Current Issues in Merger Enforcement:
Thoughts on Theory, Litigation Practice, and Retrospectives
THOMAS O. BARNETT
June 26, 2008
Good afternoon and thank you for attending this lecture in honor of Lewis Bernstein.(1) I appreciate Bob Kramer's remarks on the career and legacy of Lewis Bernstein. The purpose of this lecture series is not only to honor Lew, but also to honor the extraordinary men and women who have made the Antitrust Division such a special place. Lew is a stellar example of a deeply rooted tradition of excellence that both pre-dates and post-dates his career. While I can attest to recent experience, I also want to share with you some comments about Division staff made by one of my predecessors, Thurman Arnold, in 1940:
I cannot say too much about the quality of their services. It has required intelligence and energy of a high order. It has also required the sacrifice of personal interest by men [and women] who believe in the job they were doing. [They] have frequently turned down higher salaries because they were in the midst of cases for the Division which they were too loyal to abandon. . . [They] have gone without vacations. They have worked long hours. They have treated cases which they were working on as more important than any personal interest. They have been real soldiers in a cause in which they believe.(2)
I am happy to report that this tradition remains alive and well within the Division.
Today, I will address current issues in merger review and enforcement, and how the antitrust agencies are adapting both in substantive analysis and in procedural approaches. In particular, I will spend my time discussing the following challenges that we currently face: (I) substantive merger issues involving unilateral effects, coordinated effects, and differentiated products; (ii) litigation issues; (iii) retrospective studies; and (iv) the efficiency of the review process.
I. Substantive Issues in Merger Review
The agencies formally introduced the specific terminology of unilateral effects analysis in the 1992 revision to the Horizontal Merger Guidelines.(3) Since that time, my perception is that agency use of these theories has changed significantly. While I have not attempted to go all the way back to 1992, the last seven years are illustrative. The Division filed 58 merger complaints during the period from fiscal year 2001 to the present. Forty-one included only a unilateral effects claim, six included only a coordinated effects claim, and the remaining eleven contained both. There seems little doubt that recent merger challenges have focused more extensively on unilateral effects claims than was the case prior to 1992. It is worth examining the reasons for this apparent shift. I perceive at least two key factors.
First, the economy has evolved in a direction that makes unilateral effects more likely to be a relevant concern. Our world has become increasingly complex, with increasingly sophisticated and differentiated products and services to match. We no longer live in a world where products in a given category are virtually identical in terms of functionality, with the principal differences being price and perhaps reliability. Instead, customer relationship management, technological innovations, and other developments have led to customized, massed produced products, which I know sounds like an oxymoron. But consider the dizzying variety of choices now available for mobile phones (e.g., voice, data, camera, internet, GPS), cola soft drinks (regular, diet, caffeine free, or diet and caffeine free), or even different types of corn flake breakfast cereals.
These efforts to tailor products to consumer preferences are obviously pro-consumer, but they affect our merger analysis. We now often consider potential competitive effects within a category of products. For example, we consider the potential harm from a reduction in competition among sellers of certain types of bread in addition to all sellers of all bread.(4) Unilateral effects analysis is readily directed at just such an exercise.(5) Thus, the trend likely reflects in part a matching of the analysis to the most relevant theory of potential harm.
Second, the economic tools that we have been developing in recent years lend themselves more readily to unilateral effects analysis than to coordinated effects analysis. We can produce diversion ratios, critical loss estimates, and merger simulations with mathematical results to support a unilateral effects claim. In contrast, the Merger Guidelines provide a list of factors to consider in a coordinated effects analysis, but little that lends itself to mathematical estimation. Thus, the analytical tools that we have developed more readily enable agencies to support unilateral claims with the rigor that courts expect today.
As a result, there is a risk that we will not pay as much attention to potential coordinated effects as we should. We are by no means ignoring coordinated effects and continue to investigate such concerns on a regular basis. My point is to remind us all that the agencies and courts need to remain vigilant to the possibility of coordinated effects, particularly where only a few competitors will remain in a market. Further, we need to strive to improve our ability to assess and prove the possibility of coordination.
The increase in product differentiation has another effect on antitrust analysis: it increases the difficulty of product market definition. There are at least three reasons why this is so.
