Empirical evidence on FTC enforcement of the merger guidelines.

Coate, Malcolm B. ; Mcchesney, Fred S.

The Justice Department's 1982/1984 merger guidelines identify various factors-- concentration, entry barriers, ease of collusion, efficiency--that would thereafter determine whether the government will challenge a merger. Analysts have criticized enforcement agencies, however, for not following the guidelines, and criticize the guidelines themselves for not identifying the weights attached to the factors. Using a 1982-86 sample of seventy horizontal mergers, we examine which factors influenced Federal Trade Commission decisions to challenge mergers. The relative importance of the guidelines and other factors in merger challenges is measured, and related empirical issues are also explored.

I. INTRODUCTION

The 1982/1984 "merger guidelines" adopted by the Antitrust Division of the Department of Justice (DOJ) and followed by the Federal Trade Commission (FTC) marked important antitrust policy changes by the Reagan administration. The Herfindahl index replaced the fourfirm ratios used in the old (1968) guidelines as the measure of concentration. Other non-concentration factors (barriers to entry, ease of collusion) were elevated in importance. In the 1984 revision of the 1982 guidelines, efficiency considerations were for the first time generally included as a relevant factor.

However, commentators who applauded the newer guidelines have complained subsequently that the Reagan administration did not apply them. The objections of two veteran academic antitrusters, Krattenmaker and Pitofsky [1988, 232], are typical:

Certainly, in many respects, the announced merger guidelines are a substantial accomplishment .... This accomplishment, however, has been almost completely undercut by the Administration's behavior in ignoring those guidelines in practice and instead enforcing, without any public explanation, a merger policy that was not only exceptionally lenient but substantially at odds with professed standards.

In an earlier paper (Coate et al. [1990]), we, along with coauthor Richard Higgins, tested one hypothesis why merger policy (at least at the FTC in 1982-86) has departed from mere enforcement of the guidelines: politics. We showed how external pressures imposed on the Commission by Congress to block mergers, independent of the factors identified in the guidelines as meriting a merger challenge, are a statistically significant factor in the FTC's decision whether to challenge a merger.

In the present paper, we use the same data--seventy horizontal mergers from 1982 to 1986--to examine in greater detail the charges that the FTC has not followed the guidelines, plus several related issues not considered in our earlier article. These include the extent to which various guidelines factors are either necessary or sufficient for the Commission to vote to challenge a merger, and the role of the new efficiency criterion in merger evaluation. An econometric model is presented as one approach to determine how the FTC balances the various guidelines factors. This allows an estimate of the relative importance and influence of lawyers and economists in the evaluation process. Finally, the data also permit an appraisal of the extent to which a structuralist or Chicago approach to competition issues dominated merger votes at the FTC during the relevant period.

Section II discusses briefly the factors identified in the merger guidelines and the process by which the Commission determines whether to oppose a proposed merger. That section also identifies several testable hypotheses concerning the Commission's reliance on the guidelines variables. Section III then explains the data to be used for testing. Section IV, the bulk of the paper, presents empirical evidence concerning the FTC's use of the guidelines and related issues.

II. EVALUATING MERGERS AT THE FEDERAL

TRADE COMMISSION

The Merger Guidelines

The Federal Trade Commission enforces (along with the Justice Department) the federal antitrust laws, including those applicable to mergers. The Commission takes formal action through a majority vote of the sitting (ordinarily five) commissioners. Commissioners vote whether to challenge a merger on the basis of formal staff memoranda that the lawyers of the Bureau of Competition (BC) and the economists of the Bureau of Economics (BE) prepare and submit separately. The memoranda are based in turn on data submitted to the Commission by wouldbe merging partners as required by the Hart-Scott-Rodino Act, and on information developed independently by FTC staff lawyers and economists. Mergers that raise greater anticompetitive concerns usually elicit a supplemental FTC demand for more information, known as a HartScott-Rodino "second request." The parties to the merger also may present their own analyses to the staff, and thus may influence either the Bureau of Competition or Bureau of Economics evaluation.

In analyzing a prospective merger, the FTC claims that its own process follows the Department of Justice merger guidelines.[1] The current guidelines were promulgated in June 1982 (and fine-tuned somewhat in 1984), replacing the guidelines issued in 1968. They focus on preventing a price increase from enhanced market power due to a merger, particularly when no countervailing efficiencies are present. To determine whether such a price increase is likely, the guidelines call for examination of several factors: concentration (including definition of relevant markets), entry barriers, ease of collusion, efficiency, and failing-firm status.[2]

Concentration analysis, based on the Herfindahl-Hirschman index (HHI), is probably the guidelines' best known aspect. The 1984 version establishes three index classifications:

1. Where the post-merger Herfindahl index is under 1000, the merger will be challenged only "in extraordinary circumstances." Thus, a Herfindahl index of 1000 is a safe harbor; mergers falling below that level will rarely be challenged.

