Download as pdf or txt
Download as pdf or txt
You are on page 1of 32

Predatory Pricing and the Acquisition Cost of Competitors

Author(s): Malcolm R. Burns


Source: Journal of Political Economy, Vol. 94, No. 2 (Apr., 1986), pp. 266-296
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/1837405
Accessed: 24-07-2016 18:05 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Political Economy

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

Predatory Pricing and the Acquisition


Cost of Competitors

Malcolm R. Burns
University of Kansas

This paper investigates whether predatory price cutting reduces a


trust's cost of acquiring its competitors. A variant of the Litzenberger-Rao valuation model is estimated with the expenditures for 43
rival firms purchased by the old American Tobacco Company between 1891 and 1906. The coefficient estimates indicate that, ceteris
paribus, alleged predation significantly lowered the acquisition costs
of the tobacco trust both for asserted victims and, through reputation effects, for competitors that sold out peacefully. Although
qualified by data limitations, these results support the classical view
of predatory pricing as a systematic business practice.

This paper began as an independent study project for aJ.D. degree at the Universitv
of Kansas School of Law under the direction of Richard J. Pierce, Jr. It was completely
rewritten with the support of a National Fellowship at the Hoover Institution. I am very
grateful to Dennis L. Bark, W. Glenn Campbell, Constantin Galskoy, John H. Moore,
and Thomas G. Moore for making the resources of the Hoover Institution available to
me. The paper was subsequently revised with the generous assistance of Kurt V. Krue-

ger and Anthony L. Redwood of the Institute for Public Policy and Business Research

at the University of' Kansas. Eileen M. McChesney verified the literature citations.
Valuable comments and suggestions were contributed by Timothy Bresnahan, Thomas

J. Campbell, Harold Demsetz, Aaron Director, Michele U. Fratianni, Mark Gritz,


Robert E. Hall, 0. Maurice Joy, Frederick C. Kirby, Phillip C. Kissam, Stephen Langlois, Robert H. Litzenberger, Robert T. Michael, Lawrence Radecki, James B. Ramsey,

Al H. Ringleb, Sidney A. Shapiro, and De-Min Wu, as well as seminar participants at


the University of Chicago, Indiana, Iowa State, Kansas, Stanford, and Texas A & M.
The revision has been improved substantially by the criticisms of two anonymous refer-

ees. However, my greatest debt is to George J. Stigler, who sponsored my fellowship


application and graciously reviewed multiple drafts of the manuscript. I am responsible
for any remaining errors.

Llournal of Political Ecoioonmo, 1986, vol. 94, no. 21


? 1986 by [he University of Chicago. All rights reserved. 0022-3808/86/9402-0003$01.50

266

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

267

I. Introduction

There have been numerous empirical studies of predatory (or below-

cost) pricing as a monopolization strategy since McGee's (1958) semi-

nal article exonerating the Standard Oil trust.' However, almost all of
them apply the "case study" technique that was judiciously abandoned
long ago in other fields of economics. Typically, it involves a careful

rereading of the historical records of a classic monopoly to determine,


in only an impressionistic way, whether predatory price cutting was

actually used to discipline or eliminate smaller competitors. This approach has produced some fascinating information about business

conduct, but the analysis is so ad hoc and subjective that none of the
evidence is widely accepted.2 Put simply, economic research has developed few hard facts on the frequency or effects of below-cost pricing
by the great trusts at the turn of the century. Hence it is easy to
understand why the empirical literature has been largely ignored by
the policy-oriented, and more influential, discussions of this practice

in the law reviews.3


This paper offers an entirely new analysis of predatory price cutting with considerable promise for future empirical research. It also

expands the modest base of established facts that should be influencing proposed rules of law. Instead of reassessing contemporary ac-

counts of below-cost pricing, the analysis accepts these claims as provisionally true and examines one of their most frequently alleged
consequences: the forcing of competitors to merge with a trust at
distress asset values.4 The effects of predation on the acquisition costs

l They are Zerbe (1969), Elzinga (1970), Koller (1971), Yamey (1972), Kamerschen
(1974), Mariger (1978), and Burns (1982). See also the interesting experiments of Isaac
and Smith (1985).

2 See the review of the empirical literature in Scherer (1980, pp. 336-37).
3 The leading articles are Areeda and Turner (1975), Williamson (1977), Baumol
(1979), Joskow and Klevorick (1979), McGee (1980), Easterbrook (1981), and Zerbe
and Cooper (1982). The rule proposed by Areeda and Turner (1975, p. 733)-that

only a product price below average variable cost is predatory and therefore unlawfulhas been adopted (with modifications) by appellate courts in the following cases: Inter-

national Air Industries v. American Excelsior Co., 517 F.2d 714 (5th Cir. 1975), cert.
denied, 424 U.S. 943 (1976); Pacific Engineering and Production Co. of Nevada v. KerrMcGee Corp., 551 F.2d 790 (10th Cir. 1977), cert. denied, 434 U.S. 879 (1977); Chillicothe Sand and Gravel Co. v. Martin Marietta Corp., 615 F.2d 427 (7th Cir. 1980);
Northeastern Telephone Co. v. American Telephone and Telegraph Co., 651 F.2d 76
(2d Cir. 1981), cert. denied, 455 U.S. 943 (1982); William Inglis and Sons Baking Co. v.
ITT Continental Baking Co., 668 F.2d 1014 (9th Cir. 1981), cert. denied, 103 S.Ct. 57
(1982); 0. Hommel Co. v. Ferro Corp., 659 F.2d 340 (3d Cir. 1981), cert. denied, 455
U.S. 1017 (1982); and MCI Communications v. American Telephone and Telegraph
Co., 708 F.2d 1081 (7th Cir. 1983), cert. denied, 104 S.Ct. 234 (1983).

-1 The classic example of these allegations is Tarbell (1904, 1:63-67, 154-61, 202-7).
For the tobacco industry, see U.S. Supreme Court (1911, testimony of Gustavus A.
Puryear, 4:170-71).

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

268 JOURNAL OF POLITICAL ECONOMY

of an aspiring monopolist can be studied objectively with a multiple


regression model of competitor valuation, and the results will support

more reliable conclusions about the prevalence of such misconduct

than heretofore possible. The reason is that, when mergers for monopoly are lawful, predatory price cutting is a rational alternative
strategy only if it lowers the amounts paid for competitors enough to

offset all the costs of warfare (McGee 1958, p. 141). Therefore, a


savings in the trust's acquisition expenditures from what they would
be under peaceful conditions is a prerequisite for the intelligent use

of predatory pricing to monopolize an industry. Thus, given the various financial determinants of competitor market values, coefficient
estimates signifying reduced costs of purchasing asserted victims of a
predatory campaign would tend to validate historical reports of this
practice. Conversely, the failure to find any impact on acquisition

costs implies either that predation was not in fact widespread or that

the trust used it for some other purpose that is more difficult to
quantify, such as forestalling entry or accelerating the consolidation
of major rivals.

The analysis is carried out on the amounts paid to 43 competitors

acquired between 1891 and 1906 by the old American Tobacco Company and two affiliated corporations as they established quasi
monopolies in the plug, smoking, snuff, and fine cut branches of the

industry.5 This sample contains virtually all of the largest and most
important firms merged into the tobacco trust. The majority were
proprietorships or partnerships, and all of the corporations acquired

were closely held, so no stock price data are available. According to


contemporary accounts, several competitors sold out following a
predatory attack-allegedly financed by the trust's large profits on

the sale of cigarettes while the other firms in the sample evidently
were absorbed peacefully.

The empirical results are consistent with the classical view of below-

5 The discussion uses "American Tobacco" and "tobacco trust" interchangeably.


From 1900 to 1910, the last full year before dissolution, the average market shares of
the trust were plug, 77.2 percent; smoking, 68.8 percent; snuff, 90.7 percent; and fine
cut, 73.9 percent (U.S. Bureau of Corporations 1915, pp. 49, 84, 127, 138). Fine cut was
a finely shredded product that was chewed by most users (U.S. Bureau of Corporations
1909, p. 240).

