Art of the deal: the merger settlement process at the Federal Trade Commission.

Kleit, Andrew N.

1. Introduction

Two firms, competing with each other in at least one line of business, face a complicated regulatory regime if they attempt to merge. Under the Hart-Scott-Rodino (HSR) Act of 1975, almost all large transactions in the United States are subject to review by either the Department of Justice or the Federal Trade Commission (FTC). In the event an enforcement agency objects to the transaction, several outcomes are possible: (i) The parties could give up and abandon their transaction, (ii) take the case to a federal court and litigate the dispute to a conclusion, or (iii) the firms could settle their differences with the enforcement agency. All settlements, however, are not the same. In particular, agencies have shown some willingness to accept "compromise" settlements, defined here as consents that do not fully resolve the relevant competitive concerns. The purpose of this study is to model the interaction between an antitrust agency and firms seeking to consummate mergers. In doing so we will attempt to answer an important question in antitrust regulation: What attributes drive the final outcome of a merger challenge?

This paper will model interactions between the FTC and private parties interested in consummating horizontal mergers that may adversely affect competition in at least one relevant market. This decision to fight, fold, or settle is somewhat more complicated than an action for damages, because the prospective nature of the alleged competitive injury gives the defendant an opportunity to abandon the transaction before the injury occurs. Moreover, the conglomerate nature of most mergers creates a possibility for a "strong" settlement to resolve competitive concerns while allowing the firm to quickly consummate the innocuous portions of the transaction. Compromise settlements also are possible, under which the consent offers partial relief. The maximum-likelihood estimation procedure used in the paper is derived directly from a game-theoretic analysis that models the outcome of the interaction between merging firms and the FTC. This allows for formal estimation of "utility functions" for both merging parties and the FTC. Estimating the structural equations of the model offers more detailed insight into the regulatory system.

We have two basic hypotheses with respect to this process. First, the underlying opportunity costs, the legal merits of particular cases, and possibly the political ramifications of enforcement decisions drive FTC decisions. Second, firms' decisions depend not only on the competitive merits of the FTC's case, but perhaps more importantly on both the financial issues relevant to the specific transaction and how the nature of the case fits into the merger review process.

Section 2 presents the background for the analysis by discussing the institutional structure of the FTC, explaining the merger review process, describing the hostage effects that may affect a firm's response to an FTC enforcement decision, and introducing the idea of a compromise settlement. Section 3 models the game-theoretic interaction between firms and the FTC and then explores an econometrically tractable method of estimating the underlying utility functions for the players in the game. Section 4 describes the data and specifications to be used. Results of the maximum likelihood estimation are presented in section 5. Concluding comments are in section 6.

We hope to shed light on an important aspect of the U.S. merger review process. We also suggest that our results may have broader consequences. Since the mid-1980s, a large number of countries have established formal antitrust procedures. The results here may be of assistance to those policy regimes as well.

2. Issues in Merger Enforcement

Background on the FTC

The FTC is a government agency charged, along with the Department of Justice (DOJ), with enforcing U.S. antitrust laws. The bulk of the casework involves the evaluation of proposed horizontal mergers. To interdict a proposed merger, the FTC (or the DOJ) must obtain an injunction from a federal district court and, if required, defend the injunction in the relevant court of appeals. If the court declines to issue the injunction (or the appeals court reverses the injunction), the firms are free to merge, although the FTC may undertake further administrative action against the merger.

Commission decisions are made by a majority vote of the commissioners, each of whom is appointed by the U.S. President, subject to Senate confirmation, for terms of up to seven years. The Commission is authorized to have five members, not more than three of whom may belong to the same political party. In each matter, the Commissioners usually receive separate memoranda from both the staff and senior management of both the Bureau of Competition (BC, the lawyers' bureau) and Bureau of Economics (BE, the economists' bureau) to assist in their decision-making process. (1) These memos are usually supplemented with "white papers" submitted on behalf of the merging parties and, on occasion, formal complaints advanced by "injured" third parties. (2)

What may be called the "modern" merger review enforcement period began in 1982 with the adoption of that year's Merger Guidelines. (See, for example, Viscusi, Vernon, and Harrington 1992.) (3) These guidelines, supplemented with a series of minor revisions, present the agency staff, the merging parties, and the courts with an organized construct for how to conduct a merger evaluation. Together with the HSR Act, they acted to create a systematic regulatory structure for mergers, which is the focus of our study.

Overview of the Merger Review Process

A small fraction of proposed mergers, if consummated, may threaten competition and present a risk of antitrust injury. If the firms recognize the potential injury, they have the option to abandon the transaction. If the firms desire to proceed, the HSR Act, in most cases, requires them to notify the government of the proposed transaction and observe a 30-day waiting period. (For further details on the HSR Act, see Johnson and Parkman 1991.) If a merger raises concerns, the government can issue a request for additional information and further delay the process. At the end of the investigation, the government will either challenge the transaction or close the investigation. (See Kolasky and Love 1997 and Waller 1998 for further details.) Coate, Higgins, and McChesney (1990), Coate (1995b), and Coate (2002) have consistently found the FTC's decision to move against a merger depends on case-specific facts and political considerations. The Commission is more likely to challenge the transaction if the Herfindahl Index (4) is over the Merger Guidelines' threshold of 1800, if barriers to entry exist, or if other structural factors make the market conducive to anticompetitive behavior.

If the government decides to challenge the merger, the parties have three choices. They can "fight" by forcing the government to obtain a court order to block the transaction through litigation, "fold" and abandon the transaction, or seek to "settle" by entering into a consent agreement with the government and address the competitive concern. Not all settlements, however, completely resolve the relevant competitive concerns.

The final stage in the horizontal merger enforcement process is litigation. Studies (Coate 1995a; Kleit and Coate 1993) of federal judicial merger decisions made after the issuance of the 1982 Merger Guidelines have found that to prevail in a merger case, the government must show high concentration (as measured by the relevant Herfindahl statistic) and barriers to entry in the relevant market. Even in these cases, if the Herfindahl is not well over 1800, evidence on structural factors conducive to competition can be dispositive. Thus, it appears that proxies for the merits of a case have a significant impact on the outcome of litigation. (We note that merging firms have won roughly half of the litigated cases.)

Factors that induce firms to avoid fighting the FTC regardless of the merits of the case are an important aspect of the regulatory process. For example, litigation against the FTC may force the firm to incur substantial costs if the firm forgoes a settlement that would bring the process to a quick resolution. These costs could include the loss of skilled workers worried about their future employment, the reduction in sales as valued customers forge new business relationships, or the enhanced risk associated with the financing of the transaction. In effect, the FTC can often hold hostage the bulk of the transaction unrelated to the competitive concerns, for it can prevent integration of any of the firm's assets pending judicial resolution of a challenge to a small part of the merger. Moreover, the acquiring firm could incur a reputational cost from putting up an active defense against the regulators. Thus, the desire by firms to resolve the antitrust problem quickly can place the FTC in a strong negotiating position. Coate, Kleit, and Bustamante (1995 [CKB]) model this problem through the use of a reduced form multinomial logit model. The reduced form nature of the model, however, precluded an estimation of utility functions for the involved parties. Our structural model allows us to break out the particular effects by plaintiff and defendant.

We note that this situation would be different in a textbook example of a merger involving two firms competing in a highly concentrated market, because that scenario leaves no assets unrelated to the competitive concern to hold hostage. In the modern world of conglomerate firms, a merger almost always involves both competitive overlaps and wide areas where the firms are not related at all. In such acquisitions, firms would desire to consummate the noncontested portion of the merger quickly and deal with the contested portion later. No simple legal mechanism exists, however, to support such a partial acquisition. Instead, the HSR Act allows the FTC to hold up the entire acquisition while seeking a preliminary injunction in order to block a relatively small aspect of the deal (see Kolasky, Proger, and Englert 1985). We will test to determine how firm decisions respond to this hostage effect.