2. If the post-merger Herfindahl index is between 1000 and 1800 and the merger increases the Herfindahl by 100 points, the government is "likely" to challenge these transactions unless other factors suggest "the merger is not likely substantially to lessen competition."

3. Finally, for a post-merger index value over 1800, a challenge is "likely" if the merger increases the index by over 50 points, unless other factors suggest "the merger is not likely substantially to lessen competition." However, there will be a challenge in all but "extraordinary circumstances" if the merger raises the Herfindahl index by over 100 points to a level substantially above 1800.

The remaining (non-concentration) factors in the guidelines influence the decision within each Herfindahl index class. Ease of entry will prevent current rivals from raising price for any extended period, and so lowers the likelihood of a merger challenge. Entry is considered easy if a new firm can break into the market within two years of a merger, and difficult if entry would take longer. The guidelines note that likelihood of entry is a function of sunk assets, industry growth rate, economies of scale, and the specific capital needed to compete. But no quantitative measures of these entry determinants are provided, leaving one with the practical two-year standard.

The guidelines analyze ease of collusion or dominant-firm pricing (for reasons other than entry barriers) through proxies: product homogeneity, spatial similarities of merging firms, information available in the market, ease of fringe expansion, market conduct, and past performance. No weights for those proxies are given. AIthough consideration of collusion would appear necessary in all cases to tell an economically-plausible anticompetitive story, the guidelines state that the collusion factors will be most important when the merger decision is a close call. (For simplicity, we refer to these other factors under the heading "collusion," although they may sometimes refer to dominantfirm behavior.)

Greater efficiency (scale economies or lower transportation costs, for example) was generally excluded from consideration in the 1968 guidelines. But the 1984 guidelines listed efficiency gains as a factor that may save a merger that otherwise would be challenged. Efficiencies must be shown by "clear and convincing" evidence, a higher standard than that for anticompetitive factors (high concentration, entry barriers and ease of collusion).3

The final guidelines factor discussed is merger with a failing firm. If one merging firm will soon exit the market anyway, fewer competitive concerns arise. Since the Supreme Court has accepted this argument as a legal defense, antitrust regulators are unlikely to challenge a merger with a failing firm.

Empirical Issues

The merger guidelines are supposed to structure merger regulation to make enforcement decisions consistent, increasing predictability and lowering private transaction costs. Subsequent application of the guidelines, however, has resulted in two general types of complaint.

Failure to Follow Stated Guidelines. Many commentators complain that antitrust enforcers do not follow their own guidelines. Some, such as Leddy [1986] and White [1987], claim that the Justice Department and the FTC have ignored mergers in the concentration ranges that the guidelines label as likely to elicit challenge and have targeted only those with greater Herfindahl index numbers.4 Similarly, Krattenmaker and Pitofsky state [1988, 226], "What is clear, but is very difficult to document by people who lack access to the confidential H-S-R [Hart-ScottRodino] reports and DOJ and FTC internal memoranda, is that the agencies have...not enforced the guidelines." As described below, access to the documents Krattenmaker and Pitofsky refer to permits us to evaluate their claims.

Likewise, questions have been raised about whether some finding of important barriers to entry is required before a merger will be challenged. Almost all economists would deem the existence of significant entry barriers a necessary condition for mergers to reduce welfare appreciably. But the guidelines leave it unclear whether barriers are necessary for a merger challenge, stating that "[t]he more difficult entry into the market is, the more likely the Department is to challenge the merger." Indeed, because the guidelines state that significant mergers in concentrated markets will be challenged in all but extraordinary circumstances, enforcers presumably believe entry to be almost impossible in those cases.

The role of the other (non-entry) collusion factors also merits attention. The guidelines state that these factors "are most likely to be important where the Department's decision to challenge a merger is otherwise close." This suggests that these factors would affect the occasional case, but would not be a systematic consideration in decisions to challenge mergers. The role of both entry barriers and perceived ease of collusion can both be evaluated using the internal FTC data at our disposal.

Use of the new efficiency criterion has also drawn criticism. Some (e.g., Lande [1988]) claim that efficiency has been elevated to a favored position in merger analysis, dominating the more traditional concentration and non-concentration criteria. The data permit an evaluation of this claim as well.

The Relative Importance of Different Factors.

A second set of questions concerns the trade-offs among the different guideline factors. The guidelines leave to the discretion of antitrust enforcers how to weigh the different concentration and non-concentration factors. This is another feature that has been criticized: "Where everything is relevant, nothing is determinative" (Krattenmaker and Pitofsky [1988, 220]). If one factor (high concentration, for example) is a necessary but not sufficient condition for a merger challenge, how are high Herfindahl numbers traded off against other factors, such as entry barriers and perceived ease of collusion?