6 See U.S. Bureau of Corporations (1909, pp. 96-98). American Tobacco dominated
cigarette manufacturing with an average market share of 86.1 percent for the period
1891-1910 (U.S. Bureau of Corporations 1915, p. 153, table 52). The sample excludes
cigarette acquisitions for three reasons. In the first place, they are quite few in number.
Second, the unusually diverse outputs of these competitors make the measurement of
earnings and risk doubtful using the procedures described in the App. Finally, none of
the available cigarette firms was an alleged victim of predatory pricing before its merger into the trust. The early history of the cigarette branch is reviewed in U.S. Bureau
of Corporations (1909, pp. 324-43).

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

269

cost pricing as a systematic business practice. The coefficient estimates


of the valuation model indicate that alleged predation reduced the
acquisition costs of American Tobacco both by lowering the amounts

paid for asserted victims and by creating a reputation for misconduct


that lessened expenditures for competitors acquired peacefully thereafter. Other things being equal, the estimated direct savings range up
to 60 percent of what some targeted rivals would have cost if they had

not been preyed on, and the trust's reputation produced an additional discount averaging 25 percent. In general, these rates are statistically significant at the .01 level. Thus, in sharp contrast to the leading case studies of McGee (1958) and Elzinga (1970), the tobacco data
contain evidence that predatory pricing can be an efficient and ra-

tional monopolization strategy.


These conclusions are qualified by limited data and measurement

errors in two financial variables. The paucity of theoretical research


on the valuation effects of predatory pricing also gives the regression
models a somewhat ad hoc appearance. However, these shortcomings

are explicit and cannot be any less severe than the ambiguities inher-

ent in previous case studies. Moreover, both the sample and the results are very interesting. Since the actual effects of pricing below cost
are completely unknown, this evidence can contribute to developing
an appropriate policy response. And of greater importance, perhaps,
is the prospect that the analysis could be refined and extended to

other classic trusts if, as seems likely, a diligent search of trial records,
government reports, and trade publications yields suitable data on the

acquisition costs of their competitors.


The remaining discussion is organized as follows. Section II applies
the Litzenberger-Rao (1971) model, which underlies all the empirical
testing, to review what is known about the valuation of trust competitors. Then the regression equations are developed in Section III, and
their coefficient estimates are presented in Section IV. The paper

concludes in Section V with a brief commentary. A complete descrip-

tion of the measures and data sources used in the analysis is relegated
to the Appendix.

II. The Valuation of Competitors


While McGee (1958, pp. 138-43) fully identified all the costs of a
predatory campaign, he gave relatively little attention to the costs of a

peaceful merger strategy. As a result, subsequent studies essentially


took it for granted that "competitive firms can be purchased for competitive asset values or, at worst, for only a little more" (McGee 1958,
p. 139). This statement suggests that the amount paid for any competitor is typically fixed by the value of discounted normal profits (Telser

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

270

JOURNAL

OF

POLITICAL

ECONOMY

1966, p. 261). However, such conceptions of the valuation process are


oversimplified and obsolete. Recent research in finance implies that
the market values of competitors are neither constant nor indepen-

dent of the trust's conduct-particularly when a Standard Oil or


American Tobacco acquires rivals sequentially, one (or a few) at a

time, rather than buying all of them at once.' Under these circumstances, a series of peaceful mergers is likely to make remaining competitors more expensive to purchase later on; and a trust might
significantly reduce its acquisition costs by selectively, but systematically, cutting product prices below operating costs. Then predatory
pricing could be a rational monopolization strategy.

These conclusions follow directly from modern valuation theory.


To illustrate, consider the Litzenberger-Rao (1971) model, which can

be estimated for tobacco competitors.' Assuming a perfect capital


market, risk-averse investors, homogeneous expectations of a uniform and finite growth horizon, and no predation in the industry, the

total market value (V-) of the ith firm is given by'9

vi= E_ brSE,
+TBiL R7mRA)1 (1)
I I +~~~R-I
where EZ = expected real earnings; SE? = the standard deviation of

real earnings; I = the riskless interest rate; R, = the required rate of


return per share or the firm's cost of capital; r = the correlation
coefficient between the returns per share for the ith firm and the
market portfolio, a constant when the firms are competitors (Litzenberger and Rao 1971, pp. 269-72); b = the marginal return demanded by investors per unit of nondiversifiable risk, a positive con-

stant for all i; ri = the expected return on new investment (rr, > RJ);
B. = the anticipated expenditure on new investment; and T = the

number of years each firm is expected to earn ai > RI.


7The wholesale consolidation of established oil and tobacco competitors evidently
was precluded by the sheer number of firms that had to be absorbed for effective
market control. Prior to about 1907, Standard -Oil purchased some 223 independent
companies (McGee 1958, p. 144), while American Tobacco bought approximately 250
firms in all branches of the industry (U.S. Bureau of Corporations 1909, p. 13). The
largest single consolidation in the sample occurred during November 1898, when
American Tobacco merged with six large competitors to form the Continental Tobacco
Co.; but the new firm controlled properties that accounted for only 46.2 percent and
39.7 percent of U.S. plug and smoking output, respectively (Burns 1982, pp. 41-43).
8 Keenan (1970) critically surveys previous attempts to estimate other valuation models. Estimates of the Litzenberger-Rao model are also presented in Meeker, joy, and
Cogger (1983).

9 Only one assumption in this list can be verified for the 1891-1906 sample period. A
perfect capital market requires zero taxes, and there was no federal taxation of corporate or individual income before 1909 and 1913, respectively (Pechman 1983, pp. 302,
308).

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

71

Equation (1) asserts that a firm's market value is normally a linear


function of three components: the value of "no-growth" future earnings discounted at the pure interest rate; an offset for
nondiversifiable risk; and the present value of future earnings
growth, which depends on the expected rate of investment, its net
profitability, and the time horizon of extraordinary growth opportu-

nities. Since the capitalization of these elements generates multipliers


greater than unity, relatively small changes in earnings, risk, or

growth prospects can have large absolute effects on market value.


With the additional assumption that rivals are purchased competi-

tively,'t Vi becomes the amount paid when the ith firm is absorbed
peacefully. Then (1) will identify the theoretical impact of peaceful
merger and predatory pricing on a trust's acquisition costs.

Some of the analysis is familiar or quite straightforward. A reputed


advantage of the pure merger strategy is that additional monopoly
profits accrue immediately from the purchase of large competitors
when the trust continually restricts output as much as possible

(McGee 1958, p. 139). But this behavior increases both the expected
earnings of remaining competitors and their opportunities for expan-

sion, other things being equal (Stigler 1940, pp. 522-24; Worcester

1957, p. 340; Mallela and Nahata 1984, pp. 7, 16). Hence their market values in (1) should rise as well, such that a trust that acquired

each of its rivals peacefully for V, might exhaust all current and future
monopoly profits. In contrast, a vigorous predatory attack could
significantly reduce acquisition costs. If the trust effectively disguised
its misconduct by initiating the price cutting from a secretly controlled

subsidiary, known as a "bogus independent, the ensuing decline in


the profits of a targeted competitor probably will be attributed to

intensified and enduring competition.'" This erroneous perception


will lower earnings projections and reduce E.. The price cutting

I() This assumption is supported by the fact that rivals of American Tobacco also tried
to acquire three of the more important competitors in the sample (see U.S. Bureau of
Corporations 1909, pp. 73-75, 87; U.S. Supreme Court 1911, testimony of George A.
Helme, 2:434).

" The extensive use of bogus independents by the tobacco trust is alleged in U.S.
Bureau of Corporations (1909, pp. 13, 20-21, 110, 224-25) and U.S. Supreme Court
(1911, original petition, 1:21, 40, 66, 68-70, 72-73).
12 Most allegations of predatory pricing by the American Tobacco Co. describe warfare that was closely tailored to the business of a single rival. Evidently, many of them
sold distinctive products in small territories. Hence the trust could attack each one
effectively, while substantially limiting its own losses, by selling a close imitation priced
below cost only in the competitor's marketing area. Some examples are presented in
U.S. Supreme Court (1911, original petition, 1:72-73; answer of R. P. Richardson, Jr.
and Co., 1:269-70; testimony of Percival S. Hill, 2:283-85, 340-41; testimony of Caleb
C. Dula, 2:539-40; testimony of Gustavus A. Puryear, 4:170-71).

