An economic evaluation of federal antitrust activity in the manufacturing industries: 1980-1985.
Passage of the Sherman Act nearly a century ago codified and enlarged upon received British common-law court decisions against restraint of trade.(1) Controversy over the administration of domestic antitrust laws has ebbed and flowed ever since. Because administration of these laws exerts a perceptible influence over the national business climate, analysis of antitrust enforcement behavior attracts considerable attention from economists, policymakers and business leaders. The extent of discretion exercised by federal government agencies in interpreting and enforcing the antitrust laws has been questioned repeatedly. Recently, much attention has focused on the attenuation of antitrust under the Reagan administration (1981-1989).(2) Largely because of the Reagan administration's conservative-libertarian philosophy of government, students of antitrust policy generally recognize antitrust administration under the Reagan presidency as a distinct period in the history of these laws.
Reagan appointees to head the federal antitrust agencies have actively pursued deregulation. In particular, antitrust agency leaders appointed during the Reagan administration have made public statements expressing intentions to redirect agency efforts toward a criterion of market efficiency.(3) The views of these administrators and their successors reflect the ascendance of Chicago school tenets of antitrust in the operation of the two primary federal agencies responsible for enforcing the antitrust statutes.(4) The question we address is whether there is behavioral evidence of a radical change in the economic standards of antitrust enforcement during the Reagan administration.
Profound changes in antitrust philosophy should appear as major shifts in the administration of antitrust laws. The purpose of this study is to document and explain the observed allocation of recent federal government antitrust activity among the manufacturing industries. There are four specific objectives:
1. To document changes in the level of professional activity
expended by the Federal Trade Commission (FTC) and the
Antitrust Division of the Department of Justice;
2. To test empirically the influence of economic variables on the
distribution of federal antitrust activity among U.S. manufacturing
3. To test for structural change in the professional resource allocation
processes of the FTC and the Antitrust Division; and
4. To determine the impact of different levels of industry aggregation
on estimated relationships between antitrust activity and its
Long, Schramm & Tollison first formulated and empirically tested a model of the optimal allocation of antitrust effort using categorized case counts from the Antitrust Division.(5) Siegfried,(6) Asch,(7) and Blair and Kaserman(8) extended the theoretical and empirical work of Long et al. Each study used various economic variables to explain variation in Antitrust Division case selection across industries.(9) These studies generally found that economic variables exerted a weak influence over the allocation of antitrust activity, but none provided strong evidence that antitrust enforcement could be characterized as either economically efficient or inefficient. Moreover, several data problems have hindered the implementation and interpretation of empirical studies of antitrust enforcement. Observations on antitrust activity may have been inappropriately pooled over time. Also, overly broad industry definitions may have biased downward the statistical results of the studies because of improper aggregation of the variables. Availability of a more recent data set based on narrower industry definitions makes it possible to address these important measurement issues.
Results reported in this article improve on previously published empirical studies of the allocation of antitrust effort in three major respects. First, this study uses a more accurate measure of antitrust activity. Second, the data are aggregated at levels corresponding more closely to meaningful market definitions. Finally, an improved model formulation accounts for possible shifts over time in the observed behavior of the antitrust agencies.
II. Federal antitrust activity in the 1980's
This section responds to Posner's appeal to focus greater attention on antitrust statistics.(10) Data obtained for our study represent fairly precise measures of real antitrust resource levels, namely, Antitrust Division and FTC professional (lawyer and economist) hours devoted to antitrust activity within four-digit Standard Industrial Classification (SIC) industries from the 1980 to 1985 federal fiscal years. Data sources are the weekly time sheets filled out by agency personnel, which the Antitrust Division began collecting in 1980 and the FTC in 1981.(11) These data consist of all the hours devoted to antitrust matters by professionals: preliminary investigation, full phase investigation, litigation, and compliance. Although individuals may not be able to apportion every working hour to specific activities, it is reasonable to expect that the data are accurate to at least the nearest workday (8 hours).
These data clearly show a severe attenuation of antitrust under the Reagan administration. Professional time expended by the Antitrust Division declined by 14% from 1980 to 1985, while activity directed toward the manufacturing industries dropped by 30% over the same period (table 1). FTC professional effort devoted to antitrust in all industries declined by 42% from 1981 to 1985 and in manufacturing decreased by 59%. From 1981 to 1985, the efforts of the two agencies declined by 40% overall and by 51% across the manufacturing industries only. Thus, antitrust matters in general received less attention by both agencies, while attention to the manufacturing industries experienced even more severe declines.
[TABULAR DATA 1 OMITTED]
The sharpest declines in antitrust activity occurred between 1981 and 1983 and then leveled off. Total expenditures of professional time by both the Antitrust Division and the FTC dropped by at least 18% annually from 1981 to 1983, but the rate of decline shrank to six percent between 1983 and 1984 and three percent between 1984 and 1985. Among the manufacturing industries, however, steep rates of decline in antitrust activity continued until 1984. The rate of decrease in antitrust activity within manufacturing then slowed to a three percent drop between 1984 and 1985.