First, because products are often differentiated in multiple dimensions, the agencies may find it hard to provide a clear and succinct verbal or empirical description of the characteristics of the market. It is often difficult to articulate the clean break between the products/services that are "in" and "out" of the market, for which the courts tend to look. This is a significant challenge. For example, there is no doubt that the Division's reliance on a large number of product features to define the "high function" software at issue in Oracle/Peoplesoft was a complicating factor in its efforts to persuade the court that Oracle, PeopleSoft and SAP competed in a separate relevant market for such software products.
Second, application of the Merger Guidelines' market definition methodology in industries with differentiated products can produce relatively narrow markets that may run counter to the initial intuition or common sense of a judge. These markets may be correct as a matter of law and economics, as the Areeda Antitrust Law treatise describes:
To the extent that . . . a merger enables the post-merger firm profitably to assess a significant price increase without losing sales to other firms, we would say that the merger facilitates the emergence of a new grouping of sales, or relevant market, in which the merging firms have either a monopoly or else a dominant share.(6)
Nonetheless, common sense matters a lot to judges; therefore, where common sense and antitrust markets initially appear to diverge, antitrust enforcers face a need to produce evidence to persuade a court that the initial reaction is not correct.
Third, the sale of differentiated products may involve price discrimination, which can complicate traditional market definition by (1) making it harder to distinguish between products that are in and out of the market, because the distinction only applies to a subset of the parties' customers, and (2) implicating in more extreme cases the rarely-discussed substantiality issue, i.e., how many or what percentage of customers must a merger harm to constitute a Section 7 violation. For example, Oracle could be viewed as a price discrimination case. Oracle and PeopleSoft sold largely the same product to thousands of customers, most of whom did not demand all of the functionality available in the software. For those that did not want the most sophisticated functionality, the government agreed that the relevant market included a number of competitors in addition to Oracle, PeopleSoft and SAP. The government argued that Oracle charged a price that depended in part on how much of the functionality that each customer needed. This complex competitive process was not easy to investigate or convey to the court.
There is an additional implication to this growth of differentiated products that warrants mention. Not all "hypothetical monopolists" are alike in that they do not all have the same degree of market power. The SSNIP test calls for defining a separate relevant market based on the hypothetical monopolist's ability profitably to raise prices by at least five to ten percent. All else equal, relevant markets defined within a universe of differentiated products are likely to reflect a degree of market power closer to the minimum threshold than markets in which there are no differentiated products in the neighborhood. This fact affects the expected costs of error in determining whether a particular merger violates Section 7. It suggests that the agencies should be most concerned when a particular transaction is likely to reduce competition in a relevant market in which the hypothetical monopolist would have a relatively large amount of market power and where barriers to entry appear to be particularly high.(7)
On a final point in this regard, some have suggested that one solution to the challenges presented by differentiated products is to abandon market definition and proceed directly to a competitive effects analysis. I want to sound a note of caution about such an approach. As an initial matter, most judges are likely to expect the agencies to present and support a relevant market definition, and a failure to meet that expectation could cause the agency to lose credibility with the court. Further, the market definition exercise places a practical discipline on the analysis. As valuable as economic analyses can be, they are most valuable and most reliable when put in the context of other evidence and a sound theoretical framework. Thus, for example, merely running an ordinary least squares regression between several sets of randomly selected variables (such as sunspots) can yield correlations that are spurious or otherwise meaningless. When the regressions are run to test a clearly-posed hypothesis that has some common sense basis for being posited, resulting correlations are more likely to reflect an actual causal relationship. Thus, agencies should view the market definition exercise as an opportunity to explain to the court how and why the posited set of products compete in a particular manner before proceeding to show how this competition is harmed by the transaction.
II. Litigation: How We Do It and How We're Doing
I turn now to the topic of merger enforcement actions. The Division has been extremely successful in obtaining remedies for transactions that threaten to harm the competitive process. The annual total of merger enforcement matters is set forth in Figure 1, below:
Since FY2001, the Division has identified problems with 112 transactions. One is in active litigation today. For the remaining 111, the Division has obtained appropriate relief in 109 of those matters.(8) This amounts to a 98% success rate since FY2001, and since 2004, our win record is--so far--unblemished; we have obtained relief in 100 percent of the transactions in which a problem was identified during that period.