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

272

JOURNAL

OF

POLITICAL

ECONOMY

would also increase SE, by accentuating the downward fluctuations in


the firm's earnings trend. Finally, predation should diminish immediate growth opportunities and simultaneously curb investment as
the competitor retrenches operations to conserve cash (Telser 1966,
pp. 260-68). Any or all of these outcomes lower the victim's market
value in (1) and presumably reduce the amount it demands for a
merger (Yamey 1972, pp. 130-31).

However, theoretical research on the valuation consequences of

predatory price cutting leaves two other issues up in the air. One
interesting question arises when the trust peacefully acquires a competitor after having previously attacked other firms. Assuming that
an infamous reputation can be established through prior episodes of

below-cost pricing, does the notoriety itself lessen the amounts paid
for subsequent acquisitions? An effective predatory reputation would
create a legitimate fear that the competitor will become the next casualty of price warfare-unless, of course, it accepts the trust's merger
offer. Such intimidation may induce the firm to sell out for less than
Vs. This hypothesis, that a mere reputation for predatory pricing

generates additional savings in acquisition costs, can be studied empirically as described in Section III.
The second, and more difficult, question concerns the proper way
to test for any direct impact of predatory pricing on a trust's acquisition costs. The answer depends in part on how previous studies are

interpreted. Some of them can be read as asserting a virtual impossi-

bility theorem-that, because the practice is so costly, predatory price


cutting is self-deterring (Easterbrook 1981, pp. 267, 318) and no
thoughtful trust would ever deliberately use it (McGee 1958, p. 168).
These conclusions become suspect if predatory pricing generates
gross savings to offset at least some of its costs by substantially reducing the amounts paid for victims below V, in (1). Predation works
here, if at all, by countering the tendency for output restriction to
raise the market values of competitors. This test is very interesting
and well worth carrying out, but it is not especially demanding be-

cause V, represents the "oligopoly" value of a competitor-the highest


possible valuation, attributable to peaceful market conditions where
the firm prospers and grows larger under the monopolistic price
umbrella of the trust. Thus it should not be too surprising if predatory price cutting tends to yield acquisition costs lower than V1.
The opposing viewpoint in the academic literature asserts that predation is a strategy that, like warfare generally, rational men use only
when less expensive measures fail (Scherer 1980, pp. 336-37). Yet it
is widely regarded in the industry as an available option and so must
restrain the conduct and reduce the valuation of competitors. This
last-resort interpretation implies that a continuing danger of preda-

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

273

tory pricing makes equation (1) an inappropriate benchmark because

the actual values of every competitor are always lower. Instead they

will range from an upper bound of V, downward to the value of each


firm's assets in their next-best use, depending on the perceived proba-

bility of recurring price warfare. Now the only suitable replacement


for (1) is what will be called the "competitive" valuation level, defined

as the discounted and risk-adjusted sum of no-growth earnings, or

[(E/lI) - (brSE/II)]. 13 It assumes that the trust does not restrict output,
so ai = R, eliminates ex ante growth opportunities and makes expansion unprofitable for any competitor. Then predatory pricing is more
than just a last-resort strategy to the extent that it lowers acquisition

costs below [(E1/I) - (brSE/I)]-and toward the level of distress asset


values emphasized in historical accounts.
To summarize, there is reason to believe that previous studies have

underestimated the costs of a peaceful merger strategy as well as the


power of predatory price cutting to reduce the acquisition costs of an
aspiring monopolist. The market values of competitors should vary
with the trust's conduct and reputation in the industry. But then no

single valuation formula is valid ex ante to measure any savings in


acquisition costs resulting from actual or threatened predation. With

the data available for tobacco competitors, the Litzenberger-Rao

model yields the following pair of nonpredatory valuation benchmarks: the low or "competitive" level, where each firm is worth only

the risk-adjusted present value of no-growth expected profits, and the


higher "oligopoly" value in (1), where the competitor follows the price
leadership of a dominant firm. Both the oligopoly and competitive

benchmarks are applied in Sections III-IV to test the influential


views in the literature that predatory pricing always is irrational or

rational only under extreme conditions, respectively.

III. The Regression Model

Valuation data are limited to P0t, the actual expenditure by the American Tobacco Company for the ith competitor acquired in year t," and
13 This specification is something of a compromise. No data exist for other mergers
in the tobacco industry, either prior to the formation of the trust in 1890 or shortly
after its dissolution in 1911, that could be used to develop another valuation benchmark.

14 The 43 tobacco acquisitions were not spread out uniformly over the 1891-1906
period. No competitors were purchased during some of these years, and occasionally
the tobacco trust bought two or more firms simultaneously (see fig. 1). Since the sample

period is discontinuous, the observed expenditure for each firm should be written as

Pi,,, where ti identifies the year the trust acquired the ith competitor. The same notation

should be used on the other time-subscripted variables defined in this section. How-

ever, ti is so cumbersome that it is omitted in favor of a t subscript that means "the year
the ith competitor was acquired."

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

274

JOURNAL

OF

POLITICAL

ECONOMY

previous discussion indicates that it was probably less than V, for every
firm. This section develops two regression equations for Pit from the
oligopoly and competitive valuation benchmarks. The analysis begins
with the former by converting equation (1) to a time series and simultaneously redefining its growth component with measurable vari-

ables. The result is a formula for V,, that describes the peaceful valuation of each competitor absorbed in year t. Then the trust's acquisition

cost Pit is derived from Vjt by modeling reputation effects and any
direct savings attributable to alleged predatory pricing. Finally, these
steps are easily replicated for a competitive valuation benchmark with
the assumption that no firm has any growth prospects.

A. The Time-Series Adaptation of (1)


The revised specification of the growth component in (1) is

TB.[ R+ 1 = TB11, +I R, 1 (2)


LR(I1+RI)J Rt(l + Rt) 1
with

B,, = KQ,,tI (K>0), (3)


where NI,, = essentially, the trust's average rate of return on invest-

ment in years (t + I)-(t + 5); R, = [(D,/C,) + Gj] for the Cowles index
of common stock prices (industrials), with D, = the prorated dividends on the index in year t, C, = the corresponding value of the
index, and G. = the average annual growth rate of C, between (t - 7)
and t.

In (2), the variables Mi, and R, replace the competitor's gross return
on investment (,r,-) and cost of capital (Ri), respectively. These proxies,
the average return on investment for the American Tobacco Company (MI,-) and the marketwide cost of capital (R,), are primarily or
solely a function of time because firm-specific measurement is impossible. The trust's returns are the only available data on the
profitability of investment in the tobacco industry at the turn of the

century. Note that (2) assumes ex post values of MNI, could have been
forecasted accurately in the (t + 1)-(t + 5) period. Measuring the cost
of capital as the average dividend yield (D,/C,) plus the rate of stock

price appreciation (G,) is a common practice in finance (Weston and


Brigham 1981, p. 601). However, no data exist for either future investment (B,,) or the horizon period (T). Equation (3) makes Bi, proportional to the preacquisition output (Ql {_ <) of each competitor because sales are considered a reliable indicator of commercial success
and the ability to finance expansion (Baumol 1967, pp. 45-46). Then
K and T will be compounded in the coefficient estimate of the growth

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

275

component. The Appendix presents further details on the measurement of M.t, Rt, and 1

Substituting (2)-(3) into (1) gives

Vi, = o ( I ) + I + P( -SEI)l + P33,"t-'ljt -R1 (4)


where PI = 1, 12 = - br < 0, and P3 = KT > 0. Although (1) has no
intercept, the preferred econometric practice (see Meeker et al. 1983,

p. 179) is to include a P3o in (4). This time-series adaptation of the


Litzenberger-Rao model conforms with the limited data on tobacco
competitors while continuing to describe their market values under
entirely peaceful conditions. That leaves the remaining adjustments

to transform Vdt into observed purchase prices.


B. The Likely Effects of a Predatory Reputation
Even when no one fears industrywide price cutting, the trust might be
able nonetheless to reduce the amount paid for an individual competitor by threatening it alone with a predatory attack. The credibility of

any such threat presumably depends on the trust's reputation for

misconduct. This variable is measured as N, l/7 for the American


Tobacco Company, where N, l = the number of prior episodes of
alleged predation not involving the ith competitor. The sample contains seven episodes of alleged pricing below cost, separated by time

and geographic or product markets. Hence N, rises during the


1891-1906 period, to a maximum of 6/7, as these events accumulate.
Any effects of American Tobacco's reputation on the amounts paid

for competitors probably increased as well because the apparently

frequent resort to predatory pricing in the recent past should have

made the threat of another attack more believable.'