A related issue concerns the different agency roles of lawyers and economists. One feature frequently noted about antitrust in the Reagan years was that "staff economists at the Commission and the DOJ have gained considerable influence" (Salop [1987, 3]). Their antitrust assessments are said to differ from those of lawyers, particularly because lawyers have greater incentives to litigate in order to increase their human capital for subsequent careers in private practice (Posner [1969, 86]; Clarkson and Muris [1981, 300]). If so, the ultimate trade-off made by the Commission among guideline factors would depend on the differing evaluations by the two groups and their relative influence in Commission votes. Both issues can be examined with the available FTC data.

A final issue concerning FTC application of the different guidelines factors can be addressed with the data. The intellectual battles in antitrust for the past generation have pitted more traditional structuralists against partisans of the Chicago school of antitrust. In some areas of antitrust law, Chicago-school learning has gained ascendency among academicians and even in the courts. It is not clear, however, to what extent antitrust enforcers themselves have adopted a Chicagostyle approach. The differences between the two models, however, are registered largely in terms of the relative importance of the variables (concentration, entry barriers, collusion) identified by the guidelines, meaning that our data can be used to determine which model better predicts Commission decisions.

III. DATA

To explore the empirical issues noted above, we used a data set of seventy merger investigations at the FTC from June 14, 1982 (the date of the new merger guidelines) to January 1, 1987. The sample includes every important horizontal merger that came to the FI?C at that time, as indicated by the fact that it merited a "second request" for data under the HartScott-Rodino Act.s In forty-three of the cases, the Commission allowed the merger, while in twenty-seven the Commission voted to challenge the merger.6

For each proposed merger in the sample, we reviewed the separate lawyers' and economists' memorandum-evaluations of the different guidelines factors. Markets were defined and concentration data were available in all proposed mergers, and we noted the various estimates of the Herfindahl index and the change in the index for both the lawyers in the Bureau of Competition and the economists in the Bureau of Economics- Entry barrier data were available in most of the memos, with sixty-five observations for the Bureau of Economics and sixty-six for the Bureau of Competition. We follow the general interpretation of the guidelines, treating evidence that entry would require at least two years as indicating that important barriers exist, and evidence that entry could occur within two years as evidence that entry was easy. Evidence on the perceived ease of collusion was sparser. We found analysis of an anticompetitive effect in thirty-seven of the Bureau of Economics memos and in a somewhat different group of thirty-seven Bureau of Competition memos; information was obtained from at least one bureau in forty-six cases. We required the memo explicitly to explain how the collusion-based factors did or did not produce an anticompetitive effect before counting the analysis as information sent to the Commission. Efficiency and failing-firm factors were even more rarely considered or discussed. The Bureau of Economics addressed efficiencies in twenty-eight memos and the Bureau of Competition in twenty-three. Finally, the Bureau of Competition raised the failing firm defense in four of the reports.

Before these data are used to answer the questions about FTC merger enforcement presented above, one issue must be addressed. Guidelines factors could conceivably not be evaluated on their merits, but might instead be manipulated to push a result that the FTC staff has decided (for whatever reason) it wishes the Commission to take. If so, the guidelines themselves would afford no predictability in evaluating mergers. It is important to determine, therefore, whether the individual guidelines factors are analyzed on their own merits.

One way to check is to look at the correlations between each pair of guidelines factors. If the factors are not independently evaluated, there should be high correlation among them.? We calculated the simple correlations among the concentration, entry barriers and ease of collusion variables, and found no significant correlation between any pair of variables, as they were evaluated by either lawyers or economists. Apparently, then, variables are evaluated independently and pressure to make or close a case does not generally lead to "cooking" the data.8

IV. EMPIRICAL EVIDENCE

Role of Individual Guidelines Factors

Concentration. The internal FTC evidence allows one to determine if the Commission is actually following the guidelines with respect to concentration. Table I presents a classification for the Bureau of Competition estimates of the Herfindahl indices and changes in index values for the seventy cases. (The Bureau of Economics data is distributed similarly.) As the table shows, in twenty-two of the twenty-seven cases filed the Herfindahl index was over 1800; twelve had index numbers over 3000. There were no complaints when the Herfindahl index increased less than 100 points, and only five when it increased less than 200. In eleven complaints, the increase exceeded 500 points. There is, however, a surprisingly similar pattern for the forty-three closed cases. Twenty-nine cases were not brought even when the Herfindahl index exceeded 1800; in eleven of those cases, the index increased by more than 500 points.