Merger Settlements as Compromises

Merger challenges are resolved through litigation, the formal withdrawal of the filing (abandonment), or a negotiated settlement agreement with the FTC. Our model considers a further complication: the compromise or weak settlement. An antitrust settlement is often thought of as a resolution of a competitive problem (Coate 1996). Assets are transferred to an independent entity to restore the competitive nature of the market. The government obtains antitrust relief, whereas the firm is able to consummate the remainder of the transaction.

We suggest that at least some merger settlements occur when the government compromises and accepts less than a strong settlement. Such weak compromise settlements do not restore the premerger level of competition in the marketplace. To differentiate "strong" from "compromise" settlements, we adopt the methodology implicit in Elzinga (1969) and Rogowsky (1986, 1987). A settlement is defined as strong unless it meets the characteristics described below for a compromise settlement. The prototypical (strong) settlement involves the divestiture of a business unit(s) that is (are) linked to the competitive concerns. The business unit can be a formal entity of the acquired firm, a group of retail stores, or a key asset (e.g., an approved New Drug Application issued by the Food and Drug Administration). Assuming both the business and purchaser are viable, competition should be restored with the structural relief. As Elzinga (1969) and Rogowsky (1986, 1987) point out, however, settlements can easily fail to fully resolve concerns in the marketplace. These studies highlight a number of merger transactions in which the official relief was either unsuccessful (failed to address the competitive concern) or deficient (did not fully address the competitive concern). We classify a resolution as a "compromise" or weak remedy is the order itself or the compliance with it is either unsuccessful or deficient.

For obvious bureaucratic reasons, settlements are not labeled as compromises when they are placed on the public record by the agency. Generally, the analyst must infer the quality of the settlement from the available evidence relating to the remedy. Based on the Elzinga/Rogowsky specification, an order is deficient if the government secures only partial relief or the divestiture arrangement includes structures that limit the independence of the new entity. A divestiture is unsuccessful if the order contains no structural relief, de minimis relief, or requires a marketing order. (5) Similarly, compliance with the order is deficient when a full divestiture to an acceptable buyer is not achieved, and it is unsuccessful if no relief is achieved through failure to perform an agreement, through sale to a nonviable firm, or through a divestiture to a close competitive rival.

In our categorization, we assume the parties to the transaction and the FTC understand the weaknesses of the settlements when they are proposed and take these issues into account in their negotiations. Thus, ex post evidence of failure does not generally represent a "surprise." This may appear to be a strong assumption until one considers the fact that substantial a priori evidence is needed to trigger the classification of a settlement as weak. Even the consent decree violation investigations address a readily identifiable aspect of the remedy that can be linked to the classification of the settlement as a compromise, as it created a loophole for the firm to exploit. In contrast, implementation problems may affect strong settlements, as a range of developments may limit the impact of the relief. The coding of these remedies would not be changed to compromise, because the failure of the remedy could not be foreseen at the time of settlement. (6)

Thus, in this work we advance a theoretically and empirically sophisticated model that will attempt to predict whether a particular settlement will address the full competitive concern (strong) or accept some loss of competitive vigor (weak).

3. Modeling the Settlement and Litigation Process

The first part of this section models the settlement/litigation game between the merging parties and the FTC. The second part outlines a tractable, although somewhat lengthy, econometric model of this game.

Settlement/Litigation Game

The enforcement game is outlined in Figure 1. In Period 0, the FTC chooses whether to close an investigation and allow the merger to be consummated, or to move to block the merger. We note that this decision to attempt to block the merger does not need to take the form of a formal vote to seek a preliminary injunction. Commission staff could notify the parties of the Commission's interest in enforcement action, or the acquiring firm could infer this interest from the actions of the regulators. Our focus here is on the cases where the FTC seeks to stop the merger from being consummated, at least in its original form.

In Period 1 the firms make their initial decision. If they choose to negotiate, the game continues to Period 2. If they choose to fight, the case goes into litigation. If they choose to fold, the merger is abandoned.

In Period 2, given that the firms chose to negotiate in Period 1, the FTC chooses to either fight or negotiate. (We do not allow the Commission to fold at this point, for reasons explained below.) If fight is chosen, the game goes to Period 3 and the firms can either fight, with litigation ensuing, or fold and abandon the litigation. If the FTC chooses to negotiate in Period 2, settlement talks ensue, according to rules discussed below. The solution concept used is trembling-hand perfection, and this is a full information game. (7)

The settlement/litigation game has the following rules: First, we do not allow the FTC to fold in our model. In our experience, the Commission almost never folds once it votes to seek a preliminary injunction. Indeed, there is only one example of this happening during our sample period, and that was after the parties had accepted a formal consent decree. (8) Perhaps folding after moving to block a merger would cost the FTC too much in reputation, or perhaps the FTC's institutional constraints make this difficult. (For a discussion of how internal procedures bind an agency's decision making, see McCubbins, Noll, and Weingast 1987.)

Second, the FTC is likely to face lower delay costs than the merging parties. Firms have ongoing operations and legal expenses that are affected by the merger decision. This is a particular problem for the acquired party, as customers and staff may depart due to the increased uncertainty. In contrast, the FTC faces minimal costs as staffing can be reduced on any pending matters and reassigned later if the case goes to litigation. In such circumstances, the Commission is likely to capture the entire relevant surplus from a bargaining situation (Rubinstein 1985).

Third, we assume the FTC prefers a strong settlement to all other outcomes. A strong settlement generates positive publicity for the Commission for stopping an apparently anticompetitive transaction. It also allows the staff to have the opportunity to write a consent, through which they can enhance their private employment prospects. From the Commission's point of view, a strong settlement is naturally preferable to a compromise settlement as more resources are divested. It is likely preferable to risky litigation, because the strong settlement resolves the competitive concerns more quickly and at a lower cost than litigation. (9) It is also likely preferable to abandonment. Often, abandonment generates little, if any, positive publicity for the Commission, as the firms' press releases may not even mention the antitrust issues. The proposed merger partners simply return to their normal business operations. This means the Commission has nothing to show publicly for its months of investigation, as there is no settlement to place on the public record (or even no litigation to track on the official Commission web site). Thus, the FTC has no affirmative enforcement action to show to its oversight and appropriations subcommittees when it is time for agency funding. In terms of Noll (1985) and Olson (1995), the Commission desires a published settlement to send an "external signal" to the Congress that the agency is "on the job."

Fourth, we assume that firms always prefer a compromise settlement to a strong settlement. As discussed below, a compromise settlement is less restrictive than a strong settlement, and therefore likely to be more profitable to the firm.

To model the enforcement game, we require utilities (or payoffs) for both the merging firms and the FTC. (Firm utilities can be thought of as expected profits.)

Firms' Utilities

Let [X.sup.i] be a vector of explanatory variables and [B.sup.i] be the coefficients on those variables. Given this, the utilities of the merging firms are modeled below:

Abandonment: [U.sup.F.sub.A] = 0.

By a normalization assumption, the utility to the firms of abandonment is zero. Intuitively, firms get nothing from the deal if the deal is not consummated. The implicit utility (or profit) of the other three outcomes are defined relative to the abandonment outcome:

Litigation, [U.sup.F.sub.T] = [X.sup.1][B.sup.1] + [e.sup.F.sub.1]; Strong Settle, [U.sup.F.sub.S] = [X.sup.2][B.sup.2] + [e.sup.F.sub.2]; Compromise Settle, [U.sup.F.sub.W] = [X.sup.2][B.sup.2] + [k.sub.1] + [e.sup.F.sub.2], [k.sub.1] > 0, which implies [U.sup.F.sub.S] < [U.sup.F.sub.W].

We assume that [e.sup.F.sub.1] and [e.sup.F.sub.2] are normally distributed with mean 0 and variance 1 and are uncorrelated with each other. (10)

There are several different possibilities for ordinal rankings of the firms' preferences. We list them below and attach the relevant labels. The first six comparisons are for the firms' preferences for abandonment and litigation, respectively, over settling. The last two are descriptions of the firms' preferences for abandonment versus litigation. The labels are

A1: [U.sup.F.sub.W] < [U.sup.F.sub.A], and L1: [U.sup.F.sub.W] < [U.sup.F.sub.L],

determining that the firms prefer abandonment (or litigation) to a compromise (weak) settlement;

A2: [U.sup.F.sub.S] < [U.sup.F.sub.A] < [U.sup.F.sub.W] and L2: [U.sup.F.sub.S] < [U.sup.F.sub.L] < [U.sup.F.sub.W],

the firms prefer a compromise settlement to abandonment (or litigation) and abandonment (or litigation) to a strong settlement; and

A3: [U.sup.F.sub.A] < [U.sup.F.sub.S] and L3: [U.sup.F.sub.L] < [U.sup.F.sub.S],

the firms prefer a strong settlement to abandonment (or litigation).