The simplest way to model these potential cost savings is to assume

that they arise, if at all, from general intimidation. Suppose a credible


threat of price warfare is first communicated, or perceived, only during the final merger negotiations. Then it could not affect the behav-

ior of the competitor prior to acquisition. But the threat may so intimidate the firm that it sells out for some fraction of Vj, in (4), and less

Reputation effects also might vary inversely with the time elapsed since the last
predatory attack. Arguably, most firms would remember recent warfare but tend to
forget episodes that concluded several years earlier. The time between predatory cam-

paigns cannot be measured accurately in the tobacco industry because their beginning
and termination dates are largely unknown. However, the sample period contains at
least one episode of alleged pricing below cost every year after 1894, so the omission of
this variable should not affect the regression results.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

276

JOURNAL

OF

POLITICAL

ECONOMY

than its intrinsic market value, according to

V* = V + 1( ' ) + 2 SE, )
(5)

? 13Qit-L[R(1 + Rt) x (1 4 7 )
Equation (5) describes quasi-peaceful valuation: the total market
value of the ith competitor in year t reduced, perhaps, by a specific

threat of predatory pricing made increasingly credible by the trust's

enhanced reputation for misconduct. In (5), h13N- 1/7 is the average


percentage discount (if any) from Vi, that the tobacco trust obtained
through intimidation. Its magnitude depends on 34, and this parameter can be estimated directly. The hypotheses are

Ho: 4 = 0 A mere reputation for


predatory pricing does not

reduce the acquisition cost of (6)


competitors.

H1: 3 > 0 A predatory reputation lowers


acquisition costs.

If 4 = 0, then f34Nt 1/7 = 0 in (5) such that notoriety yields no


observed savings. This result would imply either that an aggressive
reputation could not be established by attacking other firms or that it
diminishes quite rapidly. On the other hand, predatory price cutting
frightens competitors into selling out cheaply-thereby lessening the
trust's expenditures for subsequent acquisitions-to the extent that

V*j < V-, because 4 > 0.

C. The Derivation of Observed Acquisition Cost

Classical accounts of predatory pricing imply that additional reductions in the amounts paid for competitors accrued from attacking
their businesses prior to acquisition. As with reputation in (5), the
most plausible specification of direct warfare is to assume that it gen-

erates savings, if at all, as a constant percentage of V%. Then the trust's


actual expenditure for the ith firm acquired in year t becomes

Pit = V*t X (1 - P5PPCJ) + Eit, (7)

where PPC, = 1 if the ith competitor was an alleged victim of predatory price cutting up until its acquisition, and 0 otherwise, and fit -

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

277

N(O c2) denotes random error. Substituting (5) into (7) gives

Pit= Po + It ) + 2 lSE )

R)

11

(8)

X (I - 4 N7 I X (1 - I 35PPCI) + Eit.
Equation (8) was derived from an oligopoly valuation benchmark
where the trust acts as a partial monopolist and creates strong incentives for the expansion of rivals. The corresponding equation based

on the competitive valuation level, where there is no output restric-

tion, is derived by setting (,e - R-) = 0 in (2), eliminating the growth


component from (5), and substituting the remainder in (7):

Pit = ?o j ) + (2( ,)
(9)

x (1 _ 4 N1 ) x (1 - 05PPCI) + Eit.
Estimates of (8)-(9) are presented in Section IV.

The hypotheses for the coefficient of PPCQ are


Ho: I5 = 0 Predatory price cutting does not
reduce the acquisition cost of

targeted competitors beyond

reputation effects (if any). (10)

H1: 35 > 0 Predatory pricing directly


reduces acquisition costs.

Estimates consistent with r5 = 0 in (8)-(9) mean that the American


Tobacco Company was unable to reduce the amounts paid for competitors by allegedly attacking them prior to acquisition. Instead these
purported victims typically cost the trust as much as comparable firms

that apparently were absorbed peacefully. This finding would reinforce the relatively extreme view in the literature that preying on
rivals is always irrational (McGee 1958, p. 168; Easterbrook 1981, pp.

267, 318), especially if N = 0 in (6) as well. These results would also

support Bork's (1978, p. 154) policy recommendation to deny relief in


private antitrust suits claiming injury from predatory pricing. On the
other hand, 35 > 0 supplies a defensible empirical basis, now long

overdue, for proposed rules of law against pricing below cost (see the

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

278

JOURNAL

OF

POLITICAL

ECONOMY

references in n. 3). These estimates mean that predatory price cutting


may intelligently augment the sequential purchase of competitors because the direct savings on payments to victims could more than offset
the trust's warfare expenses. Hence they would tend to validate his-

torical reports of such misconduct in the tobacco industry. In (8), 35 >


0 lowers acquisition costs below oligopoly valuation levels, and these
savings, if large enough, imply that predation can be a rational monopolization strategy when the trust also restricts output as much as
possible. The same result in (9) suggests that acquisition costs were
reduced even below competitive valuation and toward the level of
distress asset values. It would imply, contrary to Scherer (1980, pp.
336-37), that at least the old American Tobacco Company used predatory price cutting more often than just as a last resort.
To permit a full test of these hypotheses, equations (8)-(9) set PPC
= 1 for every competitor in the sample victimized by predatory pricing according to U.S. Bureau of Corporations (1909-15) and U.S.
Supreme Court (1911), the primary research sources on the tobacco
trust. In general, the allegations are sketchy, vague, and clearly less
than compelling a priori. The following accounts are typical (see also
U.S. Supreme Court 191 1, original petition, 1:72-73):
Plug.... the Rucker & Witten Company were [sic] doing a
very prosperous business until the managers of the American Tobacco Company made arrangements with D. H.
Spencer & Sons . .. [under secret trust control] to carry on a
destructive competitive warfare against the Rucker & Witten
brands. Finding it impossible to maintain itself against this
competition, the Rucker & Witten Tobacco Company sold its
entire business to the R. J. Reynolds Tobacco Company
[then a trust subsidiary] in 1905. [U.S. Bureau of Corporations 1909, p. 139]

Fine Cut. During this period [1899-1901], the Combination pursued a policy of active competition against the independent concerns in the fine cut business, which at that time
had substantially one-half of the total output of the country.
The Combination succeeded in increasing its output materi-

ally, but met with considerable losses. Following this period


of aggressive competition the leading competing concerns,
Spaulding & Merrick and D. H. McAlpin & Co., were acquired. [U.S. Bureau of Corporations 1915, p. 130]

The sample contains 17 alleged victims of predatory pricing, including the three firms described above. The incidence of these attacks is
depicted in figure 1, which plots the time series of amounts paid (in

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

279

$
13 -

12

*-Alleged

Victim

of

Predatory Pricing
o-Peaceful Acquisition
10 _
(0

w 8 -

L L
FX
7 n~~~~~~

W
U)65

t 4_
a.

0~~~~~~~~

00

00

oomommomQmQooIQ0oo0Q00000000OOO0

mm zzMF__(1_> z~mZjimttmcuMs->umz a. zum>

W Q. <CL-UU0< W< (oDn Ww< a. WW Ow wlDwuuwQL.O


IL < oUnozDE<nU.)0<OLL ozQV)noo0no<Z
"ACTIVE

PLUG COMPETITION"
WAR

IN

THE

FINE CUT BRANCH

FIG. 1.-The time series of tobacco acquisitions; 43 competitors, 1891-1906

constant dollars) for each competitor. Also shown are the approximate periods of two important episodes of alleged predation: the
infamous "plug war" of 1895-99 and the "active competition" in the

fine cut branch (1899-1901).16 Evidently, the American Tobacco


Company attempted to buy its largest and most expensive rivals first
and preyed on them if necessary, while usually acquiring smaller
firms only after it was well established in each branch of the industry.

IV. Empirical Results

Data for the 43 tobacco competitors are summarized in table 1, and a


correlation matrix is presented in table 2. The sample contains a wide

range of firm sizes: the smallest acquisition, in 1900, was the Brown
Brothers Company (Winston, N.C.), which cost less than $63,000 in
16 The plug war is recounted in Burns (1982, pp. 37-41).