Thus, the FTC's practice from 1982-86 does not corroborate the guidelines' claim that a challenge will be made in all but the most "extraordinary circumstances" when the post-merger Herfindahl index is over 1800 and the change is greater than 100. There were eight cases in which the merger was abandoned when the second request was issued (and thus could not be included in the sample here). Even if all eight involved high concentration and would have resulted in merger challenges, this would mean thirty merger challenges and twenty-seven closed cases in highly concentrated industries with changes of over 100 points. It is hard to believe that truly "extraordinary circumstances" could exist in almost half the cases? Thus, it would appear that the guidelines have been implicitly revised, with the term "challenge" replaced by "investigate."

By comparison, the data indicate that the Commission has not implicitly raised the lower bound for an investigation from the Herfindahl index level of 1000 stated in the guidelines. Five complaints were filed when the Herfindahl index did not exceed 1800. Fourteen cases with index numbers below 1800 were investigated, although closed upon further analysis. Of these, five involved Herfindahls under 1400; five had changes in the index of less than 200 points.

By itself, the table suggests that levels and changes in the Herfindahl numbers may not matter in the FTC vote. Cases were brought in all the Herfindahl classifications, from between 1000 and 1400 to over 3000. Moreover, cases were closed in all these classifications. The relatively small number of cases in most of the cells of Table I makes statistical testing difficult. However, index data can be aggregated into three cells (under 1800, 1800-3000, over 3000) and separated into complaints and closings. In a test of the hypothesis that the decision on a merger is independent of the Herfindahl level, the chisquare statistic is 2.14 well below the critical level necessary to reject the hypothesis of independence. The same test for Herfindahl change yields the same inference, with a chi-square of 1.97. These results should be interpreted with care, given the aggregation necessary to run the test.

The data are also useful for evaluating the guidelines' assertion that the government is "more likely than not" to challenge a merger when the Herfindahl index is under 1800 but "other factors" support issuance of a complaint. If the other factors are taken to mean both entry barriers and ease of collusion, there were six cases (using the Bureau of Competition's evaluations) in which a challenge should have been "more likely than not." In fact, the Commission challenged four of those mergers.

Barriers to Entry. The evidence suggests that entry barriers are virtually a necessary condition for a merger challenge. In the Commission's twenty-seven complaints, the Bureau of Competition claimed barriers would block entry for at least two years in twenty-six cases and the Bureau of Economics agreed in twentytwo of them. The evidence also suggests, however, that entry barriers were not a sufficient condition. There were sixteen cases in which the Bureau of Competition found both low Herfindahl index numbers and difficult entry. The Commission issued complaints in only five of these cases. Moreover, in the forty-three cases where the FTC voted not to challenge, the Bureau of Competition noted high barriers for thirty-one mergers; the Bureau of Economics claimed barriers were high for twenty. Evidently, the FTC has required evidence beyond entry barriers to vote a complaint.

Nor does the Herfindahl statistic interact predictably with entry barriers to generate a complaint. The Bureau of Competition reported entry barriers for twenty of the twenty-eight (71 percent) closed cases with Herfindahls over 1800 and changes of over 50, and for eleven of the fifteen (73 percent) closed cases where the Herfindahl index was below 1800 (or change was less than 50). Figures for Bureau of Economics findings of high barriers were also similar: eleven of twenty-one (52 percent) for high-Herfindahl closed cases and nine of twenty-two (41 percent) for low-Herfindahl mergers.

Ease of Collusion. The guidelines indicate that ease of collusion will be particularly important in marginal cases. In marginal cases, difficulty of collusion should result in case closings, and ease of collusion should cause complaints to be filed. To evaluate this assertion, one must define marginal cases. For example, a marginal case could be a merger with an Herfindahl index under 1800 or a change in the index of under 200.

The evidence indicates that collusion has not played the role indicated in the guidelines. For marginal cases resulting in complaints, the Bureau of Competition presented evidence of feasible collusion in five of eight mergers. The Bureau of Economics staff found collusion plausible in just two of ten marginal cases ending as complaints; the Bureau of Economics even thought collusion difficult in three of the ten. In the marginal cases ultimately closed, the Bureau of Competition found collusion difficult in only four of nineteen cases, but collusion easy in five. The Bureau of Economics found collusion difficult in ten of twenty-one marginal cases that closed, and collusion allegedly easy in only two.

The evidence thus contradicts the claim that perceived ease of collusion is a tiebreaker in marginal cases. For the Bureau of Competition, collusion seems to explain some complaints, but it cannot explain the numerous closings of marginal cases. Conversely, the Bureau of Economics collusion evaluations help explain closings but not complaints. Overall, no predictive pattern emerges from the data on collusion.

