Division of Research Graduate School of Business Administration The University of Michigan THE DETERMINANTS OF FTC ANTITRUST ACTIVITY Working Paper No. 286 Mary Claire Mahaney Adrian E. Tschoegl The University of Michigan FOR DISCUSSION PURPOSES ONLY None of this material is to be quoted or reproduced without the express permission of the Division of Research. January 1982

The Determinants of FTC Antitrust Activity I. Introduction This paper investigates the determinants of Federal Trade Commission (FTC) antitrust enforcement activity by examining the frequency with which the FTC initiates cases. The paper elaborates Posner's seminal study of regulatory enforcement' by focusing on the FTC, extending the time span, using statistical techniques, and drawing on recent developments in the theory of regulation. In his 1970 article, Posner examined primarily the activities of the Antitrust Division of the Justice Department. He failed to find a relationship between the government's antitrust enforcement efforts and the Gross National Product (GNP) or the political party occupying the White House. By contrast, we find strong relationships between, on the one hand, both changes in GNP and in party affiliation of the President and the Senate and, on the other hand, the frequency with which the FTC files complaints. In Part II of this paper we review briefly the pertinent theories of regulation. Part III describes the data and methodology. Part IV presents the statistical model and the results. Part V is a summary and conclusion. II. Theories of Regulation Scholars have proposed various theories to explain government regulation of economic activity. Some of these have come to be known as the public interest theory,2 the capture (or regulatees' self-interest) theory,3 the life-cycle theory,4 and the economic theory.5 More recent theories have attempted to explain regulatory behavior by focusing on the bureaucrat's perception of his self-interest,6 or by arguing that regulators act not so much to ensure that regulation itself will be fair, but rather that the process of regulation will be perceived as fair.7

-2 -The FTC's antitrust responsibility is not limited to any specific industry; hence, it is unlikely that in this area of its regulatory responsibility any one industry would have "captured" it. The very form of the FTC, with its prosecutorial and adjudicative structures and procedures, is consistent with the perception of fairness theory. In this paper, therefore, we focus upon the bureaucratic self-interest, public interest, and life-cycle theories. We discuss the economic theory in connection with the public interest theory. A. Bureaucratic Self-Interest According to Niskanen, one would expect to find relationships between regulatory behavior (measured by the frequency and outcome of governmental agency law enforcement) and certain variables. In order for the evidence to support Niskanen's theory, one should find a statistical relationship between political variables and bureaucratic behavior. This is so because if a regulator (specifically, here, a member of an administrative agency) is acting in his self-interest, he will want to be perceived as acting consistently with those people to whom he owes allegiance. Among those to whom he owes allegiance are the President and the members of Congress, because they have the power to reappoint him, to adjust his agency's budget, and to change the scope of his authority by changing the laws that he is empowered to administer. Thus, Niskanen's theory simply argues that bureaucrats will follow the political system's dictates, but the theory does not indicate what these dictates will be. B. Public Interest Recently, the concept of "public interest" has been revived as an explanatory factor in theories of regulation. From the early 1950s to the

-3 -mid- to late 1970s, the academic view of the regulatory process was that it consisted primarily of the use of governmental powers for private or sectional gain. As Levine has observed: The consensus characterized regulation as a device used by relatively small subgroups of the general population, either private corporations or geographical or occupational groups, to produce results favorable to them which would not be produced by the market. The regulatory services provided were variously described as organization of a cartel, wealth transfers as a form of "taxation," enshrinement of capitalistic class interests, or preservation of congressional or bureaucratic power.8 To this description we would add that the subgroups might include those based on ethnicity, race, religion, sex, and age.9 Levine argues that this view cannot explain the phenomenon of deregulation. This is not necessarily correct. The same mechanisms that gave rise to regulation could, at a later date, bring about deregulation. However, this demurrer leaves open the question of that which guides action when the direction of self-interest is unclear. Levine suggests that the previously disparaged public interest theory be revived in a modified form and asserts that the aim of the legislative and executive branches is to promote the general welfare. Regulators may be misled or captured, but the political process will exert pressure to overturn measures demonstrated to be inefficient and undesirable. What constitutes the general welfare is of course a political question. As Reynolds,10 among others, points out, regulation aims for both efficiency and equity. Even if the legislator focuses only on the question of efficiency, his perception of the costs and benefits of regulation is likely to be conditioned by his ideology, including his general beliefs about the