What is more, in the vast majority of cases we have achieved this record without having to undergo the delay, expense, and uncertainty of contested litigation. When we determine that a particular transaction threatens harm to competition, we identify the relief necessary to eliminate that harm. We are certainly willing to litigate to a contested judgment to obtain that required relief, and we sometimes have to do so. Thus, for example, as I just mentioned, we currently are litigating the merger of two newspapers in Charleston, West Virginia.(9) We prefer, however, the more efficient and more certain process of consensual resolution when we are able to obtain the required relief.
We have been successful in obtaining remedies without being forced to litigate to a contested decision for a number of reasons.
First, the Division has made great strides in improving the transparency of its merger decisions. In addition to the Horizontal Merger Guidelines, the Division has taken the following steps:
Such transparency enhances the ability of businesses to predict our enforcement response on both liability and remedy, leading them to propose fewer problematic mergers or to more quickly propose remedies that we will find acceptable.
Second, the importance of case selection and preparation cannot be over-emphasized. The Division is making sound decisions based on the evidence obtained in thorough investigations and combined with rigorous economic analysis. If we find that a transaction likely would harm competition but that the harm can be prevented by a merger remedy, we propose exactly that remedy. We don't propose any less--we don't apply a litigation-success discount as you might in a private damages case. We don't propose more than is needed--we don't attempt to create bargaining chips. We simply propose the remedy we believe is appropriate. At that point, the decision shifts to the parties. If the parties are not willing to consent to such relief, we litigate or they abandon the transaction.
While parties may consent to a remedy for a range of reasons, one important reason is their assessment of the strength of our case.(10) In many instances, we have obtained extensive divestitures, as we did in the recent Monsanto/Delta and Pine Land and GPC/Altivity transactions. At times, parties have consented to a remedy only after being informed that our investigation was completed and that we were proceeding to file a complaint. These circumstances suggest that parties are responding at least in part to their litigation prospects.
Third, the Merger Review Process Initiative adopted in 2001 and updated in 2006 has improved communication throughout the investigative process. Better communications means that both the Division and the parties are better focused on the key issues and more likely to be able to reach a common view of the merits.
III. Merger Retrospectives
I turn now to the topic of merger retrospectives. At first blush, the case for such studies is compelling. Given the time and resources devoted to merger enforcement and the overall impact on our economy, we should study our decisions as part of our continuing efforts to improve. Because most merger decisions are prospective, they seem particularly well suited to retrospective study to inform our future decisions. As I will explain more fully below, however, conducting and drawing lessons from even the most well-done retrospective is deceptively difficult.
A key benefit to retrospective studies is that they provide facts rather than uninformed opinions. One sometimes sees statements from an outside expert offering an opinion on the likely competitive effects of a particular transaction without access to the evidence from the agency's investigation. Because they are not based on the evidence upon which the agencies and courts rely to make merger enforcement decisions, such statements are, at best, meaningless.(11) Similarly, attempts to assess merger enforcement based on the total number of cases filed or upon the number of contested cases are fundamentally and irretrievably flawed.
Any useful assessment of merger decisions must begin with a careful analysis of facts in specific cases. As the future President John Adams wrote in the eighteenth century, "Facts are stubborn things; and whatever may be our wishes, our inclinations, or the dictates of our passions, they cannot alter the state of facts and evidence. . . ."(12) Adams likely would have made a good antitrust lawyer.
In September 2005, Whirlpool Corporation announced plans to acquire Maytag Corporation for approximately $1.7 billion. After a thorough investigation that focused on the market for residential clothes washers and dryers, the Division closed its investigation in March 2006 without taking any action.(13) The Division issued a closing statement explaining that the evidence indicated that, among other things, the competitors with excess capacity that would remain in the laundry market and the efficiencies substantiated by the parties were likely to be sufficient to prevent anticompetitive effects. The transaction was consummated immediately thereafter.
While I have seen this decision much discussed, I have not seen those discussions illuminated by many, if any, facts. In the spirit of President Adams, I asked the Division's Economic Analysis Group (EAG) to attempt to collect from public sources information that might shed some light about the effects of the Whirlpool/Maytag combination on sales of residential laundry machines. EAG obtained some illuminating facts.
First, there are several reasons to have expected the price of residential laundry machines to have increased in the United States during the last couple of years. Perhaps most importantly, the cost of materials--steel, energy, and others--used in manufacturing laundry machines rose during that time. For example, the Bureau of Labor Statistics (BLS) price index for the type of steel used in residential laundry machines--one of the largest cost components-- increased by 26 percent during the period. More generally, a price index calculated from BLS data for laundry machine material inputs, weighted for their relative use in household laundry equipment manufacturing, is set forth in Figure 2 below.(14) Further, there has been a shift--due at least in part to legal mandates--toward machines with higher energy efficiency ratings,(15) which are generally more expensive to make. Finally, the declining value of the dollar tends to make imports more expensive.