In manufacturing, both the breadth and intensity of antitrust effort were affected. Of the 451 four-digit SIC manufacturing industries, the number of industries subject to scrutiny by one or both of the antitrust agencies dropped from 230 in 1981 to 193 during 1985. Of the manufacturing industries subject to antitrust activity in 1981, the average level of activity was 0.95 professional-years.(12) The average intensity of effort dropped by over 40% to 0.56 professional-years per industry in 1985.
These substantial shifts in patterns of antitrust enforcement appear inconsistent with the analyses of Posner(13) and Cartwright and Kamerschen.(14) Both studies failed to detect a significant relationship between changes in the political affiliation of presidential administrations and antitrust case-bringing activity. However, both studies examined multidecade trends that may not have captured lags in case-bringing. The measures of antitrust activity employed herein are more sensitive to quickly changing levels of antitrust effort. In any case, the evidence from the 1980's confirms that the Reagan administration carried out its professed antitrust philosophy with substantial effect.
III. Economic determinants of federal antitrust activity
A. Theoretical foundations
For enforcement to be efficient, antitrust activity should be allocated across industries to maximize the expected net benefits from antitrust (Long et al.). Benefits are measured as potential reductions in monopoly welfare loss. Benefits arise from the imposition of a more competitive industry structure or constraints on noncompetitive conduct that yield more competitive performance. Costs include antitrust-related expenditures by private firms and public agencies. Sufficient conditions for existence of an optimal pattern of antitrust resource allocation are a concave benefit function and a convex cost function.
The benefit-cost approach implicitly assumes that reducing monopoly welfare loss is a function of government antitrust activity. Long et al. did not state explicitly in their development of the benefit-cost model the precise nature of this relationship. In the benefit-cost framework, the objective of antitrust policy is to minimize monopoly welfare loss or, equivalently, to maximize expected reductions in welfare loss. To assure concavity of the benefit function, it must be assumed that increases in antitrust activity have a negative but diminishing impact on industry welfare loss.
Like a conventional investment algorithm, maximization of net benefits from antitrust enforcement subject to a fixed budget requires that cases first be brought "against the industry with the highest benefit-cost ratio until the point where the ratio in the highest industry equals the ratio in the next highest. Then cases should be brought against two industries, and so forth."(15) As in Long et al.'s work, the present empirical analysis maintains the weaker assumption that a socially desirable allocation of antitrust enforcement effort implies a positive correlation between relative magnitudes of antitrust activity and measures of welfare losses across industries, ceteris paribus. Potential or expected benefits can be measured by indicators of suboptimal, long-run industry performance (a direct measure of welfare loss) or dimensions of industry structure imperfections (which are known to be determinants, and hence indirect indicators, of noncompetitive industry performance.
B. Empirical implementation of antitrust resource allocation models
1. DATA This study uses both performance and structure measures to explain the allocation of federal agency antitrust
activity across four-digit SIC manufacturing industries. Separate equations are used to explain the behavior of each agency. Independent variables include two performance measures: deadweight loss (Harberger welfare loss(16)) and income transfers from consumers to monopolists (Cowling-Mueller welfare loss(17)). The welfare loss estimates are based on 1972 Census data.(18) Pretax industry profits are used to calculate the welfare loss estimates under the presumption that individuals prefer a dollar retained to a dollar redistributed by the government. Positive coefficients on welfare loss correspond to an efficient allocation of antitrust activity according to the benefit-cost criterion that greater antitrust enforcement resources should be directed toward industries that exercise greater market power.
Industry performance, as indicated by welfare loss measures, can be inferred from industry structure components representing industry size and sources of monopoly power. This study uses a corrected four-firm concentration ratio as an indicator of monopoly power. Corrected concentration ratios attempt to adjust Census data for noncompeting subproducts, geographically fragmented markets, interindustry competition, and foreign shipments.(19) Industry value of shipments are used as a measure of industry size.(20) in accord with theoretical expectations, greater professional resources should be directed toward larger, more concentrated industries under an efficient antitrust resource allocation scheme.(21)
Changes in the internal policies of the antitrust agencies suggest that corrected four-firm concentration ratios may not be the most appropriate measure of industry concentration. The FTC and Justice Department have switched from concentration ratios to Herfindahl-Hirschman indexes in their merger analyses. However, this is probably a matter of little concern given the close correspondence between the two measures of concentration (Shepherd,(22) Nelson(23)). Moreover, our empirical research focuses on relating actual antitrust behavior to an efficiency norm rather than explaining antitrust behavior in terms of stated agency guidelines.
