) 0.523** (0.171) 0.024 (0.061) -0.807** (0.244)

Adj. R2 0.214 0.617 Standard Errors in parentheses. ** indicate those coefficients significant at the 95 percent level. 1

27