In addition, we will require notation for the relative utilities to the firms of abandonment versus litigation. If the firms prefer abandonment to litigation, we denote that as state PABAN:

PABAN, [U.sup.F.sub.L] < [U.sup.F.sub.A].

Similarly, if the firms prefer litigation to abandonment, we denote that state as PLIT:

PLIT, [U.sup.F.sub.A] < [U.sup.F.sub.L].

FTC Utilities

Given the assumption that the Commission strictly prefers strong settlements, the number of situations where the FTC's utility must be calculated is limited. The problem is further reduced because there is no reason to calculate the relative merits to the FTC of litigation versus abandonment. The reason for this is that the firms can choose either of those two options on their own without any further interaction with the Commission. This leaves two FTC preferences to be calculated: abandonment versus compromise settle, and litigation versus compromise settle. The utility function of the FTC across the three situations is denoted below:

Compromise Settle, [U.sup.C.sub.W] = 0.

For the FTC, the utility of compromise settlement is arbitrarily set to zero, defining it as the one to which other scenarios will be compared. This implies that the abandonment and litigation utilities are defined with respect to the utility of a compromise settlement:

Abandonment, [U.sup.C.sub.A] = [X.sup.3][B.sup.3] + [e.sup.C.sub.1], and Litigation, [U.sup.C.sub.L] = [X.sup.4][B.sup.4] + [e.sup.C.sub.2].

We assume that [e.sup.C.sub.1] and [e.sup.C.sub.2] are normally distributed with mean 0 and variance 1, and uncorrelated with [e.sup.F.sub.1] and [e.sup.F.sub.2], as well as each other.

Given this utility structure, there are four different scenarios, labeled C1, C2, D1, and D2:

C1: [U.sup.C.sub.W] < [U.sup.C.sub.A], the FTC prefers Abandonment to Compromise Settle;

C2: [U.sup.C.sub.A] < [U.sup.C.sub.W], the FTC prefers Compromise Settle to Abandonment;

D1: [U.sup.C.sub.W] < [U.sup.C.sub.L], the FTC prefers Litigation to Compromise Settle;

D2: [U.sup.C.sub.L] < [U.sup.C.sub.W], the FTC prefers a Compromise Settle to Litigation.

We note that this model of interactions between a government agency and private litigants has different implications than a model of litigation between two private parties. Generally, private parties are modeled to value the same thing--money earned or lost from a suit. In such circumstances, litigation (as opposed to settlement) occurs generally only when firms have different expectations about their prospects in court. (See Cooter and Rubinfeld 1989 for a general discussion of standard settlement models.) In contrast with the standard literature, we have no need to assume divergent expectations to generate our results. In our model, where the players have utilities based on different attributes, we do not require different expectations to be a necessary condition for litigation.

Solving the Game

Next, we turn to the solution of the enforcement game, given particular scenarios. Recall that both the firms and the FTC have particular advantages in this game. If the firms desire, they can either fold or fight. This gives them the ability, under certain circumstances, to force the FTC into a compromise settlement as the firms' commitment to litigate or abandon the merger may compel the FTC to offer a compromise settlement. On the other hand, the Commission captures the entire surplus from a bargaining situation. Thus, some situations allow the FTC to force a complete resolution of the competitive concern.

The model is defined by the finite list of possible combinations and outcomes, described below and in Appendix A. Our presentation matches up the assumptions on the firm's abandonment utilities (the As) with the possible firm's litigation utilities (the Ls) and integrates the FTC's utilities as needed to solve the game. We present four scenarios here, with the remainder in Appendix A.

Combination Outcome

1.1. A1, L1 a. Litigation if PLIT

The firms prefer litigation to all other scenarios, and choose Fight in Period 1. Once the FTC has challenged the transaction, it cannot stop the dispute from going to court. (11)

b. Abandonment if PABAN

Similarly, the firms prefer to abandon the merger in the face of the initial FTC opposition, and that is the final outcome.

1.2. A1, L2 Abandonment

1.3. A1, L3 Abandonment

In either of these situations, the firms refuse to negotiate, and they wall away from the deal. The FTC preferences have no impact on the outcome.

Econometric Model

Here in the text we calculate probabilities for the relevant outcomes, although we can aggregate some steps for simplicity. The gist of the model is to maximize the relevant likelihood function by calculating the probability of each case being in one of the relevant outcomes. Mazzeo (2002), among others, uses a similar method when estimating discrete choices in product spaces.

Probabilities for the Firms' Position

Due to space limitations, only the first probability for the firms' position is calculated here, with two others calculated in Appendix B. The remaining probabilities (available at http://www.meteo. psu.edu/%7Ekleit/model%20equations.doc) are calculated similarly.

Calculating Pr(A1, L1, PLIT). Note PLIT is not affected by [e.sup.F.sub.2], the error term in the settlement utility equation. This implies we can write Pr(A1, L1, PLIT)=

Pr(A1)P(PLIT)Pr(L1 | A1, PLIT) = F(-[X.sup.2][B.sup.2] - [k.sub.1])F([X.sup.1][B.sup.1])Pr([U.sup.F.sub.L] > [U.sup.F.sub.W] | [U.sup.F.sub.W] < 0 < [U.sup.F.sub.L]),

where F(*) is the cumulative normal distribution function. Note that [U.sup.F.sub.W] < 0 < [U.sup.F.sub.L] implies [U.sup.F.sub.W] < [U.sup.F.sub.L]. Thus, Pr([U.sup.F.sub.L] > [U.sup.F.sub.W] | [U.sup.F.sub.W] < 0 < [U.sup.F.sub.L] = 1. This implies

Pr(A1)Pr(PLIT)Pr(L1 | A1, PLIT) = F(-[X.sup.2][B.sup.2] - [k.sub.1])F([X.sup.1][B.sup.1]).

This scenario results in litigation.

Probabilities for the FTC's Position

Probabilities for the FTC's position are straightforward to calculate.

Pr(C1) = Pr([U.sup.C.sub.W] < [U.sup.C.sub.A]) = Pr(0 < [X.sup.3][B.sup.3] + [e.sup.C.sub.1]) = F([X.sup.3][B.sup.3]); Pr(C2) = Pr([U.sup.C.sub.A] < [U.sup.C.sub.W]) = 1 - Pr(C1); Pr(D1) = Pr([U.sup.C.sub.W] < [U.sup.C.sub.L]) = Pr(0 < [X.sup.4][B.sup.4] + [e.sup.C.sub.2]) = F([X.sup.4][B.sup.4]); and Pr(D2) = Pr([U.sup.C.sub.L] < [U.sup.C.sub.W]) = 1 - Pr(D1).

Collectively, all terms derived in this section can be placed in the likelihood function above. The parameters can then be estimated through maximum likelihood. We note that the probabilities for each outcome are the sum of the probabilities for each state where the relevant outcome occurs. Thus, the model does not (directly) estimate the probability of being at a particular end node in the decision tree, although this calculation is possible.

4. Data and Specification

Data Sources

We reviewed the FTC's 172 attempts to enjoin horizontal transactions between June 1983 and June 1999. Our data set begins roughly a year after the adoption of the 1982 Merger Guidelines, which standardized merger investigations at the FTC. Three transactions were dropped from the data set because the structure of the settlement allowed the acquiring firm to benefit even though the measured overlap (12) was very large. (13) The sample was left with 22 litigations, 34 abandoned transactions, and 113 settlements.