2. ESTIMATION PROCEDURES Long et al. and Siegfried noted that the use of overly broad levels of industry aggregation likely biases downward the estimated effects of explanatory variables on the allocation of antitrust activity. Bias arises from aggregating and averaging into major industry groups observations on industries that are heterogeneous with respect to the included variables. Compared to earlier studies, data obtained for this study are less aggregated and correspond to more meaningful economic market definitions. We will test explicitly for potential estimation bias introduced by inappropriate levels of industry aggregation. It is expected that our less-aggregated data will reduce the bias in previous studies that arises from the use of two- and three-digit SIC industry level data.(24)
Estimation bias attributable to heterogeneous industry groupings is a cross-sectional data problem. Tests of the influence of economic variables on antitrust behavior may also be biased by time-series data problems. Industry characteristics and antitrust behavior both change over time. The probability of aggregating heterogeneous observations increases as the time period considered lengthens. Previous empirical studies, which relied on data gathered by Posner,(25) aggregated statistics on antitrust activity over the period 1945-1970. Merger activity over these years ranged from an inactive period during the late 1940's and early 1950's to a boom period during the 1960's.(26) Early in the same period, a burst of antitrust activity receded into a trough during be late 1950's and early 1960S, only to be followed by another wave of antitrust activism during the late 1960's and early 1970's.(27) With such considerable cycles in industry structure and antitrust enforcement, pooling the data over the entire period may bias downward the estimated effects of industry characteristics on the allocation of antitrust activity. In contrast with previous studies, this analysis covers a shorter period characterized by surging merger activity and ebbing antitrust enforcement. We will test explicitly for possible changes over time in the estimated relationships between antitrust activity and its economic determinants.
Because many (from 45% to nearly 70%) of the four-digit industries included in the analysis were not subject to any antitrust activity during some or all of the years in the study period, ordinary least squares regression generally would lead to biased and inconsistent empirical estimates. The estimation procedure adopted in this study, Tobit analysis, is the appropriate technique for handling econometric models with censored error terms?(28)
IV. Analysis of results
A premise of this analysis is an expected shift in antitrust enforcement behavior under two presidential administrations. Thus, the data are split into two time periods: the first two fiscal years for which data are available (1981-1982 for the FTC and 1980-1981 for the Antitrust Division) and the last 2 years of the study period (1984-1985 for both agencies). The middle years are omitted from the analysis to enhance the power of statistical tests to detect changes in antitrust behavior. Although Reagan entered office in January 1981, examination of the antitrust statistics on professional hours expended suggests that the policies of the Carter administration continued to influence antitrust administration for a year or two. Besides, the new presidential administration could not fully implement major budget changes and many mid-level personnel appointments until fiscal 1982 at the earliest.
A. Changes in antitrust activity over time
1. THE FEDERAL TRADE COMMISSION Table 2 displays results examining FTC activity over time. The explanatory variables in equations 2.1 to 2.4 include Harberger welfare loss (WH), Cowling-Mueller welfare loss (WCM), the corrected four-firm concentration ratio (CCR), and the industry value of shipments (VS). Equations 2.1 and 2.3 include only economic variables to explain the allocation of antitrust activity during the periods 1981-1982 and 1984-1985, respectively. Equation 2.3 tests whether a statistically significant shift in FTC behavior occurred during the middle of the study period. Equation 2.4 addresses the question of lags in changing antitrust behavior by including both economic determinants and lagged antitrust activity as independent variables.
[TABULAR DATA 2 OMITTED]
Equation 2.1 shows that Cowling-Mueller income transfers, concentration, and value of shipments had positive impacts on the allocation of FTC antitrust activity during fiscal years 1981-1982.(29) The coefficients on income transfers and shipments are highly significant.(30) The coefficient on the Harberger deadweight loss measure is negative and significant at the 10% level, apparently conflicting with theoretical expectations for efficient antitrust enforcement. However, the effect of Harberger welfare loss cannot be interpreted independently from the effect of Cowling-Mueller welfare loss because the variables are not independent. Assuming constant marginal costs and unitary price elasticity of demand over the relevant range of quantities, the formula for the Harberger measure of welfare loss is WH = [.5TRd.sup.2], where TR is industry total revenue and d is monopoly price distortion (defined as the ratio by which the monopoly price deviates from competitive price, where 0 [less than or equal to] d [less than or equal to] 1). The Cowling-Mueller measure is defined as WCM = .5TRd. For a given level of total revenue, the two measures are positively correlated. Thus, it is not valid to determine the partial effect of a change in Harberger loss on antitrust resource allocation without considering also the effect from the concurrent change in Cowling-Mueller loss.
The combined effects of the Harberger and Cowling-Mueller welfare loss measures cart be assessed by examining changes over relevant ranges of price distortion. Estimates of d range from 0 to 0.58, with a simple mean of 0. 16. Based on equation 2.1, the joint effect of changes in Harberger and Cowling-Mueller welfare losses with respect to changes in monopoly price distortion are positive and increasing (at a decreasing rate) from d = 0 until a maximum is reached at d = 0.51. From d = 0.51 to 0.58, the effect is still positive but just below the maximum. Over the range of the data, then, monopoly price distortion had a positive impact on the distribution of FTC antitrust activity. Furthermore, the negative sign on Harberger loss and positive sign on Cowling-Mueller loss suggest that the income-transfer criterion may have taken precedence over the allocative-efficiency criterion in FTC antitrust resource allocation decisions.(31)
Comparing equation 2.2 with equation 2.1 suggests that the effects of industry variables on FTC antitrust activity changed from 1981-1982 to 19844985. The decline in the average level of FTC antitrust resources expended on manufacturing is reflected by the reduction in the intercept coefficient from -1.194 to -1.426. For the later period, coefficients on Cowling-Mueller welfare loss and the corrected concentration ratio are positive and significant. The coefficients for Harberger loss and value of shipments are not statistically significant. Smaller t-ratios for the welfare loss measures and value of shipments suggest these variables declined in importance, but the larger coefficient and higher t-ratio for concentration suggests that it assumed a larger role in the later years of the study period.