Efficiency. The internal FTC data also permit evaluation of the claim that the new efficiency criterion has come to dominate other considerations in merger policy. If efficiency considerations have really affected merger policy, one would expect to see otherwise anticompetitive transactions excused because of expected resource savings. In particular, one would expect in numerous closed cases to find efficiency explanations and evidence suggesting complaints would be favored but for the efficiency argument.

The Bureau of Economics and Bureau of Competition claims of efficiencies can be contrasted with the Commission's final decision on the merger. Perhaps surprisingly, efficiency claims were made more frequently in challenged cases, with the Bureau of Economics claiming efficiencies existed in 33 percent of the complaints filed and the Bureau of Competition in 26 percent. But the Bureau of Economics found efficiencies in only 21 percent and the Bureau of Competition in 7 percent of the closed cases. Obviously, this evidence suggests that the legal and economic staff's efficiency defenses are not generally successful.[10]

One can further explore the efficiency issue by comparing the number of factors (Herfindahl index, barriers or collusion) either favoring or disfavoring complaints in the closed cases where efficiencies were found to those where no efficiencies were claimed. For the Bureau of Economics, in the nine closed cases where efficiencies were claimed, the staff found an average of 1.44 factors indicating the merger would otherwise not have an anticompetitive effect. This is not signficantly different (t- .25) from the average number of 1.38 factors deemed not anticompetitive by the Bureau of Economics in the other closed cases. Thus, a Bureau of Economics efficiency claim apparently did not substitute for other guidelines factors that the Bureau of Economics said supported letting a merger proceed, when the Commission decided to close the case. For the Bureau of Competition, the inference is the same. In the three closed cases for which the lawyers claimed efficiencies existed, they found an average of 1.33 other factors suggesting a problem could exist, as compared to the average of 1.68 factors thought to be at anticompetitive levels for the other cases. This difference is also insignificant (t- .76). Again, at the margin the efficiency factor seems unimportant in explaining FTC merger challenges.

Multivariate Analysis

Analyzing the role of a single guidelines factor without controlling for other factors is potentially misleading. Multivariate techniques may be more appropriate. Multivariate analysis is also useful for investigating two issues in merger enforcement of interest to many: the relative importance of the different guidelines factors, and the differences between lawyers and economists in influencing Commission votes.

Weights of Different Guidelines Factors. Using the internal FTC data, we defined a probit model for the Commission's decisions on merger challenges (Coate et al. [1990]). The dependent variable (VOTE) equalled one if the Commission approved a complaint (including cases in which the parties negotiated a settlement) and zero if the investigation was closed with no action. We explained the merger vote as a function of both the Bureau of Competition and Bureau of Economics analysis of the Herfindahl, barriers to entry, and ease of collusion. To avoid extreme multicollinearity problems, we transformed the Bureau of Competition and Bureau of Economics concentration estimates into two dummy variables.11 For the Bureau of Competition, the concentration variable (BCHERFHI) takes on the value of 1 if the Herfindahl is over 1800 and the change is more than 50, and a value of 0 otherwise. The Bureau of Economics dummy variable (BEHERFLO) was defined as a mirror image of the Bureau of Competition variable, with a value of 1 if the Herfindahl was below 1800 or the change was less than 50 and 0 in all other cases.12

For the first non-concentration factor, entry barriers, we constructed two dummy variables, the first (BCBARHI) with the value 1 if the Bureau of Competition evaluated barriers to entry as high, and the other (BEBARLO) equalling 1 if the Bureau of Economics thought serious barriers did not exist. For the next nonconcentration factor, ease of collusion, again we constructed two variables, one (BCCOLHI) with the value 1 when the Bureau of Competition found collusion likely and the other (BECOLLO) with the value 1 when Bureau of Economics found collusion was unlikely. Finally, the legal failing-firm defense was included by a variable (FAILFIRM) equal to 1 in those four cases in which the Bureau of Competition claimed the defense applied.

Two political pressure variables are also included. The first (CITES) is the number of Wall Street Journal articles about the merger prior to the FTC's decision and is designed to measure the pressure to block high-profile transactions. The second (HEARINGS) is a twelve-month moving average of the number of times Congress summoned FTC commissioners or politically-appointed staff to hearings to defend their antitrust records.

The probit parameter estimates are shown as regression 1 ("base model") in Table II. As expected, Bureau of Competition analysis that concentration, entry barriers and collusion possibilities are at worrisome levels significantly enhances the probability of a complaint; Bureau of Economics evaluation that none of the guidelines factors are worrisome significantly lowers the likelihood of a challenge. The political, the Bureau of Competition and the Bureau of Economics variables all pass independent chi-square tests, indicating that each type of variable affects the Commission's decision making.