-4 -motivations of firms. Ideology provides a basis for action when one cannot unambiguously determine self-interest. As a general rule, members of the Democratic party are perceived as more willing to impose economic regulation than are members of the Republican party. Assuming that this perception reflects reality (based on Congressional and Presidential voting and veto records, for instance), it follows that, ceteris paribus, a bureaucrat would be more willing to seek out or to prosecute alleged regulatory violators during periods in which the views of public officials reflect Democratic, rather than Republican, partisanship. Thus in years in which the President is a Democrat, bureaucrats would be more likely both to prosecute and to impose liability upon regulatees than in years when the President is a Republican. Similarly, the greater the proportion of Democrats in the Senate and House of Representatives, the more likely it is that the bureaucrat will make an enforcement decision in favor of the imposition of economic regulation. Regardless of their political affiliations, the President and Congress presumably will regard the costs of regulation as more likely to exceed the benefits during periods when the economy is in recession than when it is growing.11 If the bureaucrat empowered to enforce regulations perceives that the President and members of Congress are watching the economy, the bureaucrat would alter his behavior to comport with his perceptions of the attitudes of those to whom he owes allegiance. In addition, the bureaucrat's own budget may reflect the fluctuations of the economy.12 Periods of a wartime economy may induce different behavior. During such a period the government may preempt the economy and become the major buyer of goods and services. While it will not wish to impose costs on firms, one

-5 -can infer that because it is a big consumer itself, and not merely a referee of the market, the government will be particularly sensitive to cases of monopoly power. Thus the net effect is indeterminate a priori. C. The Agency's Life Cycle Bernstein has posited that agencies go through life-cycles.l3 Such a cycle involves start-up, followed by a period of "excessive" zeal. This in turn leads to generalized political opposition and a subsequent "toning-down." Wagner's evidence suggests that this was indeed the case for the FTC.14 Established late in 1915, the FTC spent 1916 becoming organized and searching for a mission. It filed no complaints in 1915 and only one in 1916. Thereafter, it launched itself into a flurry of activity, generating a frequency of complaints not matched after 1925. Clarkson and Muris contend that by the post-World War II era, and especially during the Kennedy-Johnson years, the FTC had sunk into a torpor.15 Reports in 1969 by Cox et al. and the American Bar Association both claimed that the agency was poorly run and absorbed in trivia. The latter went on to state "if change does not occur, there will be no substantial purpose to be served by its continued existence; the essential work must then be carried on by other governmental institutions".16 President Nixon appointed Caspar Weinberger to head the commission in late 1969. Weinberger considered his mandate from the White House to be "carte blanche," allowing for major reorganization and wholesale personnel changes. Clarkson and Muris suggest that Weinberger and his successor, Miles Kirkpatrick, did indeed "revitalize" the FTC.17 By the late 1970s, this revitalization effort had again aroused widespread opposition, which may result in a second period of reduction in regulatory effort.18

-6 - III. Data and Methodology A. Data At the outset, we determined the number of restraint of trade complaints filed annually by the FTC from 1917 to 1979 inclusive. (We excluded 1915 and 1916 for reasons discussed supra.) For the years 1917 to 1969, we relied on Posner's computation of FTC restraint-of-trade cases, which he compiled from the FTC Docket of Complaints and F.T.C. Decisions.19 Posner excluded Robinson-Patman Act20 cases other than those charging predatory pricing. For the years 1970 to 1979, we relied on information provided by a representative of the FTC for the number of restraint of trade complaints filed, excluding all Robinson-Patman Act cases.21 Both Posner's and our tabulations include cases in which the parties (the FTC and the respondents) entered into a consent agreement simultaneously with the filing of the complaint (see Table 1). We next gathered political and economic data for each of the years from 1916 to 1979. We drew upon the U.S. Census Bureau's HISTORICAL STATISTICS22 and STATISTICAL ABSTRACT23 for the annual rates of U.S. economic growth, as measured by the yearly change in real GNP. For the variables indicating political party affiliation of the President and members of the Senate and the House of Representatives, we again relied upon the HISTORICAL STATISTICS24 and STATISTICAL ABSTRACT.25 Finally, for information concerning the FTC's composition (by political party), the chairman's political party, and the political party of the President who appointed a majority of the Commissioners, we drew upon information provided by the FTC and upon Wagner's chart indicating the succession of commissioners at the FTC.26