In the results just presented, comparing the size of the estimated coefficients and their associated t-ratios provides evidence suggestive of a shift over time in the structural relationship between FTC antitrust activity and measures of industry organization and performance. Equation 2.3 pools the data from both periods and incorporates a dummy variable formulation to test statistically for changes in the parameter estimates between 1981-1982 and 1984-1985. The dummy variable (EQ is assigned the value one for observations from 1984-1985 and zero otherwise. Equation 2.3 includes the intercept-shifting dummy variable (D) and its slope-shifting interactions with each of the independent variables ([D.sup.*]WH, [D.sup.*]WCM, [D.sup.*]CCR, and [D.sup.*]VS). The only significant coefficient on a term involving the dummy variable is its interaction with Cowling-Mueller loss ([D.sup.*]WCM), which has a negative sign. During the later period, then, the FTC placed less emphasis on monopoly overcharges in allocating its antitrust resources across industries. This result shows a significant shift in the relationship between FTC activity and its economic determinants. Thus, data from the two periods cannot be pooled into a single regression without dummy variables.
To consider the influence of prior allocations of antitrust activity on later allocations, equation 2.4 includes antitrust activity during fiscal years 1981-1982 to help explain antitrust activity in 1984-1985. Whereas equations 2.1 to 2.3 include only contemporaneous determinants of FTC antitrust activity allocated to manufacturing industries, equation 2.4 provides an indication of continuity in the allocation of antitrust effort from one period to the next. As expected, antitrust activity across industries during 1981-1982 exerted a positive and significant influence on the distribution of antitrust activity in 1984-1985. Broadly speaking, about a third of FTC antitrust activity allocation during 1981-1982 continued in 1984-1985.(32)
There are two ways an antitrust agency can reallocate professional effort over time: changing the choice of industries to target for investigation and changing the level of activity devoted to the targeted industries. In the model used in equation 2.4, positive coefficients would show movement over time toward an allocation of antitrust resources that is more efficiency based, insignificant coefficients would show essentially no change in the efficiency of antitrust resource allocation, and negative coefficients would show departure further from a theoretically efficient allocation of activity. The coefficients on Cowling-Mueller welfare loss and the corrected concentration ratio are positive and highly significant in equation 2.4. Thus, from the period 1981-1982 to 1984-1985 the FTC made small progress toward a more efficient allocation of professional time with respect to Cowling-Mueller welfare loss and industry concentration.(33)
Coefficients on Harberger deadweight loss and value of shipments are insignificant in equation 2.4. Overall, the evidence provided by equation 2.4 suggests that the FTC addressed concerns of noncompetitive industry performance (as indicated by Cowling-Mueller welfare loss) and concentration in reallocating professional resources between the periods 1981-1982 and 1984-1985. However, industry size did not significantly influence decisions regarding incremental changes in antitrust professional resource allocation.
2. THE ANTITRUST DIVISION Table 3 presents four equations that address changes in Antitrust Division behavior from fiscal years 1980 to 1985. Explanatory variables in these equations correspond to parallel equations for the FTC in table 2 above.(34)
[TABULAR DATA 3 OMITTED]
Equation 3.1 shows that Harberger welfare loss, concentration, and value of shipments had positive and significant effects on the distribution of Antitrust Division professional time during fiscal years 1980-1981. Unlike the FTC result, the Cowling-Mueller measure of income transfers did not have a significant influence on the allocation of Antitrust Division effort during the earlier period. Equation 3.2 suggests that the Harberger performance measure declined in importance during 1984-1985, while the coefficient on the Cowling-Mueller measure became statistically significant. The size of the coefficient on the corrected concentration ratio increased slightly from the earlier period. Value of shipments again had a positive and highly significant impact on the allocation of professional time across manufacturing industries.
A dummy variable formulation is introduced in equation 3.3 to test statistically for change over time in the ability of economic variables to explain Antitrust Division activity. The only statistically significant term involving the dummy variable is its interaction with Harberger welfare loss ([D.sup.*]WH), which is negative. Compared to the Carter years, Antitrust Division resource allocation under the Reagan administration exhibited diminished responsiveness to changes in allocative efficiency loss.
Equation 3.4 provides information regarding incremental changes in the allocation of Antitrust Division activity from 1980-1981 m 1984-1985. Like the FTC, Antitrust Division professional activity in the first 2 years of the study period influenced the allocation of activity in the last 2 years. However, the coefficient on 1980-1981 Antitrust Division activity is about half as large as the corresponding coefficient for the FTC. This difference suggests greater flexibility for marginal changes in the allocation of the Antitrust Division's budget, but also may be partially attributable to the fact that the Antitrust Division data are drawn from an earlier period than the FTC data (1980-1981 versus 1981-1982 for the FTC). Unlike the FTC, value of shipments played a significant role in explaining marginal adjustments in Antitrust Division resource allocation. Hence, it appears that the two antitrust agencies operated independently and responded to different economic stimuli in implementing marginal changes in professional resource allocations.