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

r~~~~~ t- CoK - in C

t- N- N n G4 X O

>

ts

Or . X 0 n C- z _

OU

>t

C
U

C~~~~~~~C4o

)0 00 CM ON - 0-0 O

H~~~~~~~~~

C~~~~~~~~~~~C-

-z

280

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

ZC'

'-0

ot O
H

-I

Xr C 1 I X if ,CL

>0

_l

0.

U Q0
?~~~~~~~ CZ 0~~~~~b
X ~ ~ ~ r z - > G --i

-; v
o ~~~~8

V8

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

282

JOURNAL

OF

POLITICAL

ECONOMY

real terms; but the trust paid $12,500,000 in 1899 for the original
Liggett & Myers Tobacco Company (St. Louis, Mo.), its leading rival
in the plug branch. Table 2 provides some indication of when predatory pricing can be expected, assuming (for the moment) that the

allegations in U.S. Bureau of Corporations (1909-15) and U.S. Supreme Court (1911) are true. Other things being equal, American

Tobacco tended to attack large competitors (rppc.Q = .440) with attractive growth prospects ex ante (rppc.Growth = .369). These firms
presumably became targets for price warfare because their continued
independence jeopardized the establishment or maintenance of effective market control. Moreover, they should have been the firms most
likely to insist initially on premiums above competitive valuation for
selling out (McGee 1958, p. 141). Hence predatory pricing may have
promised substantial savings in the costs of their acquisition-in addi-

tion to accelerating the consolidation of the plug, smoking, snuff, and


fine cut branches of the industry.
Equations (8)-(9) were estimated with TSP (Hall 1983) using the

two-stage procedure of Amemiya (1974) for nonlinear equations con-

taining instrumental variables. Instruments were created for (ElII,)


and (SE/II) by the grouping method of Wald (1940) to cope with
errors in the measurement of earnings and risk. The resulting estimators are consistent but not asymptotically efficient. In addition,
the standard errors of the regression coefficients were computed with
a correction for heteroscedasticity, applying the formulas of Cham-

berlain (1982) and White (1982), to permit unbiased significance tests.

Estimates of (8)-(9) with and without the intercept 13( are presented
in table 3. Also shown are the expected signs of every coefficient in
brackets. The Durbin-Watson statistics and estimates of the Markov

process in the residuals are not meaningful here and have been
omitted (see n. 14).
Equations (8)-(9) give a strong fit to the expenditures by the to-

bacco trust for competitors. The adjusted R2's and F-statistics in table
3 are quite high, probably because of the wide range of firm sizes in
the sample. The coefficient estimates for discounted earnings, risk,
and growth opportunities have the theoretically correct signs, and, in
general, they are statistically significant at or above the .10 level.
These estimates also appear invariant with respect to the alternative

valuation benchmarks and the inclusion or exclusion of an intercept

in (8)-(9). Similar results were reported by Litzenberger and Rao


(1971, pp. 272-73) and Meeker et al. (1983, pp. 184-85) for 87

electric utilities and 357 small banks, respectively, during various


years of the period 1960-75. Thus table 3 further documents the
power of the Litzenberger-Rao model to describe the observed valuation of diverse business enterprises.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

~~~~~~- 4- c *i-*
- 6 ] - 0

C
U~~~ -

r-

Xl---

~bC

ic~ ~~~~, 00 i_ *- n ~
_ s;c + t0 b *

H > > .S ) t c- S X

_ 0 v fN *

<n

-sO

<

0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0

IAX

>t

J~~~~

0~~~~~ _~
0~~~~~~~~~

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

284

JOURNAL

OF

POLITICAL

ECONOMY

The coefficient estimates for the trust's reputation and direct price
warfare are of more immediate interest. The former are uniformly
positive and significant at or above the .05 level with both oligopoly
and competitive valuation. Likewise, the I5's are positive in three of
the four equations, and one of them becomes significant at the .05
level when (8) includes an intercept. In addition, an F-test rejects the

joint hypothesis that 04 = A5 = 0 at or above the .05 level throughout


table 3. As discussed in more detail below, these results support the

alternative hypotheses in (6) and (10)-that predatory price cutting


reduces acquisition costs both directly and indirectly through reputa-

tion effects. However, a tendency for the predatory reputation of the

American Tobacco Company to yield savings on its purchases of competitors is documented more convincingly in table 3 than savings
from actual price warfare.

The robustness of these estimates was tested with several alternative

forms of (8)-(9) as follows: specifying the market risk premium (b) as


a time series; adding the Cowles index (Ca) and Time = (t - 1890)
as explanatory variables in (5) and (7)-(9); reducing the sample to 41

competitors by dropping the largest and smallest acquisitions; adding

another dummy variable, in the same manner as PPC, in (7), when


American Tobacco did not purchase the fixed assets of a competi-

tor; 8 constraining the coefficient of discounted earnings to equal


unity; 19 restricting the trust's reputation to a chilling effect on growth
opportunities rather than allowing it to reduce all valuation compo-

17 This parameter was not necessarily constant from 1891 through 1906, as assumed
in Sec. II. Empirical estimates of the risk premium vary from year to year (Litzenberger
and Rao 1971, p. 273), evidently as a function of population and national wealth
(Litzenberger and Budd 1972, pp. 861-62). None has been computed for the sample
period, and no reliable data exist for wealth or its obvious proxy, gross national product

(U.S. Department of Commerce 1975, 1:216, 219). Consequently, the analysis assumed
that

eYime

1)

where Time - 1890. The parameter y measures the average percentage change in

b, over time, but its sign is indeterminate a priori. When (11) was included in (8)-(9),

the estimates of My were indistinguishable from zero at an acceptable confidence level.


Hence b may be treated as an intertemporal constant in (8)-(9).
18 There are five acquisitions in the sample where American Tobacco did not buy a
competitor's factory and real estate along with its manufacturing business. The usual
reason was that the firm had leased these assets and could not convey them to the trust
(see U.S. Supreme Court 1911, testimony of Caleb C. Dula, 2:480; testimony of George
W. Coan, 3:649; testimony of James B. Duke, 4:348).
1' Equations (1) and (4) require f3, = 1, but 3It < 1 at the .01 level throughout table 3.
This result suggests that all the estimates of (8)-(9) are something less than ideal.
However, constraining 131 to equal unity increases the coefficient estimates for both the
trust's reputation and predatory pricing as shown in table 4.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

285

nents as in (5);20 and reestimating (8)-(9) with a cost of capital mea-

sure developed from the capital asset pricing model (CAPM).21 These
changes did not substantially alter the pattern of coefficient estimates

in table 3. Some examples are presented in table 4, where the most


noteworthy result is that the estimated coefficients for direct price
warfare become statistically significant in three alternative forms of
oligopoly valuation.

Finally, table 5 presents estimates from (8)-(9) when the sample is


divided into groups of 25 plug competitors and 18 other firms,

classified by primary outputs.22 Some variation in the amounts paid by


20 A credible threat of price warfare might simply curtail the targeted firm's expansion by discouraging investment, even though its expected return exceeds the cost of
capital in the absence of predation. Milgrom and Roberts (1982) show that a reputation

for below-cost pricing can retard entry when knowledge of the trust's response is
imperfect; presumably the same deterrence could occur in the closely analogous case of
expansion by established competitors. Then a threatened attack reduces the market
value of the firm in (4) as the growth component (2) approaches zero. This chilling
effect is modeled by replacing (8) with

P. = {o + pi( E ) + 2 SE )

33Q~t~i(1( N,7 ) R,(1 + R,)jJ (12)

x (1 - f35PPCi) + Eit.

The variable [1 - (N,_1/7)]O4 measures the average probability that each competitor
would have undertaken the investment required to realize ex ante growth opportuni-

ties. It declines over the sample period (if a chilling effect is present) at an increasing,

constant, or decreasing rate as 04 1, and otherwise has the interpretation in (6). The

distinction between the chilled growth opportunities of (12) and the general intimidation of (8) rests on whether threatened predation induces the targeted firm to compete
with restraint prior to acquisition. If so, it would still be paid V,, in (4), other things
being equal, but this amount has been reduced by the discounted profits from the
growth opportunities forgone. In contrast, general intimidation in (5) and (8)-(9)

lowers the trust's expenditure below Vi, by frightening the competitor into accepting
less than its intrinsic market value, as noted in Sec. III.
21 The alternative cost of capital measure is

Ri, = I, + Premiumi x Betalut X 0.083, (13)


where 0.083 is the Ibbotson and Sinquefield (1982, p. 71, exhibit 29) estimate of the

equity risk premium for the 1926-81 period, Betar-,st is taken from the largely contemporaneous estimates in Burns (1983, p. 353, table 6), and
1.2 if Qi,, l is among the 15 largest in the sample

Premiumi- 1.4 if Qi,,- , is among the 14 smallest in the sample


1.3 otherwise.

The major limitation of (13) is that, although the risk of the ith competitor should
increase as the firm becomes smaller, the factors used in Premium, are nonetheless
arbitrary. Note that this alternative measure has no impact on the estimated coefficients
from (9) with its competitive valuation benchmark.

22 Virtually identical results were obtained with the cost of capital measure from n.
21.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