-7 - B. Methodology We employed ordinary least squares (OLS) regression, fitting models of the form: k Yt = aX. +i Xi, + t i=1 ':L t t where Yt is the number of FTC cases brought in year t (t = 17,..., 79), Xi t is the value of explanatory variable i (i = 1,..., k) in year t, a and.i are parameters to be estimated, and et is the usual residual term. We assume the st are Normally, independently, and identically distributed, with mean 0 and variance o2. IV. Model and Results A. Variables In Part II supra, we suggested that the frequency with which the FTC brought cases would depend upon the regulatory orientation of the President and Congress, the state of the economy, and life-cycle effects. We now discuss the variables that we used to represent these broad concepts. For the identification of regulatory orientation we turned to party affiliation. This proxy is not ideal because both major parties encompass individuals with widely varying attitudes towards government regulation and FTC activism. Nevertheless, we initially created three variables: PRES, HOUSE, and SENATE. PRES is a dummy variable that took on a value of 1 if the President was a Democrat and 0 if he was a Republican. We expected a positive coefficient. HOUSE and SENATE both reflect the percentage of Democrats in each Congressional session. Here, too, we expected positive coefficients. We lagged all three variables by one year to reflect the influences on the FTC at the time that decisions to prosecute are made.

-8 -We also introduced a variable (TRANSIT) for transition from a Democratic to a Republican presidency and vice versa. The variable took on a value of 1 in the six years of such transitions, and 0 otherwise. We expected a negative coefficient to reflect the notion that an incoming President, because of his different priorities, would urge that a number of complaints that the FTC staff had been preparing under his predecessor not be filed. However, a negative coefficient could also result from the administrative turbulence surrounding the change in Administration. For the state of the economy we used the variable MAGNP, a moving average of the change in real GNP in the year in question and the previous year. We expected a positive coefficient, indicating an increased willingness to prosecute in good times and reduced activity in bad. WWII is a dummy variable for the years 1942 through 1945. We did not adjust for the undeclared Korean and Vietnam wars, or for World War I. In our opinion none of these three conflicts equaled World War II in terms of the government's control of the economy, and the intensity of the economy's conversion to war production; furthermore, the First World War coincided with the establishment of the FTC. Both for the reasons outlined earlier and because of Posner's failure to find an association between war and Justice Department antitrust activity, we have no expectation for the sign or significance of the variable. RLYYRS is a dummy variable for the years 1917 through 1924. As we discussed supra, this was a period of peak FTC activity. We therefore expected a positive RLYYRS coefficient. N70S is a dummy variable for the years 1970 through 1979. We included it for two reasons. First, these are the years for which we, rather than Posner, tabulated the FTC data. Given the slight differences between our

-9 -selection of cases and his (see note 21, supra), we would expect a small, negative, and not statistically significant coefficient, were it not for the effect of the second consideration: that the early to mid-1970s reportedly were years of FTC revitalization. Therefore, the net effect we expected was that the coefficient would be large, positive, and significant. D1919 is a dummy variable for the year 1919. This year is clearly an outlier in that the FTC filed almost twice as many cases in 1919 as it did in its next most active year, 1918. Not accounting for 1919 could distort the results for the rest of the variables. We expected to find a large, positive, and statistically significant D1919 coefficient. Finally, in response to the results from models incorporating the above variables, we introduced the variable CHAIR. This variable takes on a value of 1 if, after 1949, the President was a Democrat, and 0 otherwise. This reflects the fact that in 1950 the presidency's relationship to the FTC changed. Prior to 1950, FTC chairmen were elected by the other commissioners; after 1950, they were appointed by the President. B. Results Table II presents the results. In the table, the numbers next to the variables' names are the estimated coefficients. They represent the effect of each variable on the number of complaints filed in a year. For instance, the coefficient of SENATE implies that a one percent increase (decrease) in the percentage of senators who are Democrats is associated with nearly a one-half case increase (decrease) in the number of cases filed in the subsequent year. Similarly, the coefficient of WWII implies that, on average, the FTC filed about 12 to 17 fewer cases in those years.