In equation 3.4 the estimated coefficients on Cowling-Mueller welfare loss and the corrected four-firm concentration ratio are positive and highly significant. This result shows that the Antitrust Division made theoretically efficient incremental reallocations of professional resources from the beginning to the end of the study period. In other words, noncompetitive industry structure and performance exerted a positive influence on the direction of reallocation of Antitrust Division professional resources from 1980-1981 to 1984-1985.
B. Defining relevant economic markets
The problem of possible cross-sectional aggregation bias is considered in this section. The major industry group Food and Kindred Products illustrates the problems that may arise from overly aggregated market definitions. During 1981 and 1982, the FTC expended no professional resources on 20 of the 47 four-digit industries within Food and Kindred Products. Over 17 professional years, however, were expended on a single four-digit industry (Roasted coffee). Among the 47 industries, corrected four-firm concentration ratios range from under 0.09 (Fresh or frozen packaged fish) to over 0.9 (Malt beverages). Value of shipments range from under $230 million (Manufactured ice) to over $40 billion (Meat packing plants). Available deadweight loss estimates range from under $500,000 (Creamery butter) to over $120 million (Bread, cake, and related products) and income transfers from under $11 million (Raw cane sugar) to over $620 million (Bread, cake, and related products). This illustration clearly demonstrates the great amount of diversity that can exist among the industries of a major industry group. Yet previous analyses have relied on precisely such highly aggregated data.
Because of the heterogeneity within grossly aggregated industry definitions, the expectation is that increasing levels of industry aggregation will reduce the ability of economic variables to explain the allocation of antitrust resources. Aggregation suppresses vital information about industry characteristics. As an extreme example, consider the case of an industry group Z consisting of two separate four-digit industries (x and y) of equal size. If industry x is atomistically competitive and industry y is a monopoly, then the four-firm concentration ratio in the two-digit industry Z will equal slightly more than 0.50. Thus, the concentration ratio for industry Z is unrepresentative of concentration within its component industries. The added complexity introduced by including more industries and more measures of industry structure or performance exacerbates the problem of unrepresentativeness because of aggregation across diverse industries. To determine how different levels of industry aggregation affect the empirical analysis of antitrust resource allocation, our models of antitrust enforcement are reestimated using a more aggregated data set. In the results presented below, our sample of 420 manufacturing industries is aggregated into a sample of 20 major industry groups.(35)
1. THE FEDERAL TRADE COMMISSION The four equations shown in table 4 correspond to the four equations shown in table 2, except for the level of aggregation of the observations. Also, equation 4.3 differs from equation 2.3 because it includes only the intercept-shifting dummy variable and not the slope-shifting interaction terms. The interaction terms are omitted to conserve degrees of freedom.
Overall levels of explanation for the equations are reasonable, with adjusted coefficients of determination (adjusted-[R.sup.2]) ranging from 0.58 to 0.68. Conversely, standard errors of the estimated coefficients are substantially larger than was the case for the more disaggregated data. Fewer of the coefficients in table 4 are significantly different from zero than in comparable equations in table 2. In particular, in no equation is the coefficient on concentration significant. The results confirm expectations of the downward bias introduced by aggregation of groups of heterogeneous industries into single observations. Although Siegfried addressed the aggregation problem by providing a more disaggregated data set (with a sample of 65 rather than 20 observations), the contrast in the size of comparable t-ratios between four- and two-digit level data provides even more striking evidence of the bias of improper aggregation.
2. THE ANTITRUST DIVISION The results for the Antitrust Division are qualitatively similar to those for the FTC, but the estimated relationships are statistically weaker. Adjusted coefficients of determination for the equations shown in table 5 range from 0.26 to 0.43. In equations 5.2 and 5.4, none of the coefficients is significantly different from zero, yet the adjusted coefficients of determination are above 0.40. This is further evidence of the downward bias introduced by overly broad industry aggregation. Also, the small number of observations heightens the sample problem of collinearity between the two welfare loss measures, further increasing estimated standard errors of the coefficients and lowering the associated t-ratios.
These results cast doubt on the validity of drawing conclusions about the efficiency of antitrust resource allocation across broadly defined industries. If antitrust resources are allocated in stages along the branches of a decision tree, then consideration should be given to industry diversity that may be present at succeeding levels of disaggregation. The implication is that inefficiency may result from allocation decisions made at broad levels of industry aggregation, unless antitrust administrators incorporate into the decision-making process a priori information about further disaggregated industry and firm characteristics.
The empirical results of this study concur with prior studies in at least two respects. Foremost of these is the strong role played by industry size in explaining the allocation of antitrust activity across manufacturing industries. Secondly, this study verifies the expected positive impacts of deadweight loss or income transfers due to monopoly.