~~~~~0~ ~ ~ 0
_
,

t
I0
c t~~~~~

rIt

<z

t~

00 K 2~~~~~~~~~~~~~~~~~~~0a

i.n

,:

OC

,,.-

,24

<t

= s ~~~~~* * *
~br

04X;

<

>G

1:

C)~~~~~~~~~~~~~~~~I

:E; o 5 -C s6,

00
It

bz

<

04- 00 00 000
Sc

>

o~

001

~C 0o 22 tG <

000

S **~~~~~~~~~~
U)~~~~~0

C) =
H

C)

C)~ 28

N2

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

ct U + bJ' X A- O 5 Qn??
Cv4

~~~

OC

CJ

00

.~~ ~ ~ ~ ~ . .~ - ,. -, . P

}^ t l-~~~~~~~~~c t

X)

Xi

cq

>

- > 's
.D

Xi

- <
. 'S .

CM
.

O
. GS~~~~~~~~Z
.

*~~~~~~~~~~~~~~~~~~~~~~~~~ ~ >

cq
.

.~~~~~~~~~~~~~~~~~~~~~~~

- ~ ~ ~ ~ ~ ~ ~ ~ ~~ ~~~~~~~~~~~~~~~~~h

*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~C

*C

*n

*8*

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

V> ,I cl ,I!o
t- ts~~~~~C oo < O

in X It ) Gnfo 0C) cqt- 00n


s~~~~~~~1 C n G C, Ot oo 00 C >K
v~~~~~~~~\

C1

cn

G.

CC

~~~~~~~~~C4 Z e) x 001 e <s o*


Z X ;.# 4 | GM GM GA G~~~~n C : t

tw

C,4

G.t00

e)sCAO1

v~~~~~~~~~~~c~

C~~~~~~~~~~~~~~~~~~~~~~O * * *

cr*

v: - r - *^ rt r^,* * -

: ~~~ ~ ~~~~~~ C, i n Cf ol 01 o.- :

p~~~~~

C>

in

Q0

D.

Cs

in

z > a; ,_ t n t ~~~~-G 00 cn U- > > G C

*M ' I O ~cq, - l X 0 GMX * -_ -

5C

oi

Ot O e_ > oX cn G X z 00 ol O t.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

289

American Tobacco might be expected because the plug manufacturers typically were larger and more expensive and were acquired earlier in the 1891-1906 period than competitors that produced fine cut,

smoking tobacco, and snuff. Unfortunately, the partitioning also reduces degrees of freedom and makes the estimation less precise, espe-

cially for the risk and growth components. Otherwise the estimates
from the plug acquisitions in panel A are similar to those in table 3

with statistically significant coefficients for discounted earnings and


the trust's reputation yet no discernible impact for the alleged warfare

itself. In particular, the long and costly plug war evidently failed to
yield any direct savings from the 10 victims purchased at its conclu-

sion. The general pattern continues in panel B with the expenditures

for 18 other firms, but, in addition, the estimated coefficients of predatory pricing are consistently significant at the .01 level under both

valuation benchmarks. They represent the strongest evidence in the


tobacco data that preying on competitors directly reduces their cost of

acquisition.
The results in tables 3-5 tend to substantiate classical accounts of

below-cost pricing by the old American Tobacco Company. They


indicate that predatory price cutting directly reduced its expenditures
for some alleged victims and, more certainly, created a notorious
reputation that intimidated other competitors into selling out

cheaply. Since the values of 04 and 5 are essentially invariant with


respect to the oligopoly and competitive valuation benchmarks, it

appears that the trust repeatedly preyed on rivals both to offset the

escalation of acquisition costs from peaceful output restriction and to


lower those costs toward distress asset values. The estimated savings
are quite large as well. The average discount from reputation effects

alone is 25 percent in table 3;23 and an additional discount averaging


56 percent accrued from preying on the relatively smaller fine cut,

snuff, and smoking tobacco firms in table 5, panel B. These rates


signify reductions in acquisition costs that are much greater than the

trust's operating losses during at least one of the price wars in the
sample: American Tobacco lost approximately $200,000 in the "active competition" for control of the fine cut branch that was described

briefly in Section III; but this campaign saved an estimated $1.07


million in the subsequent purchase of the McAlpin and Spaulding &

Merrick companies.24 The corresponding net savings of almost


23 This rate was computed by multiplying the average of the I4'S in table 3 (1.00) and
the sample mean of N,- 1/7 (0.249).
24 The savings were computed from table 5, panel B, using average values of 4 =

0.919, 5 = 0.562, and N,- 1/7 = 0.286 from two prior episodes of alleged pricing

below cost. The estimate of the trust's operating losses was calculated from U.S. Bureau
of Corporations (1915, p. 129, table 44).

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

290 JOURNAL OF POLITICAL ECONOMY

$900,000 likewise exceed the estimated prewar earnings of these


firms by about 2: 1. Hence the trust may have been fully compensated
for its direct losses and forgone monopoly profits from the fine cut
contest, especially when the enhancement of its aggressive reputation
promised to yield additional savings in the future. Thus, now that a
rational basis for such misconduct has been documented, there are
good reasons for believing that the numerous allegations of predatory
pricing by the tobacco trust were fundamentally true.
These findings raise two new questions that merit brief considera-

tion. First, if predatory price cutting actually produced such large


savings in merger expenditures, why didn't the American Tobacco
Company prey on every firm in the sample? A sensible answer seems
to be that the estimated discounts are known only ex post, and there
may have been much uncertainty ex ante to discourage indiscriminate
price warfare (Easterbrook 1981, p. 286). Second, is there any nonstatistical evidence of below-cost pricing to reinforce the inferences
from tables 3-5? This question recognizes that the estimated savings
attributed to predation are also consistent merely with intensified, but
lawful, price competition. In other words, tables 3-5 could be interpreted narrowly as confirming only that increased rivalry lowers the
market values of all competitors. However, there is other evidence
that clearly supports a predatory pricing interpretation of the regression results. It consists of remarkably candid expressions of predatory
intent in several business letters exchanged between trust officers.
These documents became the foundation of the government's Sher-

man Act case against American Tobacco, but they are too voluminous
to be fairly reviewed here.25 Some choice examples are presented in
U.S. Supreme Court (1911, testimony of Caleb C. Dula, 2:536-40).
V. Summary and Conclusions

This paper has presented a new analysis of pricing below cost as a

monopolization strategy that applies the Litzenberger-Rao (1971)


valuation model to study the amounts paid for 43 competitors pur-

chased by the old American Tobacco Company between 1891 and


1906. The regression results identify whether the tobacco trust reduced its merger expenditures by allegedly attacking some of these
firms prior to their acquisition. While subject to measurement errors
and other qualifications, the coefficient estimates are consistent with
the classical view that predatory price cutting lowers acquisition costs
directly as well as indirectly through reputation effects on rivals that
are subsequently bought out peacefully. These results cannot prove
2- A paper analyzing these documents is currently in preparation.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

291

that each episode of predation was profitable because the data necessary for comparing estimated merger savings with all warfare expenses are unavailable. But they clearly point in that direction and
thereby supply a rational basis for the reported misconduct of the
tobacco trust.