As a test of the role of efficiencies, a variable not included in the first model we now insert a dummy variable (EFFCY) equal to 1 for any case in which either the Bureau of Competition or Bureau of Economics claimed that efficiencies were present. As shown in regression 2 ("efficiency model") of Table II, efficiencies themselves are an insignificant factor in FTC votes, and their inclusion in the model has only trivial effects on the size and significance of the other variables. With other guidelines factors controlled for, staff efficiency claims have no apparent influence on a Commission merger decision.

The model is robust with respect to other specifications. One important issue concerns the missing values for the independent variables. As noted above, for example, out of the seventy total cases, the Bureau of Competition discussed collusion in only thirty-seven and the Bureau of Economics in a somewhat different set of thirty-seven mergers. Regressions 1 and 2 are based on a default coding of zero for the Bureau of Competition and Bureau of Economics independent variables when the bureau failed to mention a particular factor. But the results are not sensitive to this coding option, as regression 3 ("recoded model") in Table II indicates. Reversing the coding of the Bureau of Economics variables to match the Bureau of Competition variables, so the missing values would be treated identically, results in only minor differences from the prior estimates. Finally, if the concentration dummies for the Bureau of Competition and Bureau of Economics are multiplied by the Herfindahl index estimated by each bureau so as to create a truncated continuous variable, the estimated coefficients are insignificant, as shown in regression 4 of Table II. This may suggest that the FTC responds less to continuous Herfindahl changes than to discrete changes in classification (e.g., a post-merger HHI above 1800).13

Using the first probit equation (regression 1) and holding the political variables constant at their means, one can investigate the relative importance of the merger guidelines factors in a Commission decision to challenge a merger. If lawyers and economists agree that the Herfindahl index, entry barriers and ease of collusion are all at levels deemed worrisome under the guidelines, the probability of a merger challenge is 97 percent. Suppose, next, that both the Bureau of Competition and Bureau of Economics agree that one of the factors is not a concern. If both bureaus find that the Herfindahl level is low, the probability of an FTC challenge falls from 97 to 43 percent, a statistically significant decline. Since every merger in the 1982-86 FTC sample had an Herfindahl index above 1000 and all but three increased the index by at least 100 points, it is interesting that the Commission would challenge only 43 percent of those mergers, even when both entry barriers and an ability to collude allegedly were present.

The Herfindahl index was apparently of less importance to the Commission during this time than other guideline measures. Although there is still a 43-percent chance the Commission will challenge a merger when lawyers and economists agree that the Herfindahl is low (and the two other factors are high), bureaucratic agreement that either of the other guideline factors is low reduces the probability of a merger challenge even more. When both bureaus agree that the Herfindahl is high and collusion likely, but also that entry barriers are low, there is only a 21 percent probability of a merger challenge. Likewise, when concentration and barriers are judged high, but collusion is thought unlikely, the probability of a complaint is only 27 percent. Ceteris paribus, satisfying the FTC staff that the Herfindahl is under 1800 (or its change is less than 50) appears less likely to shield a merger from challenge than a bureaucratic finding that entry barriers or the likelihood of collusion are low. (Of course, the best way to prevent a merger challenge under the guidelines is to show that the Herfindahl index is below 1000.)

These results are consistent with the conclusion above that, despite the guidelines, the Commission was not "likely to challenge" but rather "likely to investigate" mergers pushing the Herfindahl index over 1800. Even if the Herfindahl is assumed to be high, the decision whether the merger will be challenged depends heavily on the evaluation of the other guidelines factors. Indeed, the regression model shows, if concentration alone were judged high but barriers judged low and collusion difficult, a complaint would have no measurable chance of success.

Bureau of Competition and Bureau of Economics Agreement on Guidelines Factors. Table III notes the positions taken by both the Bureau of Economics and Bureau of Competition with respect to the Herfindahl index, barriers, ease of collusion and efficiencies. There was considerable agreement among lawyers and economists concerning the facts presented to the Commission, ranging from 83 percent on the Herfindahl index to 61 percent on the ease of collusion. Disagreements occurred almost exclusively when the Bureau of Competition thought a variable indicated anticompetitive problems, but the Bureau of Economics did not.

However, the data mask some disagreements, such as differences in market definition or market share. Only 59 percent of the Herfindahl statistics were absolutely identical. Moreover, missing values indicated large potential for further disagreement between the Bureau of Economics and Bureau of Competition. Counting as disagreements any situation where one bureau ventured an opinion and the other did not lowers the agreement rate to 73 percent for barriers, 37 percent for ease of collusion, and 52 percent for efficiencies.