Second, we obtained the BLS price indices for household laundry equipment manufacturing, which estimate the change in wholesale prices for laundry machines sold in the U.S. over time. Notwithstanding the upward pressure on prices from the trends described above, the BLS statistics indicate that wholesale laundry machine prices have dropped by 1.7 percent in the two years since the transaction closed in 2006.(16) As shown in Figure 2 below, the result is particularly striking when compared to the apparent significant increase in costs during the same time period.
Third, we looked at residential washer/dryer prices and features from the Internet sites of Lowe's, Home Depot, Best Buy, and Sears at a number of points in time.(17) This data provides some corroboration of the BLS data. The average quality of available washing machines generally increased; an example of this trend is the increased efficiency shown in Figure 3 below. Comparing the pre-merger and post-merger periods, there was no apparent increase in price for washers of comparable quality.
Samples: (1) October 2005 (premerger); (2) March 2007; (3) December 2007; (4) June 2008
These results, while not dispositive, are consistent with the predictions made based on our investigation that sufficient competition would remain and that the merger would enable significant efficiencies, which could offset other cost increases, such as the rise in the price of steel.
This exercise also illustrates some of the difficulties in conducting a merger retrospective. As an initial matter, there is the difficulty of gathering the information necessary to conduct an ex post evaluation of the impact of a transaction. There are legal obstacles (i.e., the Division lacks authority to compel production of information for such a study), there are burden concerns for both the respondents and the agency, and there may be multiple forces at work in a market that render it difficult to discern the separate impact of the transaction.
The facts discussed above are consistent with the predictions the Division made in Whirlpool/Maytag and suggest that consumers may have benefitted as a result of the transaction. Assume for the moment that we knew with complete certainty that the Whirlpool/Maytag transaction benefitted laundry consumers. What lessons could we draw from this fact? You may be surprised to hear that my answer is "not as much as you might think."
We could learn potentially valuable lessons about whether specific predictions turned out to be correct, such as whether the parties' claimed efficiencies materialized. The reasons why (or why not) claimed efficiencies were achieved might be relevant to future cases.
We cannot, however, assess our overall merger enforcement policy based on the outcome of a single transaction. Appropriate calibration of merger enforcement policy involves (in addition to trying to improve our analytical tools to reduce the risk of error) striking the appropriate balance between Type I and Type II errors. Even in the most perfectly calibrated system, any given decision might turn out to have been erroneous in retrospect notwithstanding that it was correct based on the information available at the time. Former Deputy Assistant Attorney General for Economics Dennis Carlton analyzed this problem in a recent EAG discussion paper, which I highly recommend.(18) As he explains,"[a] retrospective study of an individual merger tells the analyst nothing about whether there is a systematic bias in antitrust policy."(19)
In order to draw broader conclusions about merger enforcement policy, one would need a larger sample, but selecting the sample is deceptively difficult. Professor Carlton examines this issue as well. As he explains, even if one were to correctly assess the outcome in a retrospective study of every unchallenged merger, the results would be biased.(20) My point is not to suggest that we should decline to engage in retrospective study. To the contrary, I think such efforts are important and valuable. Rather, my point is that we need to be careful how we conduct such studies and how we interpreting any results.
III. Update on Merger Process Reforms
I want to switch gears now and talk about the administrative side of merger review, including electronic production issues and our merger process reforms. I'll start with the challenges we face with the revolution in electronic data-keeping and its cousin, electronic production and discovery. The information technology revolution not only has made sharing information quicker, but also has vastly increased the amount of information that business entities can produce, analyze and store. By and large this is a good thing: more information, shared better and analyzed by ever more powerful tools, plays a major role in the increased efficiency and productivity of the modern economy. More information and faster analytical tools are also good things for the Division staff conducting merger reviews. But storing and sifting the information is a major challenge for everyone involved in the process.