The analysis and conclusions differ sharply from the pioneering


research of McGee (1958), and they suggest it is time to reevaluate the
prevailing views that predation is always, or almost always, an
inefficient and irrational business practice. The techniques applied
here show considerable promise for generating hard facts about the
frequency and consequences of below-cost pricing by the great trusts
at the turn of the century. The ongoing debate in the law reviews over
the Areeda and Turner (1975) rule illustrates how little real knowledge is produced by weak theoretical discourses, hypothetical exam-

ples, and seemingly endless rhetoric. Instead more and better facts
are needed to develop an intelligent policy for dealing with predatory
price cutting.

Appendix
The Data Sample

The 43-firm sample is the largest that can be assembled from the plug, smoking, snuff, and fine cut branches of the tobacco industry at the turn of the

century. As noted, it contains practically all of the leading domestic competitors merged into the American Tobacco Company along with many smaller
acquisitions. They were classified as follows on the basis of primary outputs:
plug, 25; smoking tobacco, 13; snuff, 4; and fine cut, 1. Multiple outputs were
common, with 20 competitors operating in two branches and five selling three
or more distinct products. Evidently, none of the firms was failing prior to
acquisition.
Information about these competitors is rather limited and comes almost
exclusively from U.S. Bureau of Corporations (1909-15) and U.S. Supreme
Court (1911). The following data exist for each firm: the purchase price, the
merger date, whether the trust obtained the factory and real estate along with
the competitor's business, the preacquisition output(s) of the firm, and
whether it was an alleged victim of predatory pricing. Annual earnings were
reported by only 15 competitors and vary from one to eight observations per
firm in the years immediately preceding acquisition. Fragmentary estimates
of unit profit margins for independent manufacturers are also available between 1890 and 1906. Finally, remarkably complete data exist during the

sample period for the trust's annual output and profit margins in the plug,
smoking, snuff, and fine cut branches; its annual rates of return on investment; and the monthly prices on the New York Stock Exchange (NYSE) of its
outstanding securities (Commercial & Financial Chronicle 1891-1904). There is
no alternative to measuring the financial variables in equations (8)-(9) with
these sets of data. However, it cannot be done in a completely satisfactory

manner for earnings (Ei) and risk (SE,). The general measurement proce-

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

292

JOURNAL

OF

POLITICAL

ECONOMY

dures are described below, but they were altered occasionally to improve the
specification of one or more variables for a particular competitor.
Acquisition Cost (Pit)

These data are listed in U.S. Bureau of Corporations (1909, pp. 179-93).
American Tobacco paid cash for 31 firms in the sample, issued securities to
four others, and gave cash as well as stock to eight competitors. For cash
purchases, Pit is simply the amount paid. When the competitor received securities, Pt was calculated by multiplying the corresponding number of shares
and the average market price(s) of the stock(s), plus any cash payment. Where
possible, these prices were averaged over the 7 months bracketing the merger
date (month 4) both to reflect information that may have influenced the
number of shares paid and to measure the cash value of the business acquired. For acquisitions financed with new trust securities not yet listed on the
NYSE, the average prices were computed during the first 3 months of actual
trading in these shares after the merger date. All averaging was limited to
lagging periods of 90 days to minimize the effects of stock price changes from
subsequent acquisitions, which no competitor could reliably anticipate. Two
other accommodations were necessary. The trust obtained only a controlling
interest in three firms and not full ownership as with the remainder of the

sample; in such cases, Pi, was measured as the amount paid divided by the

stock fraction actually purchased. And four competitors were purchased


from syndicates of financiers who supplied the trust with additional cash and

retained some of the shares paid as a commission; the P-,'s of these firms were

calculated from their net receipts as reported in U.S. Bureau of Corporations


(191 1, pp. 102-8). All purchase prices were deflated by the Bureau of Labor
Statistics wholesale price index (WPI) resealed to 1899 = 100 (U.S. Department of Commerce 1975, ser. E40, 1:200), except when (8)-(9) experimented

with the Cowles index (Ct) as an additional explanatory variable.


Discount Rate (It)

The pure interest rate was defined analogously to Meeker et al. (1983, p. 180)
as the yield to maturity on long-term U.S. government bonds traded continuously on the NYSE during the sample period. The averaging interval for the
bond prices was always the 7 months surrounding the merger date. The data
were taken from the Commercial & Financial Chronicle (1891-1907).
Expected Earnings (Ei) and Nondiversiflable Risk (SEi)

The major problem with these variables is developing measures from limited
data that largely eliminate the effects of alleged predation on current income.
Otherwise Ei will be unduly low and SE, artificially high for asserted victims of
predatory attacks by the tobacco trust. Following Litzenberger and Rao (1971,
p. 271) and Meeker et al. (1983, p. 179), the cleansing process relies primarily
on averaging the known or estimated earnings of each competitor over the
years preceding its acquisition.
Where feasible, Ei was computed as the arithmetic mean of reported earnings after deflating all values by the WPI and then excluding any figure for a
targeted competitor that appeared extremely low. The maximum averaging
interval was (t - 5)-(t - 1). Shorter periods in advance of the merger year t

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

293

were used as the earnings data became less complete. For five competitors,
including three alleged victims of predation, E was fixed as the earnings
reported only in the year immediately preceding their acquisition.
Having exhausted the available income data, the premerger earnings for

the remaining 28 firms were estimated from

EZT A pmjTQVT3, T= t - 5, t 4, . t -1, (A 1)


i
such that expected earnings becomes

Ei-

=>

r=t-5

5,

(A2)

where QT =the output of the ith competitor sold in the jth product branch

(plug j = 1], smoking [j = 2], snuff [j = 3], or fine cut [i = 4]) during a year
prior to acquisition, and pmjT = the corresponding average profit margin per
unit of output and deflated by the WPI. The average profit margins come
from detailed records of the trust's sales in each branch and less complete

surveys of independent manufacturers, both reported in U.S. Bureau of


Corporations (1915). The latter do not exist before 1900 outside the plug

branch. For competitors purchased earlier, equation (Al) typically applies the
trust margins, which should not reflect monopoly pricing because its market

shares were usually less than the 40 percent considered necessary for effective
control (Scherer 1980, p. 232). The profit margins for acquisitions after 1900
were computed by averaging the trust and independent margins in each
branch. This procedure strikes a balance between monopoly pricing, since the
independent margins were lower, and the trust's policy of buying up primarily its most successful competitors (see U.S. Bureau of Corporations 1909, p.
13; U.S. Supreme Court 1911, testimony of William R. Harris, 3:448). The
output data are scattered throughout U.S. Bureau of Corporations (1909).
They are often limited to one set of observations per firm, usually for the year

preceding its acquisition; in such cases, Qij, was fixed at Qijt- I for allJ in (Al)

so that the time-series variation in E2T originates only from pmJT. The firm's

total output in (3) and (8)-(9) is Q- =

The output data must also be handled carefully for asserted victims of
below-cost pricing with earnings generated by (Al). The reason is that these
firms should have curtailed production and sales during a predatory attack.

Without adjusting their Qij,'s, the average earnings from (A2) will tend to be

inordinately low, thereby reducing the prospect that the trust's reputation

and/or PPC, will show any cost savings in (8)-(9). The adjustments continue
to rely on an averaging of firm outputs in (A 1)-(A2) over the years preceding
each merger. But output data exist only in year (t - 1) for five alleged victims
of predation, where averaging over time was not feasible. Including the three

alleged victims with only 1 year of reported earnings, the retention of these
eight firms in the sample imparts a conservative bias to the coefficient esti-

mates of (8)-(9), making them less likely to exhibit P4, P5 > 0 than if output
and earnings data were known prior to (t - 1).
Nondiversifiable risk was measured as
t-

SE,

ITi

>

1/2

(E?-

E2

(A3)

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

294

JOURNAL

OF

POLITICAL

ECONOMY

for the majority of firms in the sample. Where (A3) could not be calculated

because of insufficient data, then SE, = (Qk t I/Qit- 1)SE , where k denotes a
roughly comparable firm acquired at approximately the same time.