The major difference between the present and earlier findings is the strength of the estimated relationship between antitrust activity and its economic determinants. The present findings reflect more favorably toward the efficient management of the two federal antitrust agencies. The results show that both the FIC and the Antitrust Division responded to determinants of efficient antitrust resource allocation in distributing professional time across manufacturing industries. Compared to the earlier set of studies, our results show a more decisive role for welfare loss measures and industry concentration in explaining the allocation of antitrust agency activities the added information provided by the data and methods used herein probably explains this divergence in results.
The results of this study clearly demonstrate the difficulties associated with attempts to aggregate observations that are heterogeneous over time and with respect to industry definitions. Statistical tests show that FTC and Antitrust Division response functions differed under the Carter and Reagan administrations. Under the Reagan administration, the FTC became less responsive to changes in monopoly overcharges and the Antitrust Division grew less responsive to changes in allocative efficiency losses. Analyses that have used antitrust data drawn from a 25-year period erred in pooling observations representing substantially different behavioral relationships. By employing a more disaggregated data set than previously available, this study demonstrates also that overly broad industry aggregation biases statistical results downward. This measurement bias impedes the ability to draw conclusions about the efficiency of antitrust resource use when data are too broadly aggregated.
This study introduces a more accurate measure of antitrust activity than was previously available. Each of the earlier studies merely counted cases brought by the antitrust agency, while this study uses the more precise measure of professional time expended on antitrust activities. Simple tabulation of the number of cases brought against an industry fails to distinguish major from minor ones. Routine cases count the same as precedent-setting, landmark cases. Also, this data set also accounts for the deterrent effect of "spotlighting" by the antitrust agencies, even though the investigation may not proceed beyond the preliminary stage. Such activities are not captured by counting numbers of cases brought. Although counting professional time expended provides an improved measure of the quality of antitrust activity, it is not ideal in all respects. First, measuring professional time expended by the antitrust agencies provides an indication of the quality of antitrust effort only to the extent that the agencies correctly assess the precedent-setting potential of a case many months prior to a decision. Second, many observers note that after 1980 professional time in the agencies was increasingly allocated to studies of economic deregulation rather than antitrust per se.
Results of empirical studies of the allocation of antitrust activity may also differ because of significant changes in underlying antitrust philosophy and administration. Our data clearly showed the diminution of antitrust activity during the 1980's. Statistical analysis confirmed that a significant shift in the response of antitrust resource allocation to changes in welfare loss occurred during the middle of the study period. Even greater differences should surface when comparing results for antitrust activity in the 1980's to results from the earlier studies. The analysis also showed that both the FTC and the Antitrust Division moved marginally closer to an economically efficient allocation of antitrust activity from the beginning to the end of the study period. During times of declining budgets, perhaps more careful attention is directed toward decisions regarding the allocation of antitrust resources across industries.
Although we may belabor the point, let us reiterate that the results presented herein are more encouraging toward antitrust administration than those reported by similar preceding studies. Despite potential limitations of theoretical models as appropriate guides to efficient antitrust behavior, the empirical results are useful in their own right. The ability to determine whether the empirical outcomes conform to an efficient standard is an important issue, yet it is not requisite to conducting the empirical analysis. One may disagree with the theoretical foundations of Long et al.'s benefit-cost model of antitrust, but the empirical findings remain, namely, that the FTC and the Antitrust Division expended more professional time on industries where conventional economic measures of deadweight losses, income transfers, concentration, and value of shipments were higher. Either the academic studies linking industry structure to subsequent performance accurately reflect complaints filed by the FTC and the Antitrust Division, or their staffs behave as if they were followers of such academic literature.