The table indicates that when presented with the same facts, attorneys are more likely than economists to claim that a merger raises issues of antitrust concern. This finding is consistent with the hypothesis that different career incentives induce lawyers to support cases more often than economists. This hypothesis can be tested directly. The internal files reveal that the Bureau of Competition staff supported a complaint in 54.3 percent of the sample cases, while the Bureau of Economics staff supported a complaint in only 30.0 percent. This difference is statistically significant at the 1-percent level.

The effects of bureaucratic disagreement tend to indicate that lawyers have greater influence over the ultimate Commission vote. Table IV (estimated from regression model 1 of Table II with political variables held constant at their means) presents the effects of bureau disagreements over the Herfindahl index, entry barriers and ease of collusion under alternative assumptions concerning evaluation of each of these factors. Suppose hypothetically that both groups at first agree that one variable--the Herfindahl, barriers to entry, or likelihood of collusion--is low while the other two are high. Then let the lawyers change their opinion and also evaluate the first variable as high. As shown in line 1 of Table IV, the probability of a challenge rises by 40 to 64 percent, depending on the variable. Suppose alternatively that both the Bureau of Competition and Bureau of Economics initially find that all merger guidelines factors are high, but that the economists then re-evaluate one of these factors as low. Line 2 of Table IV indicates that the probability of a challenge falls by 12 to 14 percent.[14]

In short, under our econometric model of Commission decision making the lawyers' evaluation of the merger guidelines variables apparently have had a greater impact than those made by economists. Moreover, all the predicted probability changes for lawyers are statistically significant at the .05 and .10 levels. The estimated effects for economists are not significant.[15]

Structuralist vs. Chicago Approaches at the FTC

The probability models above use econometric techniques to determine the simultaneous impact of particular variables on FTC merger decisions. That approach is usually the only one available, because economic theory itself does not specify a priori the exact relationship between the dependent and explanatory variables. In this case, theory suggests two deterministic models that can be constructed with data on Herfindahls, entry barriers and ease of collusion to model FTC merger decisions.

Economists differ on those conditions necessary for a merger to be anticompetitive, although entry barriers would probably be given by all as a necessary condition. If barriers are present, evaluation of a merger hinges on how easily one believes firms could coordinate their actions. So-called "Chicago school" economists typically maintain that something other than mere concentration is necessary for successful, long-term collusion. "Structuralists," on the other hand, tend to hold that once entry barriers have been shown, high concentration can substitute for evidence of the ease of collusion as a sufficient predictor of anticompetitive effect.

The debate is a long-running one. It is interesting, therefore, to use the FTC data on merger challenges to see whether the Chicago-school or structuralist model better predicts FTC merger challenges. We used two definitions that fit the competing theories. A Chicago-school approach would require evidence of all three of the principal guidelines factors: high Herfindahl index, entry barriers and ease of collusion. If any of these factors was missing, a Chicagoan would likely infer that the merger could not be anticompetitive. But a structuralist approach would deem a merger anticompetitive if, in addition to entry barriers, either the Herfindahl index was high or collusion was perceived to be easy. Neither of these approaches is the same as the probit model set out in the regressions above. In that third model, the bureaucracy's evaluation of mergers under its own guidelines is treated as making no one factor either necessary or sufficient; external pressure variables are included as significant predictors as well.[16]

The issue thus is which model better predicts Commission decisions, the Chicago or structuralist model. We used each model to predict the anticompetitive effect of a merger for both the Bureau of Competition and Bureau of Economics data.[17] Under the Chicago model, a challenge would be predicted whenever (under the Bureau of Competition and Bureau of Economics data, alternatively) all three factors were deemed worrisome: concentration, entry barriers and ease of collusion. Under the structuralist model, a challenge is predicted when barriers and either high concentration or ease of collusion is posited.

It is useful to separate the sample into three periods. The first, from the announcement of the 1982 merger guidelines to the 1984 revision, contains sixteen cases. The second period (with twenty-seven cases) runs from the revision to the resignation of FTC Chairman James C. Miller III. A question naturally of interest to economists is whether the FTC evaluated mergers differently during the tenure of Miller, the only professional economist ever to chair the Commission. The third period (also twenty-seven cases) runs from Miller's resignation to the end of the sample.

Table V compares, for all three subperiods, the predictive success of each model. The structuralist model with the Bureau of Competition data predicts Commission merger decisions 67 percent of the time, while the Chicago model is correct 74 percent of the time. Similar results are found by using the Bureau of Economics data with either model. Thus, the deterministic models of Commission decision making do not show that lawyers have more influence than economists on enforcement decisions.