Although the definition of litigations, abandoned filings, and settlements were obvious from the facts surrounding the specific cases, the identification of compromise settlements required a detailed review of each situation. As noted above, we classified a settlement as a compromise whenever the order itself or the compliance with the order could be seen as deficient or unsuccessful. Our sample identified five transactions that lacked clear structural relief and two that could be characterized as marketing orders. Six other cases involved partial divestitures. Moreover, the FTC has had numerous compliance problems with the consent docket. Divestitures would fail if the assets could not be sold at all (three cases), sold only at a negative price (two cases), or the divested entity proved unsuccessful in the market (five cases). Eight other consents resulted in only partial relief. The order could be undermined by a lack of cooperation from the respondent or a structural flaw with the original settlement that limited the procompetitive impact of the divested assets. Overall, a total of 31 settlements were found that could clearly be defined as compromises. (14)

The staff memos presented a wealth of background data such as the sales of the acquiring and acquired firms, the transaction price, the magnitude of the competitive overlap, measures of the potential anticompetitive effect of the acquisition, and the date of the acquisition. Table 1 breaks down the data by overlap fraction and case disposition, and gives a suggestion of the "hostage effect." For example, of the 29 cases brought by the Commission where the overlap fraction was 100%, 13 of these cases (44.8%) resulted in litigation. Of the 87 cases where the overlap fraction was less than 20%, however, no cases resulted in litigation.

We note that the cases in our data set are conditional upon the FTC actually voting out an enforcement complaint. Thus, it is possible that our data suffers from sample selection bias. However, a test using an extension of the data set described in Coate (2002) and following the procedure discussed in CKB (1995), we find no evidence of sample selection bias in a settlement equation.

Specification

The model requires specification of four utility functions: one for the firms' litigation versus abandonment preferences, the second for the firms' settlement versus abandonment preferences, the third for the FTC's litigation versus compromise settlement preferences, and the fourth for the FTC's abandonment versus compromise settlement preferences.

The Firms' Model

Modeling the firms' behavior requires considering the incentives for litigation and settling, both evaluated in comparison to the base situation of abandonment. Litigation implies the firm must incur opportunity costs to obtain the benefits of an eventual consummation of the transaction. The most obvious costs are linked to the delay in finalizing the transaction. By taking a merger challenge to trial, the firm accepts the costs associated with delaying the innocuous portions of the transaction. The log of the value of the transaction not subject to the FTC's competitive concerns (Not-at-Issue) represents a proxy for this cost. If the entire merger is subject to the competitive concern, no marginal delay costs accrue (because any portion of the transaction can only be accomplished by prevailing in the litigation), whereas if only a small portion of the transaction is linked to the concern, the bulk of the proposed merger is held hostage to the litigation. Thus, we expect the coefficient on Not-at-Issue to be negative.

Litigation may also generate opportunity costs related to the firm's reputation in dealing with antitrust regulators. The importance of this reputation al effect is likely linked to the number of expected interactions with the government. The log of the size of the acquiring firm (SALES) represents a good proxy for this effect, with large firms expected to have more regular interactions (some large acquisitions and some small acquisitions) with regulators than small firms (only an occasional small acquisition). It is difficult to predict a priori how an active antitrust defense will affect reputation. The regulators could shy away from future interactions with firms that actively defend proposed mergers or may focus more aggressively on firms known to litigate.

Domestic and foreign firms may differ on their expected returns from litigation. We hypothesize that foreign firms may be more reluctant to engage U.S. government regulators in legal conflict, as foreign managers may be culturally inclined to defer to government regulation. Alternatively, they may perceive their chances of success in U.S. courts as limited. (See Helland and Tabarrok 1999 for a similar effect.) An indicator variable (USA) accounts for acquisitions undertaken by domestic parent firms. We expect the coefficient on this term to be positive.

The benefits of litigation are linked primarily to efficiencies that the respondents can capture if the merger is consummated. Two efficiency variables serve to proxy this effect. First, the log of the overall transaction price of the proposed transaction (VALUE) acts as an index of the general efficiencies associated with the merger. This variable is interacted with the share of the transaction involved in the settlement (OVERLAP) to define a proxy (CONCERN) for the part of the general efficiencies that would be lost were the merger to be blocked. Second, merger-specific efficiencies are inferred from the information available to the Commission. Although it would be preferable to have a direct estimate of the value of the cost savings, this information is not readily available. Instead, we define an index (EFFICIENCY) that counts the number of pages in the staff memos needed by either the Bureau of Competition or Bureau of Economics to explain the respondent's efficiency claims. (15) As noted in CKB (1995), the extent of the analysis is often a direct response to efficiency claims made by the merging parties. Both variables are expected to take on positive signs, as efficiencies increase the returns to the firms from successful litigation.

The firm's expected benefits from litigation also depend on the probability of success on the merits and the gains from consummating the transaction. The higher the anticompetitive potential of a merger, the lower the probability of the firms' success in litigation. The higher the anticompetitive potential of the merger, however, the higher the likely anticompetitive gains from the merger. We use a variable, COURT%, to proxy the anticompetitive potential of a transaction. COURT% estimates the probability that a particular merger would be enjoined by a federal court using a customized version of a federal court decision model (Coate 1995a). (16) Because of the two impacts of the effects of an anticompetitive market structure, we use both the linear and quadratic form of COURT%. We expect the sign on the linear term to be positive and the sign on the quadratic term to be negative.

In addition, prior to June 1995, the FTC routinely pursued administrative complaints against merging firms that chose to fight the FTC, even after a federal court decision allowing the merger to proceed (see Coate and Kleit 1998 and Lopatka and Mongoven 1995 for a description of this policy). To proxy the additional burden on fighting the FTC under this policy regime, we define ADCOURT as an indicator variable equal to 1 prior to June 1995, and 0 thereafter. (The current policy calls for case-by-case consideration of the merits of further litigation, a policy more similar to that of the DOJ.) The FTC has closed two investigations and declined to pursue another administrative litigation under the revised policy. Given an unsuccessful preliminary injunction merger challenge, the FTC is yet to undertake a full trial on the merits. We expect this coefficient to have a negative sign, as the costs of litigation were likely higher when the FTC routinely pursued such administrative complaints.

This leaves us with

Utility Litigation versus Abandonment = [U.sup.F.sub.A] = [X.sup.1][B.sup.1] + [e.sup.F.sub.1] = g (Not-at-Issue, SALES, USA, CONCERN, EFFICIENCIES, COURT%, COURT% (2), ADCOURT). (17)

The respondents also possess a separate settlement utility function. To settle, the firm must give up the efficiencies associated with the markets of concern (along with any anticompetitive benefits). In general, transaction costs preclude renegotiation of the overall price, so the partial transaction allowed by the settlement may not be as attractive to the acquiring firm. This loss of value can be proxied by the two proxies for market-specific efficiencies (CONCERN and EFFICIENCIES) foregone by the settlement, and we expect the coefficient on these variables to be negative.

Similarly, a settlement allows the acquiring firm to immediately take control of the assets unassociated with the competitive concern (proxied by the variable Not-at-Issue). This allows the acquiring firm to immediately accomplish the general efficiencies associated with their broad strategic vision. Moreover, a settlement allows the firm to minimize the transaction uncertainties associated with delaying the consummation of the deal. We thus expect the coefficient on Not-at-Issue to be positive, as it measures the overall value that the firm is able to take away from the settlement.

In addition, we add an indicator variable RETAIL to denote when the settlement concerns retail properties. Retail units are discrete entities, and therefore may be easier to divest. On the other hand, retail properties may be subject to brand name capital. Divestiture would therefore impose greater costs on the merging firms. Moreover, economies of scope across retail properties may limit the number of potential purchasers, and thus the potential price. Therefore, we have no priors on the sign of this coefficient. Our settle-abandonment equation is therefore

Utility of (Strong) Settle versus Abandonment = [U.sup.F.sub.S] = [X.sup.2][B.sup.2] + [e.sup.F.sub.2] = g (CONCERN, EFFICIENCIES, Not-at-Issue, RETAIL).

The FTC's Utilities

The model requires two different FTC utility equations to be estimated: abandonment versus compromise settle, and litigation versus compromise settle. We will address the question of abandonment versus compromise settle initially.