In FY1998, the Antitrust Division had just enough electronic storage capacity to support typical second requests, which in total brought in about 0.5 terabytes(21) of electronic production. Five years later in 2003, the need for electronic storage capacity had grown exponentially to 12 terabytes. Currently in FY2008, we have increased electronic data storage capacity to support 70 terabytes of information. The related expenditures made from FY2005 through FY2007 totaled over $2.1 million. The Division anticipates that its electronic storage capacity requirements will grow to 180 terabytes by FY2013--a 36,000% increase in electronic data storage capacity in just 10 years. We obviously devote large resources to stay on top of the issue. And, of course, we are well aware that firms and their counsel are incurring costs to generate and store the information before it ever gets to us.
So how can we reduce these burdens? I will mention three ways.
First, we work internally to improve our understanding of what information is most useful to the majority of mergers, and therefore what other information we generally can forego. Some of this is informal: our attorneys and economists develop tremendous industry-specific expertise, and part of that expertise comes in the form of constantly reevaluating what information is grain and what is chaff.
Second, we work closely with the parties to make requests more targeted and make our searches smarter. Our second requests require parties to consult with our electronic data experts before producing documents in electronic form, and often we also have parties consult with our attorneys and economists who need the information. We seek to minimize the burdens on the parties and on the Division by narrowing requests, allowing parties to employ keyword-targeted searches where appropriate, ensuring that file formats are compatible, and conducting discovery in stages so that we minimize the production of information that we may not need. This is an iterative process. We constantly seek to refine it, and we make a serious goal of working cooperatively. I emphasize, though, that minimizing burdens is a two-way street and must start with the parties because the parties know and control the information in the first instance.
Which brings me to my third point, which is that parties sometimes contribute to the burden by electing to produce more documents than called for in the second request--sometime constituting a very significant portion of the production. Parties may elect not to take the time to cull out non-responsive documents in order to expedite the review process. This is a particular issue in the area of attorney-client and work product privileges, where the problem occurs in both directions: overly broad privilege claims, where we see privilege logs that list even newspaper articles and letters written by our own investigative staff; and under-inclusive privilege claims, where we see, for example, large numbers of emails between counsel and client, and where the attorneys' names are clearly listed in the "to" and "from" fields and would have been caught by even the most basic search.
While parties may seek by such an approach to expedite the investigation, such efforts can be counterproductive in that they distract everyone from the substantive analysis of the transaction. The goal of both the Division and the parties should be to identify the relevant information as quickly as possible so that we can engage in a meaningful discussion on the merits.
On a related topic, I provide a brief update about merger process reforms. In 2001 the Division released its Merger Review Process Initiative and in 2006 it announced the first major revision to that initiative.(22) One of the most significant revisions is the new "Process & Timing Agreement" merger review option, under which parties may be able to limit document searches required by a Division second request to certain central files and a targeted list of 30 employees whose files must be searched for responsive documents. This option will be made available to companies that provide certain critical information to the Division early in the investigation, agree to an investigation schedule, and agree to a sufficient period for the Division to conduct post-complaint discovery should the investigation become one of the few that result in contested litigation. To date, only one company has taken advantage of this agreement. This could mean that, despite the burdens of second request productions, parties believe those burdens are worth bearing rather than to agree to a post-complaint discovery process. It also could mean that parties believe that they can obtain much of the reduced burden through our general investigative process, which is something that we strive to achieve in any event.
We continue to streamline the merger review process. We are making better use of the initial 30 day post-filing period before the deadline for a second request. But there are limits. As the agencies increasingly deal with compressed trial schedules and difficult proof problems such as those I have mentioned today, there is a tension between minimizing the burdens on parties and ensuring that our staff will be prepared in the event of contested litigation. I can assure you that the Division nevertheless remains committed to improving merger review efficiency.
As a final consideration, I step back to look at the bigger picture of merger review and enforcement. The Division has been able over many years, and in coordination with the FTC, to make extraordinary improvements in the efficiency and effectiveness of our merger enforcement system. Comparing our current system to the past, the progress that we have made has been remarkable on multiple fronts:
Our system is not perfect, and we continuously strive to improve it. And given the longstanding tradition of excellence at the Division, I am confident that we will continue to make great progress. Thank you.
1. I thank Hill Wellford for his help in preparing these remarks, as well as numerous others within the Division who provided useful comments. I remain solely responsible for the content of these remarks.
2. Thurman W. Arnold, The Bottlenecks of Business ixx (1940).
3. The concepts underlying "unilateral effects" analysis predate 1992 by some years, reflecting the fact that the expanded focus on unilateral effects has been a long-term evolutionary process.