-10 -The numbers in parentheses are the t-statistics for the coefficients and represent measures of statistical significance. R2 is the coefficient of variation. It ranges between 0 and 1 and represents the percentage of the variation in the number of filings accounted for by the explanatory variables. 1. Political Orientations In all four models, PRES has the reverse of the expected sign, and is statistically significant in model 3. This would imply that a President who was a Democrat would influence the FTC to bring fewer cases than would a Republican. When we introduced CHAIR, the coefficient of PRES, while still negative, became smaller in absolute terms, and not statistically significant. It is therefore clear that much of the anomalous behavior of the first variable is due to the post-1950 period. As we discuss in Part B3 infra, we suspect that the two variables are picking up agency life-cycle effects and that if we could correctly account for these the sign would be as expected. The behavior of SENATE lends credence to the view that Democrats are more likely to favor an activist role for the FTC than are Republicans. SENATE's coefficient preserves its sign and magnitude across all four models. It is also significant in all, although to varying degrees. It is interesting that HOUSE has a negligible effect. Apparently the Senate is more influential in FTC activity; perhaps in part this is because it is the body that confirms appointments to the FTC. TRANSIT has the expected sign, although generally it is not statistically significant. In tests not reported here, distinguishing between Republican to Democrat and Democrat to Republican transitions had no effect on the magnitude or signs of the coefficients.

-11 - 2. Economic Influences MAGNP is consistent as to magnitude, sign, and statistical significance across all five models. A 1 percent increase (decrease) in real GNP is associated with a one-case increase (decrease) in the number of complaints filed. One can infer that the FTC is sensitive to the state of the economy. WWII has a large, negative effect and is statistically significant across all variants of the model. The FTC filed fewer complaints on average during the war than it did in other years. Our results are not consistent with Posner's Antitrust Division conclusions,27 perhaps because we examined the effects of World War II separately from other periods of war and employed a more sensitive statistical approach. 3. Life-Cycle Effects RLYYRS is large, positive, and significant across all models. The FTC was more active, by an average of 25 complaints per annum, during the period 1917-1924 than it was in later years. N70S is also positive and significant across the models. The reported increase in activism associated with the FTC's revitalization is statistically detectable. The variable CHAIR, which has a strong, negative, and statistically significant effect, takes on a value of 1 during the last 2 years of Truman's term and during the Kennedy, Johnson, and Carter presidencies. The KennedyJohnson years coincided with the nadir of FTC effectiveness. The Carter years coincided with a period of lowered activism relative to the revitalization of the early 1970s. The variable's behavior, together with the behavior of the three consciously chosen life-cycle variables, suggest the need for a better theory to account for the process of agency birth, zeal, restraint, decline,

-12 -and effective death (or rebirth); if the latter, the process would presumably recycle through the subsequent phases of zeal, restraint, etc. 4. Outliers The 121 complaints filed in 1919 have a great effect on the overall fit of the model, although not on the signs or significance of most of the other variables. Because the President in 1918 was a Democrat, one effect of dropping the variable (model 4), was to reduce in absolute value the magnitude of the coefficient of PRES. We believe the reason for the relatively high number of complaints in 1918 and 1919, over and above the effect of the agency's youthful enthusiasm, was the influence of Woodrow Wilson. Wagner reports that President Wilson instructed the agency to examine the price of food and materials related to the war effort when the United States entered World War I in 1917.28 The data for 1918 and 1919 may represent the filing of complaints following investigations launched in 1917 and 1918. V. Summary and Conclusion We have examined some determinants of FTC antitrust enforcement activities by regressing the frequency of complaints issued by the FTC from 1915 to 1979 on political, economic, and life-cycle variables. We have found strong relationships between the variables and the frequency with which the FTC has issued complaints. These results support aspects of the regulation theories of Niskanen, Levine, Reynolds, and Bernstein. This research has addressed the issue of the determinants of FTC prosecutorial activity. With respect to the FTC itself, further research calls for a similar statistical examination of the determinants of FTC adjudicatory activity. Moving beyond the FTC, statistically refined research into the prosecutorial activity of the Justice Department's Antitrust Division, as well