The cost savings from a compromise settlement are likely linked to the efficiencies of the merger. As discussed above for firms' utility, two efficiencies exist, CONCERN and EFFICIENCY. If the FTC is interested in promoting an efficient economy, the variables should negatively impact the FTC's utility from abandonment. In contrast, if the FTC acts to reduce efficiency, as the public choice advocates might suggest, we would expect the opposite sign on these coefficients. (18)

The relative utility of abandonment versus compromise settle is likely to be a function of the anticompetitive aspects of the merger. The more anticompetitive the merger, the less attractive a compromise settlement would be to the Commission. We expect that as a merger becomes more anticompetitive, the relative utility to the FTC of abandonment with respect to a compromise settlement increases, as the chance of the Commission being embarrassed by subsequent price increases grows. Again, we use COURT% to model this, and we expect the coefficient on this variable to be positive.

The Commission's overall shadow value of settlements also could change over time as the political composition of the Commission changes. For example, each administration could have different goals for the agency, and thus different opportunity costs. We test for these effects with categorical variables for the three presidential administrations represented at the FTC--Reagan administration (Chairmen Miller and Oliver), Bush administration (Steiger), and Clinton administration (Pitofsky). Our specification will add indicator variables for Bush (COMM1) and Clinton (COMM2) chairmen, leaving the Reagan chairmen as the control variable. We have no priors on COMM1 and COMM2.

Partisan congressional oversight may also impact on FTC preferences. In addition to affecting case selection, political considerations could influence the type of outcome. Multiple theories could be advanced. For example, a heavily Democratic congress could expect the agency to maximize litigation, whereas a more Republican congress could prefer settlements. On the other hand, it could be the Democrats in Congress that prefer some type of settlement, and the Republicans that insist on litigation to preclude the agency from devolving into an unofficial regulator. To test for the various possibilities, we follow CKB (1995) and use a variable CONGRESS, measuring the average Democratic control of the Congress. No clear expectations exist for the sign on this coefficient.

The FTC may also be interested in the reaction of the general public to its actions, and thus high-profile deals in consumer goods markets would obtain the most attention. The variable VALUE, defined as the log of the value of the transaction, highlights large transactions more likely to obtain media attention. We hypothesize that the Commission may be reluctant to "take the blame" for large transactions that are not consummated, as large firms tend to have more affected stakeholders. Similarly, we expect the FTC may be reluctant to take the blame for transactions in consumer products industries that are not consummated, as represented by the indicator variable CONSUMER. We therefore expect that the coefficients on both VALUE and CONSUMER will be negative.

This leaves us with the FTC's utility of abandonment versus compromise settle as

FTC Utility of Abandonment versus Compromise Settle = [U.sup.C.sub.A] = [X.sup.3][B.sup.3] + [e.sup.C.sub.1] = g (CONCERN, EFFICIENCIES, COURT%, COMM1, COMM2, CONGRESS, VALUE, CONSUMER).

The FTC's utility of litigation versus compromise settle will also have CONCERN, EFFICIENCIES, COURT%, COMM1, COMM2, CONGRESS, and VALUE, for similar reasons, for their inclusion in the FTC's abandonment equation. In addition, opportunity costs can affect the Commission's incentive to settle with respect to litigation. Holding staff constant, an increase in workload places pressure on the FTC to settle more cases. Thus, we computed the ratio of the number of HartScott-Rodino Act filings per full-time equivalent staffer at the FTC (WORK) for each fiscal year in the sample. This variable was expected to have a positive impact on the utility of any settlement, and therefore take on a negative coefficient in our regression. This leaves us with

FTC Utility of Litigation versus Compromise Settle = [U.sup.C.sub.L] = [X.sup.4][B.sup.4] + [e.sup.C.sub.2] = g (CONCERN, EFFICIENCIES, COURT%, COMM1, COMM2, CONGRESS, VALUE, WORK).

5. Empirical Results

Table 2 presents the results of the regressions estimating the firms' utility functions. The left side of the table reports the results for firms' litigation versus abandonment utility. The coefficient on Not-at-Issue is positive, as expected, but only marginally significant. This suggests that value-based hostage effects impose costs on the firm, so transactions with large noncontested aspects are relatively less likely to end in litigation. The coefficient on SALES is significant and negative, indicating large firms perceive a negative effect to their reputation caused by fighting the government. The coefficient on USA, however, is insignificant, suggesting foreign firms do not perceive an additional disincentive to litigate. The coefficient on CONCERN is positive and significant, as is the coefficient on EFFICIENCIES. These results are consistent with the hypothesis that the opportunity to capture efficiencies is an important factor in firms' desire to litigate. In contrast, the coefficients on the two COURT% variables are individually insignificant (as well as jointly, using a likelihood ratio test), implying that market structure considerations do not have a net affect in the decision to litigate. Finally, the coefficient on ADCOURT is insignificant. Thus, the Commission's 1995 change in litigation policy did not have a net effect on the firms' litigation propensity.

The right side of Table 2 reports the estimation results for the firms' settlement versus abandonment equation. The coefficients on the two efficiency variables, CONCERN and EFFICIENCY, are both negative and significant, indicating that as the consent adversely affects the efficiency returns from the transaction, the firms' interest in a settlement decreases. On the other hand, the coefficient on Not-at-Issue is positive and significant, suggesting the larger the value of the uncontested assets retained by the acquiring firm, the more likely a consent will be chosen. The coefficient on RETAIL is negative and significant. This result suggests that divesting retail outlets is relatively more expensive than divesting plants and equipment. It may imply that the divesting firm is forced to accept "fire sale" prices as the result of either the reduction in the value of the assets as part of another retail network or the relative scarcity of bidders for the stores. (19)

Table 3 presents an analysis of the FTC's utility of abandonment versus compromise settlement choice. The results of the equation's estimation generate a vector of (arbitrarily close to) 0s and 1s for the probability that the FTC prefers abandonment to a compromise settle. In this circumstance, as Lien and Rearden (1990) and Caudill and Jackson (1989) indicate, coefficients and t-statistics are unreliable. The problem is that the estimation has reached a "comer solution," where changing the relevant coefficients does not change (to any appreciable amount) the probability of the relevant preference.

To obtain some insight into the effects of the variables, the authors cited above suggest focusing on the marginal changes generated by the relevant variables. What we have done, therefore, is to calculate the mean change in probability in the following way: If the relevant variable is an indicator variable, we calculate the relevant probabilities given that that variable is either 0 or 1. For other variables, we let the value of the variable be either the mean value minus twice the standard deviation, or the mean plus twice the standard deviation. We report the mean difference in probabilities over the entire data set. (20) To determine statistical significance, we estimate the model with and without each of the variables in question. We then calculate likelihood ratio test statistics, which equals twice the difference in the relevant log likelihoods. These statistics axe distributed chi-square with 1 d.f., have a 5% confidence threshold of 3.84, and have a 10% confidence threshold of 2.72.

Five variables are significant using this test. CONCERN and EFFICIENCIES (the latter with marginal significance) have positive signs on their coefficient, implying that the FTC acts to reduce efficiency when it has the opportunity. The coefficient on COMM1 is significant and negative, indicating the Steiger FTC was more willing to accept compromise consents. As the first Bush administration was generally considered more aggressive than the previous Reagan regulators, the compromise settlement may have been used to cushion some of the impact of additional enforcement. The coefficients on both VALUE and CONSUMER are negative, indicating the Commission is more likely to accept compromise settlements for larger transactions and transactions that deal with consumer products.

We also estimated the probability that the FTC prefers litigation to a compromise settlement. The results of the estimation produced a row of numbers arbitrarily close to 0. We deduce from this result that the FTC's preferences in this regard are deterministic: It always prefers a compromise settlement to litigation.

Table 4 compares the actual outcomes to the ones predicted by the model. Seventeen of 22 litigations (77.3%) were predicted correctly, as well as 28 of 34 abandonments (82.4%). In addition, 69 of 82 strong settlements (84.1%) were predicted correctly. In contrast, only 10 of 31 compromise settlements (32.3%) were predicted correctly. We note, however, that compromise settlement is the second highest predicted probability for all 21 remaining cases that resulted in compromise settlements. In total, 124 of 169 (73.4%) cases were predicted correctly. (It is also possible to calculate the probability of which path the negotiations took, as illustrated in Figure 1.)