4. United States v. Interstate Bakeries Corp., No. 1:95-cv-04194 (Jan. 9, 1996), http://www.usdoj.gov/atr/cases/f0300/0305.pdf.
5. Unilateral effects analysis can also apply, of course, to cases that do not involve product differentiation.
6. 4 Phillip E. Areeda, Herbert Hovenkamp & John L. Solow, Antitrust law,
¶ 914f, at 77 (2d ed. 2006). While the authors refer to the merger creating a new, more narrow relevant market, the point applies more generally to market definition.
7. See generally Ken Heyer, A World of Uncertainty: Economics and the Globalization of Antitrust, 72 Antitrust L.J. 375 (2005).
8. The data includes transactions in which the Division filed a complaint and sought relief (whether or not the transaction was ultimately consummated), and in which the parties restructured or abandoned the transaction for antitrust reasons during the Division's investigation but before a complaint was filed. The relief in ten of these matters is contained in consent decrees currently pending court approval in Tunney Act proceedings. In two transactions, the Division was unsuccessful.
9. United States v. Daily Gazette Co., No. 2:07-0329 (S.D. W. Va. filed May 22, 2007). The Antitrust Division's case filings in are available on its website. Antitrust Division, Antitrust Case Filings, http://www.usdoj.gov/atr/cases/daily.htm (last visited June 26, 2008). The Division recently prevailed against the defendants' motion to dismiss. Memorandum Opinion & Order, Daily Gazette, No. 2:07-0329 (June 19, 2008).
10. Without attempting to be comprehensive, other factors certainly include the time, expense, and uncertainty of litigation.
11. I sometimes see supposed antitrust experts opining on particular merger enforcement decisions based principally or exclusively on market shares (or perceived market shares, given that they do not have access to the investigative record). While I welcome debate on issues, there is a potentially pernicious aspect to the former kind of comment. To the extent that they foster a perception that we should return to the era of United States v. Von's Grocery Co., 384 U.S. 270 (1966), and that merger decisions should be made solely or principally on an assessment of market shares, such a regression would end up harming consumers.
12. John Adams, Argument for the Defense, in 3 Legal Papers of John Adams 242, 269 (L. Kinvin Wroth & Hiller B. Zobel eds., 1965).
13. Press Release, U.S. Dep't of Justice, Department of Justice Antitrust Division Statement on the Closing of Its Investigation of Whirlpool's Acquisition of Maytag (Mar. 29, 2006), available at http://www.usdoj.gov/atr/public/press_releases/2006/215326.pdf.
14. This result is suggestive and not definitive as there are reasons why the changes in costs to particular manufacturers might not be the same as the index. For example, a particular manufacturer might substitute away from more expensive products and/or might have more favorable contractual arrangements (e.g., long-term contracts with lower prices) than other purchasers of the inputs. It is nonetheless highly plausible that the input costs of manufacturing laundry machines have risen in the last several years.
15. The minimum efficiency required for residential washing machines increased on January 1, 2007. ENERGY STAR, Clothes Washers Key Product Criteria, http://www.energystar.gov/index.cfm?c=clotheswash.pr_crit_clothes_washers (last visited June 26, 2008).
16. EAG obtained wholesale prices from the BLS Producer Price Index series "PCU335224335224P," which indexes prices received by domestic establishments manufacturing "household laundry equipment," available at http://data.bls.gov/cgi-bin/srgate. EAG calculated the materials cost index for household laundry equipment manufacturing from BLS and Census data, including Table 7 of the 2002 Economic Census Industry Report for Household Laundry Equipment Manufacturing (Census Report EC02-31I-335224) and BLS PPI indices for input materials, normalized by date and weighted according to the use of materials in the relevant manufacturing.
17. These retailers account for the majority of household laundry machines sold in the United States. This data is of prices listed for models available online and does not reflect quantities sold.
18. Dennis W. Carlton, The Need to Measure the Effect of Merger Policy and How to Do It (U.S. Dep't of Justice, Paper No. EAG 07-15, 2007), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1075707.
19. Id. at 3 (emphasis added).
20. Id. at 4. It is, of course, difficult to study what would have happened where the government successfully challenges a merger.
21. A terabyte is equal to one trillion bytes of information.
22. These and other merger policy documents can be found on the Division's website at Antitrust Division, Merger Enforcement, http://www.usdoj.gov/atr/public/premerger.htm.