-13 -as into the enforcement activities of other administrative agencies, may uncover more evidence to support the revived public interest and life-cycle theories. Finally, similar statistical techniques applied to Antitrust Division activities would allow us to compare the direction and magnitude of the President's impact on the FTC, an independent regulatory agency, with his impact on a Division that accounts to a member of the President's cabinet.

-14 - TABLE 1* FTC RESTRAINT OF TRADE CASES Year Initiated *1917 *1918 *1919 *1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 *1933 *1934 *1935 *1936 *1937 *1938 *1939 *1940 Number 20 64 121 18 26 32 50 51 21 4 8 10 17 12 4 3 4 14 30 33 18 28 31 33 Year Initiated *1941 *1942 *1943 *1944 *1945 *1946 *1947 *1948 *1949 *1950 *1951 *1952 1953 1954 1955 1956 1957 1958 1959 1960 *1961 *1962 *1963 *1964 Number 32 16 14 8 6 9 11 11 10 5 18 16 7 11 29 22 16 13 12 26 7 15 9 12 Year Initiated Number *1965 *1966 *1967 *1968 1969 1970 1971 1972 1973 1974 1975 1976 *1977 *1978 *1979 *1980 18 19 9 15 15 19 26 28 24 24 31 28 20 16 33 NA Source: See text at notes 18-20. *Denotes years in which the White House was occupied by a Democratic President.

-15 - TABLE II REGRESSION RESULTS Model Variables 1. 2. 3. 4. CONSTANT PRES HOUSE SENATE TRANSIT MAGNP WWI I RLYYRS N70S D1919 CHAIR R2 SER F DW -8.44 (-1.23) -3.42 (-1.26) -0.01 (-0.06) 0.43 (1.80)* -4.96 (-1.57) 0.91 (4.46)* -15.02 (-3.39)* 24.03 (6.93)* 7.78 (2.73)* 80.33 (10.55)* -6.87 (-2.73)* 0.86 7.05 32.81 2.17 -8.37 (-1.25) -3.42 (-1.27) 0.45 (3.36)* -4.95 (-1.59) 0.91 (4.52)* -15.11 (-3.67)* 23.96 (7.41)* 7.83 (2.92)* 80.34 (10.66)* -4.95 (-2.76)* 0.86 6.98 37.15 2.16 -8.95 (-1.26) -5.90 (-2.20)* 0.46 (3.25)* -6.15 (-1.88)* 0.80 (3.83)* -11.66 (-2.81)* 25.99 (7.78)* 6.66 (2.37)* 81.26 (10.19)* 0.84 7.40 36.38 2.00 -11.84 (-1.01) -1.35 (-0.29) 0.49 (2.12)* -6.70 (-1.23) 0.88 (2.51)* -16.63 (-2.30)* 34.11 (6.28)* 8.66 (1.84)* -8.03 (-1.84)* 0.57 12.27 8.94 2.12 NOTE: 1) *indicates significance at the 10% level. 2) R2 is the coefficient of determination; SER is the standard error of the regression. F is the F statistic. DW is the Durbin-Watson statistic for first-degree autocorrelation.