Mueller argues against studies ". . . preoccupied with optimum allocation of scarce antitrust resources. Such myopic preoccupation may lead to a search for the least soggy spot in a swamp, when one should be out surveying the high country."(36) His concern for more attention to broader antitrust policy issues is justified, but it is also important to evaluate antitrust behavior given existing political and institutional constraints. Following Mueller's metaphor, it is reasonable to assert that gaining the high country may require slogging across the swamp. The record of antitrust administration under the Reagan White House affords an exceptionally apt laboratory to investigate the consequences of a sea-change in political philosophy. This study addresses the question of whether campaign rhetoric of 1980 was translated into actual antitrust policy; the empirical results show that the answer is yes. (1) 4 W. Blackstone, Commentaries (1765-1770, reprinted 1976). (2) J. M. Connor, R. T. Rogers, B.W. Marion & W.F. Mueller The Food Manufacturing Industries: Structure, Strategies, Performance, and Policies (1985); Subcommittee on Commerce, Consumer, and Monetary Affairs, U.S. House of Representatives, Subcommittee Oversight Hearing on Federal Trade Commission Operation (Staff memorandum 1983); and Gallo, Craycraft & Bush, Guess Who Came to Dinner-An Empirical Study of Federal Antitrust Enforcement for the Period 1963-1984, 2 Rev. Ind. Org. 106 (1985). (3) William Baxter, the first Reagan nominee confirmed to be assistant attorney general for antitrust, stated "The role of antitrust policy must be to insure that the competitive process is permitted to function. . . .An equally important goal is to minimize government interference in the marketplace through regulation that is more intrusive than it need be." Reilly, Big Shift in Antitrust Policy, 118 Dun's Rev. 38 (August 1981). James C. Miller III, Reagan's first appointee to chair the FTC, hastily replaced existing bureau heads with administrators willing to carry out market-oriented reforms. See, for example, T. F. Walton & J. Langenfeld, Regulatory Reform Under Reagan--The Right Way and the Wrong Way (paper presented at the Annual Conference of the Southern Economic Association, Washington, D.C., November 1987). (4) Posner, The Chicago School of Antitrust Analysis, 127 U.P. L. Rev. 925 (1979). (5) Long, Schramm & Tollison, The Economic Determinants of Antitrust Activity, 16 J. L. & Econ. 351 (1973) [hereinafter Long et al.]. (6) Siegfried, The Determinants of Antitrust Activity, 18 J. L. & Econ. 559 (1975). (7) Asch, The Determinants and Effects of Antitrust Activity, 18 J. L. & Econ. 575 (1975). (8) Blair & Kaserman, Market Structure and Costs: An Explanation of the Antitrust Authorities, 21 Antitrust Bull. 691 (1976). (9) Asch examined case selection behavior of the FTC as well as the Antitrust Division. (10) Posner, A Statistical Study of Antitrust Enforcement, 13 J. L. & Econ. 365 (1970). (11) These data were obtained under Freedom of Information Act requests by the authors. (12) One professional-year is 2,080 professional hours. (13) Posner, supra note 10. (14) Cartwright & Kamerschen, Variations in Antitrust Enforcement Activity, 2 Rev. Ind. Org. 1 (1985). (15) Long et al., supra note 5. Game-theoretic alternatives to the benefit-cost framework for analyzing the allocation of antitrust activity are offered in Lee, Some Models of Antitrust Enforcement, 47 So. Econ. J. 147 (1980) and Perri, The Social Loss From Private Monopoly and Optimal Antitrust Enforcement, I Rev. Ind. Org. 276 (1984). Comparative statics results of these models based on unconstrained antitrust budgets fail to support the benefit-cost model's conclusion that more antitrust resources should be devoted to industries with larger monopoly welfare losses. However, a game-theoretic model of the allocation of a limited antitrust budget upholds the conclusions of the benefit-cost model. See ch. 3 in W. P. Preston, An Economic Evaluation of Federal Antitrust Activity in the Manufacturing Industries: 1980-1985 (unpublished Ph.D. dissertation, Purdue University, 1987). (16) Harberger, Monopoly and Resource Allocation, 44 Am. Econ. Rev. pt. 2, at 77 (papers and proceedings, May 1954). (17) Cowling & Mueller, The Social Costs of Monopoly Power, 88 Econ. J. 727 (1978). Excess profits, as measured using the Cowling and Mueller method, includes components of income redistribution toward monopoly and costs of Posnerian rent-seeking activity. See Posner, Social Costs of Monopoly and Regulation, 83 J. Pol. Econ. 807 (1975). (18) The welfare loss estimates are scaled to units of $100 million and were calculated by other researchers as described in Olson & Bumpass, An Intertemporal Analysis of the Welfare Cost of Monopoly Power: U.S. Manufacturing 1967-1981, 1 Rev. Ind. Org. 308 (1984). (19) L. W. Weiss & G. Pascoe, Adjusted Concentration Ratios in Manufacturing-1972 (FTC Line of Business Publication #20, 1982). Corrected concentration ratios for 1982 were obtained by multiplying unadjusted 1982 Census concentration ratios by factors of proportionality equal to Weiss and Pascoe's corrected concentration ratios for 1972 divided by unadjusted 1972 Census concentration ratios. The concentration ratios range between zero and one. (20) Value of shipments are based on 1982 Census data and are measured in billions of dollars. Other measures of structure such as number of firms, average shipments per firm, and shipments by the top four firms also were investigated as independent variables in more general models to explain antitrust activity over the entire study period. These results are not reported herein. (21) See, however, Asch, supra note 7 for difficulties associated with interpretation of the effect of industry size on the allocation of antitrust enforcement. (22) W. G. Shepherd, The Economics of Industrial Organization 330 (2d ed. 1985) at appendix A3. (23) R. L. Nelson, Concentration in the Manufacturing Industries of the United States (1963). (24) Long et al., supra note 5; Asch, supra note 7; and Blair & Kaserman, supra note 8 used two-digit level data. Siegfried, supra note 6 used dam aggregated at approximately the three-digit SIC industry level. (25) Posner, supra note 10. (26) Shepherd, supra note 22, at ch. 16. (27) Posner, supra note 10. (28) Censoring refers to models in which several observations on the dependent variable are clustered at some limiting value, such as zero. See Tobin, Estimation of Relationships for Limited