By implicitly allowing the consummation of the uncontested portions of a transaction, the model is able to simulate the elimination of the overlap hostage effect. In effect, contested portions of the transaction are held separate and resolved through a game with the FTC. To compute the relevant probabilities for each outcome, the Not-at-Issue variable is set to zero in the firm utility functions (to represent the situation in which no assets are related to the competitive concern, and thus no hostage effect exists). In addition, the VALUE variable is replaced with CONCERN in the FTC utility function, as only a smaller portion of the deal is contested. Of the 113 settlements in the sample, 55 (roughly 50%) would be resolved with compromise settlements, whereas 47 (about 40%) would be abandoned and 11 (roughly 10%) would be litigated. (21) No strong settlements would be predicted, a result consistent with the simulation structure of a strong settlement effectively meaning the contested portion of the transaction would be completely blocked. These results suggest that the hostage effect clearly impacts the outcome of the enforcement process, with the FTC entering into forced settlements in some cases and avoiding litigation on the merits in others.

6. Conclusion

The methodology and data set presented here create a unique opportunity to examine the interaction between the regulators and the regulated. It allows analysis of which factors affect outcomes in the decision to litigate, fold, or settle (including insight into the type of settlement) in the context of antitrust enforcement at the FTC. In particular, it allows the testing of two important hypotheses: (i) whether FTC decisions in this settlement/litigation game are driven by the underlying opportunity costs and competitive merits of particular cases; and (ii) whether firms' decisions depend not only on the competitive merits of the FTC's case, but perhaps more importantly on both the financial nature of the relevant transaction and how the nature of the case fits into the merger review process. The combination of an analytic approach to this negotiation sequence, an econometrically tractable model, and access to important data allows for testing of these hypotheses. To the extent that the FTC settlement process is much like the process at the DOJ (something that is more relevant after the FTC's 1995 policy change on administrative trials on unsuccessful preliminary injunction cases), this research also sheds light on enforcement decisions by that agency as well.

We observe that merger-related efficiencies play a significant role in firms' decision-making process. The two efficiency variables, one for general and the other for specific efficiencies, are positively linked to the relative probability of litigation and negatively tied to the relative probability of settlement. This result tends to support the theory that mergers are driven by the opportunity to capture efficiencies. In contrast, the structural characteristics of the merger, at least as measured by the FTC's economists, do not appear to significantly impact firms' utility of litigation. This seems to imply that strict legal issues do not deter firms from testing their mergers in court. Firms, however, are apparently deterred from fighting the FTC by the potential for negative impacts on their reputations. Similarly, hostage effects associated with the size of the uncontroversial portion of the acquisition held up by the competitive concern do affect firm utilities. These results suggest that institutional structures allow the FTC to hold certain mergers "hostage" and force divestitures.

The FTC's preferences, although somewhat difficult to establish, indicate that when it has the opportunity, the Commission is more likely to prefer that efficient acquisitions be abandoned rather than move forward with a compromise settlement. This result is compatible with a capture theory of the bureaucracy and perhaps suggests that the FTC should be more critical of competitor complaints. Compromise settlements were also more likely to be preferred during the FTC's Steiger era and for transactions that were large and/or dealt with consumer products. In addition, results indicate that the FTC prefers to take a compromise settlement, rather than to face the risks of litigation.

Appendix A

In this Appendix, the scenarios of the game not discussed in section 3 are presented:

2.1. A2, L1 Litigation

indicates that the firms insist on a litigated outcome. The FTC utilities are not relevant.

2.2. A2, L2 a. Compromise Settlement if PLIT, D2,

where the firms prefer a compromise settlement to all the other outcomes, with litigation as the second choice (as the FTC would rather see a compromise settlement than litigation, that is what will occur).

b. Litigation if PLIT, D1,

where, on the other hand, the firm's second choice of litigation will occur if the Commission prefers litigation to a compromise settlement.

c. Compromise settlement if PABAN, C2,

where the firm's second choice becomes abandonment. As the Commission prefers to avoid an abandonment, it is willing to offer a compromise settlement.

d. Abandonment if PABAN, C1,

If the firm cannot obtain a compromise settlement from the FTC, the firms choose to abandon the transaction.

2.3. A2, L3 a. Compromise Settlement if C2

b. Abandonment if C1,

where the firms will not fight and will fold if the FTC insists on a strong settlement. In this instance the FTC preferences are controlling, with an offer of a compromise settlement accepted, whereas the offer of a strong settlement leads to the abandonment of the transaction.

3.1. A3, L1 Litigation,

in which, again, the parties insist on a fight, so the FTC must oblige.

3.2. A3, L2 a. Compromise Settlement if D2

b. Litigation if D1,

where the firms will offer the FTC a compromise settlement. The offer is accepted if the FTC prefers the compromise to litigation. Otherwise, the dispute is resolved in court.

3.3. A3, L3 Strong Settlement,

where the firms prefer a strong settlement to both litigation and abandonment, and the FTC is happy to take them up on it. Although the firms would prefer a compromise settlement to a strong one, the assumption that the FTC captures the entire surplus from a negotiation implies that a strong settlement will occur here.

With these results in hand, we can write down the probability for each event of abandonment, litigation, strong settlement, and compromise settlement. Let [Y.sup.A], [Y.sup.L], [Y.sup.S], and [Y.sup.W] equal indicator variables for whether abandonment, litigation, strong settlement, or compromise settlement occurred. The probability of each event is

Pr([Y.sup.A]) = Pr(A1,L1,PABAN) + Pr(A1,L2) + Pr(A1,L3) + Pr(A2,L2, PABAN) x Pr(C1) + Pr(A2,L3) x Pr(C1), Pr([Y.sup.L]) = Pr(A1,L1,PLIT) + Pr(A2,L1) + Pr(A2,L2, PLIT) x Pr(D1) + Pr(A3,L1) + Pr(A3,L2) x Pr(D1), Pr([Y.sup.S]) = Pr(A3,L3), and Pr([Y.sup.W]) = Pr(A2,L2,PLIT) x Pr(D2) + Pr(A2,L2, PABAN) x Pr(C2) + Pr(A2,L3) x Pr(C2) + Pr(A3,L2) x Pr(D2).

This implies the log likelihood function to be maximized is

ln L = [Y.sup.A] x ln[Pr([Y.sup.A])] + [Y.sup.L] x ln[Pr([Y.sup.L])] + [Y.sup.S] x ln[Pr([Y.sup.S])] + [Y.sup.W] x ln[Pr([Y.sup.W)].

Appendix B

Here two additional probabilities discussed in Section 3 are calculated. The probabilities of the remaining scenarios are available upon request from the authors.

All Other Instances of A1

All other instances of A1 generate abandonment. These occur with probability Pr(A1) x [1 - Pr(L1, PABAN | A1)] = F([-X.sup.2][B.sup.2]-[k.sup.1]) F([-X.sup.1][B.sup.1]), solving for 1.2 and 1.3.

Calculating Pr(A2, L1)

First, the probability of A2:

Pr(A2) = Pr([X.sup.2][B.sup.2] + [e.sup.2.sub.F] < 0 < [X.sup.2][B.sup.2] + [k.sub.1] + [e.sup.2.sub.F] = Pr ([-X.sup.2][B.sup.2] - [k.sub.1] < [e.sup.2.sub.F] < [-X.sup.2][B.sup.2]) = F([-X.sup.2][B.sup.2]) - F ([-X.sup.2][B.sup.2] - [k.sub.1]).

Thus, given A2, we know that [e.sup.F.sub.2] is distributed normally, truncated above and below

[e.sup.F.sub.2] ~ (f(e')/Pr(A2)[e.sup.F.sub.2][epsilon][[-X.sup.2][B.sup.2] - [k.sub.1], [-X.sup.2][B.sup.2].