-16 -1. Posner, A Statistical Study of Antitrust Enforcement, 13 J. LAW. & ECON. 365 (1970). 2. Levine, Revisionism Revised? Airline Deregulation and the Public Interest, 44 LAW. & CONTEMP. PROB. 179 (1981). 3. See Posner, Theories of Economic Regulation, 5 BELL J. ECON. & MNGMT. SCI. 335, 341-43 (1974). 4. M. BERNSTEIN, REGULATING BUSINESS by INDEPENDENT COMMISSION (1955). 5. Stigler, The Theory of Economic Regulation, 2 BELL J. ECON. & MGNMT. SCI. 3 (1971); Posner, Theories of Economic Regulation, 5 BELL J. ECON. & MNG4T. SCI. 335 (1974). 6. W. NISKANEN, BUREAUCRACY and REPRESENTATIVE GOVERNMENT (1971). 7. Noll, What Is Regulation? California Institute of Technology, Social Science Working Paper #324 (June 1980). 8. Levine, supra note 2, at 180. 9. Thurow, Equity, Efficiency, Social Justice, and Redistribution, 20 NEB. J. ECON. & BUS. 5 (1981). 10. Reynolds, Foundation of an Institutional Theory of Regulation, 15 J. ECON. ISSUES 641 (1981). 11. Several investigations of Justice Department and FTC enforcement activity have found no relationship between frequency of complaints by industry and measures of industry monopoly power. Moreover, we have no evidence that links the magnitude of the benefits of antitrust activity to phases of the business cycle. See Long, Schramm & Tollison, The Economic Determinants of Antitrust Activity, 16 J. LAW & ECON. 351 (1973); Siegfried, The Determinants of Antitrust Activity, 18 J. LAW & ECON. 559 (1975); Asch, The Determinants and Effects of Antitrust Activity, 18 J. LAW & ECON. 575 (1975).

-17 -12. The extent to which Congress and the President have exercised control over the FTC's budget is not entirely clear. Typically, the Office of Management and Budget (OMB) has sliced the FTC's budget requests (13 percent on average from 1971 to 1979), and the President has perfunctorily approved OMB's suggestions. Congress seems generally to have been a bit more generous towards the FTC, and until the late 1970s granted the agency at least the (OMB) proposed amounts. On the issue of budgetary oversight of the FTC, see R. KATZMANN, REGULATORY BUREAUCRACY 122-25 (1980) and K. CLARKSON & T. MURIS, THE FEDERAL TRADE COMMISSION since 1970 at 18-22, 55-58,301-02 (1981). 13. M. BERNSTEIN, supra note 4. 14. S. WAGNER, THE FEDERAL TRADE COMMISSION 27 (1971). 15. K. CLARKSON & T. MURIS, THE FEDERAL TRADE COMMISSION SINCE 1970 at 3 (1981). 16. Id. 17. K. CLARKSON & T. MURIS, supra note 15, at 3-5. 18. Id. at 302-03. 19. Posner, supra note 1, at 369. 20. Robinson-Patman Act, amending 15 U.S.C. ~13 (1964). 21. Due to the form in which the FTC had compiled the relevant data, we were unable to separate predatory pricing Robinson-Patman Act cases from other Robinson-Patman Act cases. In this sense, then, our sample potentially excludes cases that Posner's sample includes; we have reflected that disparity in a dummy variable. Furthermore, because the FTC records the complaints it issues on a fiscal-year basis, and we continued to use Posner's style of recording cases (on a calendar-year basis), we assigned a proportionate number of cases from each fiscal year to each calendar year.

-18 -22. U.S. BUREAU OF THE CENSUS, 1 HISTORICAL STATISTICS COLONIAL TIMES TO PRESENT 226-27 (1975). 23. U.S. BUREAU OF THE CENSUS, STATISTICAL ABSTRACT OF 440 (1980). 24. U.S. BUREAU OF THE CENSUS, 2 HISTORICAL STATISTICS COLONIAL TIMES TO PRESENT 1083 (1975). 25. U.S. BUREAU OF THE CENSUS, STATISTICAL ABSTRACT OF 509 (1980). 26. S. WAGNER, supra note 14, at 25. 27. Posner, supra note 1, at 368. 28. S. WAGNER, supra note 14, at 22. OF THE UNITED STATES, THE UNITED STATES OF THE UNITED STATES, THE UNITED STATES