Among the three subperiods,[18] one finds that the Chicago model seemingly outperformed the structuralist model while the 1982 guidelines were in effect, but again there was no significant difference. When the 1982 guidelines were revised, the Chicago model's predictive success fell for both the Bureau of Competition and Bureau of Economics data, with the Bureau of Economics decline being statistically significantly (t= 1.90). However, the opposite result occurred following Chairman Miller's departure, with the increase in the Bureau of Economics's predictive ability again significant (t=3.53).[19]

The predictions of the Chicago and structuralist models can be compared to the results obtained from the bureaucraticpolitical probit model (regression 1 of Table II). Assuming a fitted probability from the regression of over (under) 50 percent predicts a complaint (closing), the probit model correctly forecasted the Commission's decision with 84 percent accuracy. Figures for the three subperiods are 88, 81 and 85 percent, respectively. The probit model generally outperforms the Chicago and structuralist models, not a surprising result since the probit model incorporates significant political variables.

V. CONCLUSION

As the merger wave of the 1980s rolled on, commentators alleged that antitrust agencies had tacitly revised the merger guidelines, challenging only extreme increases in concentration. Also, efficiencies supposedly had been elevated above the other guidelines factors. Moreover, there were rumors that economists were actually taking part in antitrust enforcement decisions.

It is useful to look back and separate fact from fiction. Our data on Federal Trade Commission merger challenges from the mid-1980s provide evidence that for the most part the merger guidelines have not been applied as written. The use of concentration measures is perhaps the best example. However, the data do not reveal an increase in the critical Herfindahl levels. Instead, the evidence suggests simply that concentration has not been used to establish a presumption of guilt. Rather, it has served to determine which cases should be investigated. The cases appear to be examined on their merits and some proof of anticompetitive effect-beyond mere concentration numbers-often required before a complaint is issued. In that respect, the FTC's approach thus has been consistent with economic theory: concentration has been a necessary but not sufficient condition for a merger challenge.

As for non-concentration factors, there is considerable evidence that a finding of barriers to entry was a necessary but not sufficient condition for a merger challenge. On the other hand, there is no evidence (despite critics' claims to the contrary) that explicit inclusion of efficiencies in the guidelines has made any difference. Closed cases where efficiencies were allegedly present in 1982-86 presented competitive concerns similar to those in which no efficiency claims were made.

Moreover, both economist and attorney evaluations of guidelines factors appear to have an impact. At the margin, attorneys seem to have more influence at the Commission if one accepts the econometric model. Finally, both the structuralist and Chicago models predict the Commission's decisions reasonably well over the sample period. But both are inferior to the probability model, shown to be rather robust here, that includes political variables as predictors of FTC merger decisions.

REFERENCES

Clarkson, Kenneth W., and Timothy J. Muris. 'Commission Performance, Incentives, and Behavior," in The Federal Trade Commission Since 1970: Economic Regulation and Bureaucratic Behavior, edited by Kenneth W. Clarkson and Timothy J. Muris. Cambridge: Cambridge University Press, 1981, 280-306.

Coate, Malcolm B., Richard S. Higgins, and Fred S. McChesney. "Bureaucracy and Politics in FTC Merger Enforcement. Journal of Law and Economics, October 1990, 463-82.

Cooter, Robert D., and Daniel L. Rubinfeld. "Economic Analysis of Legal Disputes and Their Resolution." Journal of Economic Literature, September 1989, 1067-97.

Demsetz, Harold. "Barriers to Entry." American Economic Review, March 1982, 47-57. Krattenmaker, Thomas G. and Robert Pitofsky. 'Antitrust Merger Policy and the Reagan Administration." Antitrust Bulletin, Summer 1988, 211-32. Lande, Robert H. "The Rise and (Coming) Fall of Efficiency as the Ruler of Antitrust." Antitrust Bulletin, Fall 1988, 429-65.

Leddy, Mark. "Recent Merger Cases Reflect Revolution in Antitrust Policy." Legal Times, 3 November, 1986, p. 2.

McChesney, Fred S. "Be True to Your School: Conflicting Chicago Approaches to Antitrust and Regulation," Cato Journal, Winter 1991, 775-98.

Posner, Richard A. |The Federal Trade Commission." University of Chicago Law Review 37(1), 1969, 4789.

Priest, George L., and Benjamin Klein. "The Selection of Disputes for Litigation." Journal of Legal Studies, January 1984, 1-55.

Rogowsky, Robert A. "The Justice Department's Merger Guidelines: A Study in the Application of the Rule." Research in Law and Economics, 6, 1984, 135-66.

Salop, Steven C. "Symposium on Mergers and Antitrust." Journal of Economic Perspectives, Fall 1987, 3-12.

U.S. Department of Justice. "Merger Guidelines" (1968, 1982 and 1984 versions), in Horizontal

Mergers: Law and Policy, Section of Antitrust Law, American Bar Association, 1986, 264-336. White, Lawrence J. "Antitrust and Merger Policy: A

Review and Critique." Journal of Economic Perspectives, Fall 1987, 13-22.