L1 implies [X.sup.1][B.sup.1] + [e.sup.F.sub.1] > [X.sup.2][B.sup.2] + [k.sub.1] + [e.sup.2.sub.F]. The probability of this occurring given A2 is

Pr(L1 | A2) = Pr([X.sup.1][B.sup.1] + [e.sup.F.sub.1] > [X.sup.2][B.sup.2] + [k.sub.1] + [e.sup.F.sub.2] | A2) = Pr(e.sup.F.sub.1] < [-X.sup.2][B.sup.2] + [X.sup.1][B.sup.1] - [k.sub.1] - [e.sup.F.sub.2] | A2)

Let [e.sup.F.sub.2] = e; then

Pr(L1 | [e.sup.F.sub.2 = e) = F([-X.sup.2][B.sup.2] + [X.sup.1][B.sup.1] - [k.sub.1] - e).

Over the distribution of [e.sup.F.sub.2],

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.],

which implies

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.].

Other probabilities are available at http://www.meteo.psu.edu/%7Ekleit/model%20equations.doc.

Table 1. Merger Outcomes By Overlap Fraction

Overlap Mergers Mergers Strong Compromise
Fraction (%) Total Abandoned Litigated Settlements Settlements

 0-9.9 59 1 0 47 11
10-19.9 28 2 0 20 6
20-29.9 13 4 1 6 2
30-39.9 8 1 0 4 3
40-49.9 6 2 2 1 1
50-59.9 7 3 0 2 2
60-69.9 11 3 3 2 3
70-79.9 1 1 0 0 0
80-89.9 2 2 0 0 0
90-99.9 5 2 3 0 0
100 29 13 13 0 3
Total 169 34 22 82 31

Table 2. Firm Utility Functions

 Litigation vs. Abandonment

Variable Coefficient t-statistic

Constant -2.898 (a) -2.015
Not-at-Issue -0.190 (b) -1.861
SALES -0.478 (a) -3.306
CONCERN 0.831 (a) 3.826
EFFICIENCY 1.265 (a) 4.042
ADCOURT 0.765 1.290
COURT% -3.576 -1.130
COURT% (2) 2.165 0.875
USA 0.113 0.180

 Settlement vs. Abandonment

Variable Coefficient t-statistic

Constant -0.408 -1.079
CONCERN -0.232 (a) -2.896
Not-at-Issue 0.400 (a) 6.525
EFFICIENCY -0.327 (a) -2.810
RETAIL -0.582 (a) -2.053

In threshold
 ([k.sub.1]) 0.603 (a) 3.778

(a) Significant at the 5% level.

(b) Significant at the 10% level.

Table 3. FTC Utility of Abandonment versus Compromise Settlement


 Log Likelihood of Marginal Likelihood Ratio
Variable Restricted Model Impact Statistic

Constant 108.664 -0.131 3.228 (b)
CONCERN 110.785 0.799 7.466 (a)
EFFICIENCIES 108.674 0.781 3.244 (b)
COURT% 107.052 -0.067 0.000
COMM1 110.743 -0.201 7.382 (a)
COMM2 108.322 -0.242 2.540
CONGRESS 107.052 0.012 0.000
VALUE 109.505 -0.072 4.906 (a)
CONSUMER 109.247 -0.188 4.390 (a)

Variable

Constant Log likelihood of full
CONCERN model: -107.052
EFFICIENCIES
COURT% Mean Probability of FTC
COMM1 preferring Abandonment:
COMM2 0.403
CONGRESS
VALUE
CONSUMER

(a) Significant at the 5% level.

(b) Significant at the 10% level.

Table 4. Actual versus Predicted Outcomes

 Predicted
 Litigation Abandonment Strong Settlement
Actual
 Litigation 17 5 0
 Abandonment 3 28 1
 Strong Settlement 0 4 69
 Compromise Settlement 2 1 18
 Total 21 38 89

 Predicted
 Compromise Settlement Total

Actual
 Litigation 0 22
 Abandonment 2 34
 Strong Settlement 9 82
 Compromise Settlement 10 31
 Total 21 169

This article is based on nonpublic data obtained from Federal Trade Commission internal files. The Commission's General Counsel has authorized publication of such data in aggregated form under Commission Rule 4.11(g), 16 C.F.R. 4.11(g) (2002). The views expressed herein are solely those of the authors and do not constitute the views of the Commission or any individual Commissioner. We would like to thank Tim Brennan, Mark Williams, seminar participants at West Virginia University and the University of British Columbia, and two referees for helpful comments.

(1) The structure at the DOJ is similar, although less formal, as the enforcement decision is made by the Assistant Attorney General for Antitrust (an executive branch official subject to Senate confirmation). On occasion, a specific State Attorney General conducts a merger investigation; however, the state regulators do not benefit from the provision of the HSR act, which delays firms from consummating their transactions during federal investigations, as discussed below.

(2) Customer complaints are generally considered more credible than competitor complaints, because customers may suffer if they are unable to pass on the entire potential anticompetitive effect to final consumers, whereas competitors should benefit from noncompetitive pricing. Similarly, competitors are more likely to be harmed by merger-specific efficiencies, and thus their concerns should be discounted to some degree.

(3) The FTC also released formal guidelines in 1982 (FTC 1982), with the document noting the DOJ guidelines will be given "considerable weight by the Commission and its staff." In practice, the FTC quickly adopted the DOJ guidelines as its own. The Commission formally accepted the structure of the DOJ guidelines in 1992 (see Scheffman, Coate, and Silvia 2003; USDOJ 1992).

(4) The Herfindahl Index equals the sum of the squares of the market shares of the firms in the relevant market and is now the generally used measure of market concentration in antitrust enforcement. Implicit in the use of the Herfindahl in merger analysis is some material increase in the index caused by the merger. The guidelines specify 50 or 100 points as the threshold of concern, although most challenged transactions exhibit much higher increases.

(5) A marketing order involves the divestiture of output to another entity that then markets the product in competition with the firm under order. In contrast to a license agreement, the firm under order manufactures the divested product itself.

(6) One of the three strong settlements studied in the Commission's divestiture study was found to have failed to significantly impact the market. Although the case-specific information is confidential, the broad details of the divestiture review are covered in FTC (1999).

(7) As discussed later in footnote 11, using backward induction as a solution concept would generate the same results, but in a somewhat more complicated fashion.

(8) In Nestle Food Company (File No. 941 0124, June 7, 1995), the Commission justified its decision to reject the signed consent agreement that had already been placed on public record based on new evidence relating to the definition of the canned cat food market, the relevant concentration, and ease of entry. See http://www3.ftc.gov/opa/1995/06/nestle2.htm. Note that our assumption that the FTC does not fold relates only to the FTC's position as a prosecutor, and not to FTC actions when the FTC acts as an administrative court. For a discussion of the FTC as an administrative court, see Coate and Kleit (1998).

(9) Here some staff preferences could differ from the Commission's preferences. Although certain staff might prefer litigation to a strong settlement, other staff might prefer the experience of structuring a consent order, and still other staff might prefer the quiet life of the regulator.

(10) Throughout the model, the subscript W is used for compromise (i.e., weak) settlements to avoid the confusion with the superscript C used for utilities of the FTC.

(11) If the solution concept used were backward induction, the firms would be indifferent between Fight in Period 1 and Negotiate, Fight, Fight, or Negotiate, Negotiate, Fight. However, using trembling hand perfection, the firms Fight in Period 1 to avoid the possibility that they (the firms) "tremble" in Period 3 and choose Fold or (if applicable) a settlement outcome. For reasons of space, descriptions of the other nonrelevant actions and equilibria in the other scenarios are omitted.

(12) The overlap index was generally defined as the ratio of the sales associated with the problematic overlaps to the total sales of the acquired entity. Other share proxies involved market value, number of retail outlets, or a count of business units.

(13) in these deals, the problematic assets acquired were larger than the assets associated with the competitive concern held in the acquiring firm. (For example, a firm with a 1% market share could have acquired a stand-alone firm with a 99% market share.) In these deals, our overlap variable is a poor proxy for the significance of the problem facing the acquiring firm, because the concern can be resolved with a simple divestiture of the acquiring firm's 1% share. In contrast, if both firms had relatively equal shares or the acquiring firm had a larger share, the competitive concern could only be resolved by divesting the acquired business. However, for a high overlap variable, this would imply a full divestiture effectively rescinds the transactions.