Dependent Variables, 26 Econometrica 24 (1958). (29) Coefficients estimated by the Tobit procedure can be decomposed to determine two separate effects, as demonstrated in McDonald & Moffitt, The Uses of Tobit Analysis, 62 Rev. Econ. & Stat. 318 (1980). In the context of the current study, the probability effect measures the extent to which changes in an independent variable induce changes in the probability of expending any antitrust activity at all on an industry. The conditional effect shows the extent to which changes in an independent variable induce marginal changes in antitrust enforcement among those industries already under antitrust investigation. The sum of the probability and conditional effects shows the total effect of a change in the value of an independent variable on the level of antitrust activity. Evaluated at the means of the independent variables, results of the decomposition show that about one-third of the total effect is attributable to changes in antitrust activity conditional on already having some FTC resources spent on an industry; the remaining two-thirds of the total response comes from changes in the probability of expending any resources at all on an industry. In other words, the major immediate impact of changes in the independent variables is changing the probability of whether the FTC expends any resources on an industry, rather than changing the level of resources expended on industries in which it is already active. These general results of the decomposition procedure are consistent across the various explanatory variables, across the different time periods, and across both the FTC and the Antitrust Division. (30) Coefficients significant at the five percent level or better are defined as highly significant. (31) When deadweight losses or income transfers are included in separate equations, estimated coefficients on both industry performance measures are positive and significant. Long et al., supra note 5 obtained a similar result. They identified collinearity as the likely cause of the sign reversal on deadweight loss when both measures were included in a single equation, but did not offer further interpretation of the coefficients. Also, because collinearity is often a sample problem, the sample size of 420 observations in the current study ameliorates some of the difficulties posed by the sample size of 20 used in the Long et al. analyses. (32) Coefficients estimated by the Tobit procedure cannot be interpreted in the same manner as those obtained by ordinary least squares regression. The Tobit estimate of 0.329 for the coefficient on antitrust activity during fiscal years 1981-1982 does not imply that a third of this allocation of professional time carried directly into fiscal years 1984-1985. The coefficient may be attributable to a reduction of activity in some industries as well as a shift in the mix of industries subject to any activity at all. See McDonald & Moffitt, supra note 29 for a detailed description of the proper interpretation of Tobit coefficients. (33) Apparently, this result contradicts the finding in equation 2.3 of a decidedly smaller role for the Cowling-Mueller measure in the later years of the study period. This seeming anomaly can be explained by the difference in the formulation of the two equations. Compared to the 1981-1982 period, the Cowling-Mueller welfare loss estimates exerted significantly less influence over total allocations of FTC activity during the 1984-1985 period (equation 2.3). During 1984-1985, Cowling-Mueller losses nonetheless exerted a positive and significant impact on incremental reallocations of FTC resources from the pattern of activity observed during 1981-1982 (equation 2.4). (34) Estimates also were obtained for a parallel set of equations which had as dependent variables the sum of professional hours expended by both the FTC and the Antitrust Division. Because the relative magnitudes of FTC activity dominated those of the Antitrust Division, results were virtually identical to results for the FTC reported herein. Analyzing the behavior of the FTC and the Antitrust Division separately serves as an acknowledgment of the historic division of enforcement responsibilities between the two agencies. (35) Two-digit level observations on the variables representing professional hours expended on antitrust, Harberger and Cowling-Mueller welfare loss, and value of shipments are obtained by simple summation over the relevant four-digit industries. Corrected concentration ratios for the two-digit major industries are computed by weighting the four-digit industry concentration ratios according to the percentage of total value of shipments held by each four-digit industry within the two-digit major industries. Further, it should be emphasized that only the 420 observations for which complete information was available are included in the two-digit level aggregation. The reason is to maintain consistency with the disaggregated data set and to minimize possible bias arising from incomplete information. Except for one industry, both the Antitrust Division and the FTC expended some effort on all two-digit industries in each of the relevant 2-year periods. Because the aggregation thus eliminates the problem of limiting zero values on the dependent variable, ordinary least squares regression is used to estimate the equations presented in this section. (36) W.F. Mueller, Current Policy Issues in Antitrust, The Antitrust Dilemma 35 (Dalton & Levin eds. 1974).
WARREN P. PRESTON Assistant Professor of Agricultural Economics, Virginia Polytechnic Institute and State University.
JOHN M. CONNOR Professor of Agricultural Economics, Purdue University.
AUTHORS' NOTE: The authors are grateful to James Binkley, Deborah Brown, Keith Brown, Eluned Jones, Stephen Lovejoy, Joseph Uhl and anonymous reviewers for commenting on early versions of this work; to Keith Golden and Alan Proctor at the Federal Trade Commission and Leo Neshkes and Ron Wierciock at the Department of Justice for their assistance in providing data requested under the Freedom of Information Act; to Donald Bumpass for providing industry welfare loss estimates; and to Pat DeFeo for consultation on the statistical analysis. Purdue Journal Paper No. 11 748, April 1988.
|Printer friendly Cite/link Email Feedback|
|Author:||Preston, Warren P.; Connor, John M.|
|Date:||Dec 22, 1992|
|Previous Article:||The competitive role of import penetration.|
|Next Article:||Economics, the guidelines and the evolution of merger policy.|