Prospective Return from Growth (Mi,)


These data measure the average return on investment for the American
Tobacco Company and were calculated from U.S. Bureau of Corporations

(191 1, pp. 305, 308, tables 101-2). Separate rates of return are available for
the trust's general operations and, after 1900, its snuff business. Unfortu-

nately, the former includes the investment in and profits from cigarettes,
which none of the 43 competitors manufactured in significant quantities. The

values of Mi, in (2) were computed as the weighted average of these returns

between (t + 1) and (t + 5), with the weights fixed by the relative outputs of

snuff and other products for each firm. This specification assumes that any
growth by the ith competitor would have involved a proportionate increase in

all outputs, and it is preferable to restricting growth opportunities to a branch


of the industry where the firm sold its primary output.

Cost of Capital (Rt)


The source of these data is Cowles (1939, ser. P-2, Ya-2, pp. 68-69, 372-73).
They include the securities of American Tobacco and its affiliated corporations during most of the sample period. Lorie and Hamilton (1973, p. 30)
describe the Cowles index as the "best" of the early stock market indices
because of its comprehensive selection of common stocks and weighting by

relative market values. An alternative measure of R, is described in note 21.


Assessment

The sample could be generated only with some restrictive assumptions that
were a precondition for the regression analysis. Nor is it possible to indepen-

dently verify the margin data in U.S. Bureau of Corporations (1915). Hence
significant measurement errors cannot be ruled out. But great care was taken
to assemble the data properly, and they do yield plausible coefficient estimates in tables 3-5.
References

Amemiya, Takeshi. "The Nonlinear Two-Stage Least-Squares Estimator." J.


Econometrics 2 (July 1974): 105-10.
Areeda, Phillip, and Turner, Donald F. "Predatory Pricing and Related Prac-

tices under Section 2 of the Sherman Act." Harvard Law Rev. 88 (February
1975): 697-733.
Baumol, William J. Business Behavior, Value and Growth. Rev. ed. New York:
Harcourt, Brace and World, 1967.
. "Quasi-Permanence of Price Reductions: A Policy for Prevention of

Predatory Pricing." Yale LawJ. 89 (November 1979): 1-26.

Bork, Robert H. The Antitrust Paradox: A Policy at War with Itself. New York:
Basic, 1978.
Burns, Malcolm R. "Outside Intervention in Monopolistic Price Warfare: The
Case of the 'Plug War' and the Union Tobacco Company." Buts. Hist. Rev. 56
(Spring 1982): 33-53.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

PREDATORY

PRICING

295

"An Empirical Analysis of Stockholder Injury under ?2 of the Sher-

man Act."J. Indus. Econ. 31 (June 1983): 333-62.


Chamberlain, Gary. "Multivariate Regression Models for Panel Data." J.
Econometrics 18 (January 1982): 5-46.

Commercial & Financial Chronicle. New York: Dana, 1891-1907.


Cowles, Alfred. Common-Stock Indexes. Bloomington, Ind.: Principia (for
Cowles Comm. Res. Econ.), 1939.
Easterbrook, Frank H. "Predatory Strategies and Counterstrategies." Univ.

Chicago Law Rev. 48 (Spring 1981): 263-337.


Elzinga, Kenneth G. "Predatory Pricing: The Case of the Gunpowder Trust."
J. Law and Econ. 13 (April 1970): 223-40.

Hall, Bronwyn H. Time Series Processor. Version 4.0. Stanford, Calif.: TSP
Internat., 1983.
Ibbotson, Roger G., and Sinquefield, Rex A. Stocks, Bonds, Bills, and Inflation:
The Past and the Future. Charlottesville, Va.: Financial Analysts Res. Found.,
1982.
Isaac, R. Mark, and Smith, Vernon L. "In Search of Predatory Pricing."J.P.E.

93 (April 1985): 320-45.


Joskow, Paul L., and Klevorick, Alvin K. "A Framework for Analyzing Preda-

tory Pricing Policy." Yale LawJ. 89 (December 1979): 213-70.


Kamerschen, David R. "Predatory Pricing, Vertical Integration and Market

Foreclosure: The Case of Ready Mix Concrete in Memphis." Indus. Organization Rev. 2, no. 3 (1974): 143-68.

Keenan, Michael. "Models of Equity Valuation: The Great SERM Bubble."J.


Finance 25 (May 1970): 243-73.
Koller, Roland H. "The Myth of Predatory Pricing: An Empirical Study."
Antitrust Law and Econ. Rev. 4 (Summer 1971): 105-23.
Litzenberger, Robert H., and Budd, A. P. "Secular Trends in Risk Pre-

miums."J. Finance 27 (September 1972): 857-64.


Litzenberger, Robert H., and Rao, Cherukuri U. "Estimates of the Marginal
Rate of Time Preference and Average Risk Aversion of Investors in Elec-

tric Utility Shares: 1960-66." Bell]. Econ. and Management Sci. 2 (Spring
1971): 265-77.
Lorie, James H., and Hamilton, Mary T. The Stock Market: Theories and Evidence. Homewood, Ill.: Irwin, 1973.

McGee, John S. "Predatory Price Cutting: The Standard Oil (N.J.) Case."J.
Law and Econ. 1 (October 1958): 137-69.

"Predatory Pricing Revisited." J. Law and Econ. 23 (October 1980):


289-330.

Mallela, Parthasaradhi, and Nahata, Babu. "Effects of Horizontal Merger on


Price, Profits, and Market Power in a Dominant-Firm Oligopoly." Mimeographed. DeKalb: Northern Illinois Univ., 1984.

Mariger, Randall. "Predatory Price Cutting: The Standard Oil of New Jersey
Case Revisited." Explorations Econ. Hist. 15 (October 1978): 341-67.
Meeker, Larry G.; Joy, O. Maurice; and Cogger, Kenneth 0. "Valuation of

Controlling Shares in Closely Held Banks."J. Banking and Finance 7 (June


1983): 175-88.

Milgrom, Paul, and Roberts, John. "Predation, Reputation, and Entry Deter-

rence."J. Econ. Theory 27 (August 1982): 280-312.

Pechman, Joseph A. Federal Tax Policy. 4th ed. Washington: Brookings Inst.,
1983.

Scherer, Frederic M. Industrial Market Structure and Economic Performance. 2d


ed. Chicago: Rand McNally, 1980.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

96

JOURNAL

OF

POLITICAL

ECONOMY

Stigler, George J. "Notes on the Theory of Duopoly." J.P.E. 48 (August


1940): 521-41.

Tarbell, Ida M. The History of the Standard Oil Company. 2 vols. New York:
McClure, Phillips, 1904.

Telser, Lester G. "Cutthroat Competition and the Long Purse." J. Law and
Econ. 9 (October 1966): 259-77.
U.S. Bureau of Corporations. Report on the Tobacco Industry. 3 vols. Washing-

ton: Government Printing Office, 1909-15.


U.S. Department of Commerce. Historical Statistics of the United States. 2 vols.

Washington: Government Printing Office, 1975.


U.S. Supreme Court. Briefs and Records. United States v. American Tobacco
Company, 221 U.S. 106 (1911).

Wald, Abraham. "The Fitting of Straight Lines if Both Variables Are Subject
to Error." Ann. Math. States. 11 (September 1940): 284-300.
Weston, J. Fred, and Brigham, Eugene F. Managerial Finance. 7th ed. Hinsdale, Ill.: Dryden, 1981.
White, Halbert. "Instrumental Variables Regression with Independent Observations." Econometrica 50 (March 1982): 483-99.
Williamson, Oliver E. "Predatory Pricing: A Strategic and Welfare Analysis."

Yale LawJ. 87 (December 1977): 284-340.

Worcester, Dean A., Jr. "Why 'Dominant Firms' Decline." J.P.E. 65 (August
1957): 338-46.

Yamey, Basil S. "Predatory Price Cutting: Notes and Comments."J. Law and

Econ. 15 (April 1972): 129-42.


Zerbe, Richard. "The American Sugar Refinery Company, 1887-1914: The

Story of a Monopoly."J. Law and Econ. 12 (October 1969): 339-75.

Zerbe, Richard O., Jr., and Cooper, Donald S. "An Empirical and Theoretical
Comparison of Alternative Predation Rules." Texas Law Rev. 61 (December
1982): 655-715.

This content downloaded from 86.164.156.132 on Sun, 24 Jul 2016 18:05:03 UTC
All use subject to http://about.jstor.org/terms

You might also like