(14) In theory, the FTC could settle a dispute by obtaining full relief in one affected market, but no relief in another adversely affected market. Based on our review of the files, this does not appear to be a significant empirical concern.

(15) We also created a variable EFFICIENCY2, which is the total number of efficiency pages divided by the total number of pages in the staff memos. The results of the models below were not qualitatively different when EFFICIENCY2 was used in place of EFFICIENCY.

(16) The model used is COURT% = F(-6.2 + 1.97 * BECOLLUDE + 4.26 * BEBARR + 0.000816 * HERF). BECOLLUDE is an indicator variable for whether the staff of BE found market elements consistent with competitive concern, BEBARR is a variable for whether BE staff found a barrier to entry in the market, HERF is the Herfindahl index of concentration found by BE staff in the market, and F() is the cumulative normal distribution function. The coefficients were estimated with data on court findings taken from a review of federal court decisions, and all the variables were significant in magnitude. Thus. the variable defines the probability a court will block the proposed transaction and hence represents a proxy for the merits of the case. Although the estimated coefficient on COURT% in the regressions above is efficient and unbiased, the standard errors are uncertain, as COURT% is an estimated regressor.

(17) Ideally, we also would like to know how the Commissioners voted on each matter to proxy the parties' likelihood of success in court. Unfortunately, some abandonments occurred before an official Commission vote, and thus the information is not always available.

(18) A public choice analysis might suggest that the FTC is captured by complaining competitors concerned with specific acquiring firms gaining competitive advantages through merger-specific efficiencies. Thus, these special interests would induce the FTC to prefer abandonment over a compromise settlement when large efficiencies are present. For a general discussion of the public choice theory of regulation, see Stigler (1971).

(19) We attempted using several variables to model [k.sub.1], the threshold constant that measures the difference in utilities between strong and compromise settlements. None were close to significant.

(20) Conceptually, one could be interested in the mean probability change over the set of data where the choice could be relevant, that is, mergers where the outcome was either an abandonment or a compromise settlement. The results are similar over this data set.

(21) Interestingly, the simulation results imply that 14 of the 31 compromise settlements would be abandoned were the hostage effect to be eliminated. This implies that the firms were not particularly interested in the assets of competitive concern once the transactions costs associated with resolving the overall dispute are eliminated by the hypothetical ability to consummate the innocuous portion of the transaction without delay.

References

Caudill, S. B., and J. D. Jackson. 1989. Measuring marginal effects in limited dependent variable models. The Statistician 38:203-6.

Coate, Malcolm B. 1995a. Merger analysis in the courts. Managerial and Decision Economics 16:581-92.

Coate, Malcolm B. 1995b. The shifting sands of merger enforcement at the Federal Trade Commission. International Journal of the Economics of Business 2:393-408.

Coate, Malcolm B. 1996. Winners and losers in the settlement of merger litigation. In The Economics of the antitrust process, edited by Malcolm B. Coate and Andrew N. Kleit. Boston, MA: Kluwer Academic Press, pp. 153-73.

Coate, Malcolm B. 2002. A test of political control of the bureaucracy: The case of mergers. Economics and Politics, 14:1-18.

Coate, Malcolm B., Richard S. Higgins, and Fred S. McChesney. 1990. Bureaucracy and politics in FTC merger challenges. Journal of Law and Economics 33:463-82.

Coate, Malcolm B., and Andrew N. Kleit. 1998. Does it matter that the prosecutor is also the judge? The administrative complaint process at the Federal Trade Commission. Managerial and Decision Economics 19:1-11.

Coate, Malcolm B., Andrew N. Kleit, and Rene Bustamante. 1995. Fight, fold, or settle? Modeling the reaction to FTC merger challenges. Economic Inquiry 33:537-51.

Cooter, Robert D., and Daniel L. Rubinfeld. 1989. Economic analysis of legal disputes and their resolution. Journal of Economic Literature 27:1067-97.

Elzinga, Kenneth G. 1969. The antimerger law: Pyrrhic victories. Journal of Law and Economics 12:43-77.

Federal Trade Commission. 1982. FTC statement on horizontal mergers. Trade and Regulation Reporter (CCH) 4 paragraph 13,200.

Federal Trade Commission. 1999. A study of the Commission's divestiture process. Available http://www.ftc.gov/os/1999/08/ divestiture.pdf. Accessed 30 January 2004.

Helland, Eric, and Alex Tabarrok. 1999. Court politics: The political economy of tort awards. Journal of Law and Economics 42:57-88.

Johnson, Ronald, and Allen Parkman. 1991. Premerger notification and the incentive to merge and litigate. Journal of Law, Economics, & Organization 7:145-62.

Kleit, Andrew N., and Malcolm B. Coate. 1993. Are judges leading economic theory? Sunk costs, the threat of entry and the competitive process. Southern Economic Journal 60:103-18.

Kolasky, W. J., Jr., and J. W. Love. 1997. The merger review process at the Federal Trade Commission: Administrative efficiency and the rule of law. Administrative Law Review 49:889-913.

Kolasky, W. J., Jr., P. A. Proger, and R. T. Englert. 1985. Anticompetitive mergers: Prevention and cure. In Antitrust and regulation, edited by F. M. Fisher. Cambridge, MA: MIT Press, pp. 49-84.

Lien, D., and D. T. Rearden. 1990. A remark on "'an advantage of the linear probability model over probit or logit." Oxford Bulletin of Economics and Statistics 52:223-5.

Lopatka, John E., and James Mongoven. 1995. After preliminary relief in merger cases is denied, what then? Research in Law and Economics 17:149-224.

Mazzeo, M. J. 2002. Product choice and oligopoly market structure. RAND Journal of Economics 31:221-42.

McCubbins, M. D., R. G. Noll, and B. R. Weingast. 1987. Administrative procedures as instruments of political control. Journal of Law, Economics, and Organization 3:243-77.

Noll, R. G. 1985. Government regulatory behavior: A multidisciplinary survey and synthesis. In Regulatory policy and the social sciences, edited by R. G. Noll. Berkeley, CA: University of California Press, pp. 9-63.

Olson, M. K. 1995. Regulatory discretion among competing industries: Inside the FDA. Journal of Law, Economics, and Organization 11:379-406.

Rogowsky, R. A. 1986. The economic effectiveness of section 7 relief. Antitrust Bulletin 31:187-233.

Rogowsky, R. A. 1987. The Pyrrhic victories of section 7: A political economy approach. In Public choice and regulation: A view from inside the Federal Trade Commission, edited by R. Mackey, J. Miller, and B. Yandle. Stanford, CA: Hoover Institution Press, pp. 220-39.

Rubinstein, A. 1985. Perfect equilibrium in a bargaining model. Econometrica 53:1151-72.

Scheffman, David T., Malcolm B. Coate, and Louis Silvia. 2003. 20 years of merger guideline enforcement at the FTC: An economic perspective. Antitrust Law Journal 2003:217-36.

Stigler, George J. 1971. The theory of economic regulation. Bell Journal of Economics 2:3-21.

U.S. Department of Justice. 1992. Department of Justice and Federal Trade Commission horizontal merger guidelines. Antitrust Trade and Regulation Report, No. 1559.

Viscusi, W. Kip, John M. Vernon, and Joseph E. Harrington, Jr. 1992. Economics of regulation and antitrust. Lexington, MA: D.C. Heath and Company.

Waller, S. W. 1998. Prosecution by regulation: The changing nature of antitrust enforcement. University of Oregon Law Review 77:1383-449.

Malcolm B. Coate * and Andrew N. Kleit ([dagger])

* Bureau of Economics, Federal Trade Commission, Washington, DC 20580, USA; E-mail [email protected].

([dagger]) Department of Meteorology, The Pennsylvania State University, University Park, PA 16802-5010, USA: E-mail [email protected]; corresponding author.

Received January 2003; accepted June 2003.