Did the Trusts Have Market Power? Evidence from Distilling, 1881-1898

by Werner Troesken, Hoover & Pitt, and Karen Clay, Carnegie Mellon

Full length versions of this paper can be obtained promptly by email. Send requests to troesken+@pitt.edu


1. Introduction

During the late nineteenth century, the American economy was transformed by the emergence of trusts and other large combinations. Although economists and historians have studied the trusts extensively, there is little scholarly consensus as to how these combinations affected consumer prices and welfare. Broadly construed, existing studies fall into one of two camps, market power interpretations and efficiency interpretations. According to market-power interpretations, the trusts were able to influence prices in both the short and long run because they controlled large market shares and because they acted strategically. Several studies, however, raise questions about the efficacy of the trusts’ predatory strategies and suggest the trusts were only able influence prices in the short run because of market entry. According to efficiency interpretations, the trusts reduced operating costs by introducing new technologies and by exploiting scale economies. Scholars disagree as to whether these cost reductions were passed along to consumers.

The controversy over the trusts stems largely from the fact that nearly all existing studies rely on evidence that is subject to multiple interpretations. For example, authors advancing market power interpretations often cite statistics showing the trusts controlled large shares of industry production. Yet the trusts might have controlled large market shares because they adopted the most efficient production and management techniques or because they brought consumers lower prices. Data on market share could also be deceiving indicator of market power if the threat of market entry were significant. Along similar lines, authors advancing efficiency interpretations often cite statistics showing that output expanded and prices fell with the rise of the trusts. Yet, while expanding output and falling prices are consistent with the view the trusts reduced operating costs and enhanced consumer welfare, they are also consistent with alternative explanations and do not rule out the possibility the trusts exercised significant market power. For example, if trust-dominated industries were characterized by declining costs, increases in consumer demand might have also caused output to rise and prices to fall, regardless of the level of market power the trusts possessed.

In this paper, we provide direct evidence on the market power of the Whiskey Trust, which dominated the production of alcoholic spirits from 1887 through 1895. At both the firm and industry level, we have monthly data on sales; output prices; and the prices of the three primary determinants of (marginal) cost in whiskey distilling: corn; malt; and the federal excise tax. We also have data that allows us to control for technical improvements that reduced the cost of distilling. With these data, market power is estimated using methods from the new empirical industrial organization. These methods include: examining how changes in factor prices affected output prices; examining how changes in output affected output prices; and examining how output prices changed during periods of competition, periods of pooling, and periods when the trust operated. In addition, the firm-level data allow us to estimate the demand curve the Whiskey Trust faced, which in itself would be revealing given the paucity of data usually available on the trusts.

The results suggest that various cost-reducing techniques implemented by distillers enhanced consumer welfare, and that the Whiskey Trust generally exercised little or no market power, though it did increase prices significantly for two short periods of time. Entry and the threat of entry prevented the trust from exercising market power for longer periods.

2. Background: History, Technology, and Market Definitions

2.a. The History of the Whiskey Trust

The Whiskey Trust produced a widely consumed commodity–namely, distilled spirits. In 1880, the average American adult consumed 2.4 gallons of spirits annually. And even though Americans consumed more beer (11.1 gallons annually), the majority of absolute alcohol (58 percent) was consumed through spirits because spirits had a much higher alcohol content.

The formation of the trust was preceded by a series of unsuccessful pools. After the pools failed, distillers organized the Distillers and Cattle Feeders’ Trust, better known as the "Whiskey Trust" in May 1887. Modeled after Standard Oil, the Whiskey Trust was bona fide trust so that when a distillery joined the trust it surrendered control of its operations to a board of trustees. Of the eighty-six distilleries that eventually joined the trust, only ten or twelve were kept in operation; the remainder were shut down.

Perhaps surprisingly given its large market share, in January 1895 the trust entered receivership. The primary cause of the demise of the Whiskey Trust was market entry. According to industry observers, as the trust gained market share it also tried to raise prices. This attracted new firms to the market, who in turn, undercut the trust. Although the Whiskey Trust reorganized in August 1895 as the American Spirits Manufacturing Company, it never regained its former market dominance. A government investigation of the distilling industry conducted in 1900 found that "since the new company was organized there has been very little profit in the business," because "outside distilleries that were being built all of the time came into direct competition" with the company.

2.b. Market Definitions

In the paper, we suggest two measures of the trust’s market share, one narrow and one broad. The narrow market definition suggests the Whiskey Trust’s market share peaked at 90 percent in 1893, while a broad definition suggests the trust’s market share never rose above 50 percent and was as low as 20 percent in 1894 and 1895.

2.c. The Technology and Costs of Distilling

There are two key points in the description of the technology of production. First, the technology of manufacturing alcoholic spirits was characterized by constant returns to scale and low fixed costs. These features help explain the frequency and ease with which firms entered the industry. Second, the technology of manufacturing alcoholic spirits was simple. This simplicity enables us to estimate the cost shares of various inputs, which in turn can be used to inform our empirical analysis.

The manufacture of alcoholic spirits entailed grinding corn into meal; soaking the meal in water; adding malt, which converted the corn starch into sugar; and using a small amount of yeast, which initiated fermentation and converted the sugar into alcohol. After fermentation, the corn mash was twice distilled and charcoal filtered to yield alcoholic spirits. As this description implies, the manufacture of spirits employed a fixed-proportions production technology. There was little, if any, substitutability across inputs.

The primary determinants of variable costs were a federal excise tax, the price of corn, and the price of malt. Because corn was central to the production of alcoholic spirits, most distilleries located in the Midwest and there was an especially high concentration of distilleries around Peoria, Illinois. There were few economies of scale in distilling. Once a distillery reached minimum efficient scale (the capacity to supply 5 to 10 percent of the market), there were few, if any, efficiency gains associated with expanding capacity. Fixed costs included labor, the cost of containers, and plant and capital. A distillery with the capacity to supply 10 percent of the market could have been constructed for $150,000 in 1895. In current dollars, this is about the cost of opening a modest restaurant. There was one form of entry that was particularly easy and low cost: the creation of small and illicit (untaxed) stills, which were encouraged by the large federal tax on spirits.

Government witnesses and accounting data suggest that, for trust-affiliated distilleries, the pre-tax cost of producing a gallon spirits (c - t) was between $.08 and $.15 per gallon from 1888 through 1895. From 1878 through 1894, the federal tax on spirits was set at $0.90 per gallon; from 1895 to 1900, it was set at $1.10. As we show in the paper, these tax rates and cost estimates imply the cost share of the tax (St) was between .86 and .93; and that the combined cost share of corn and malt (Sc + Sm) was between .07 and .14. (That is, the tax constituted roughly 90 percent of the cost of producing one gallon of spirits; and corn and malt constituted the remaining 10 percent of the cost of producing one gallon of spirits.) In the empirical analysis below, we use these estimated cost shares to make predictions about how the price of spirits would have responded to changes in factor prices if distilling were perfectly competitive.

3. Estimating Market Power: A Review of Methods

In the new empirical industrial organization (NEIO), four procedures have been used to estimate market power. The first procedure exploits non-proportional shifts in industry demand to identify how prices respond to changes in the elasticity of demand. In a perfectly competitive industry with constant returns to scale, changes in the elasticity of demand would not induce a change in equilibrium price. A second procedure uses regime shifts. For example, in his study of a railroad cartel, Porter compares railroad rates during price wars, when rates were set at marginal cost, to rates during periods of effective collusion. A third procedure, first suggested by Panzar and Rosse, examines how output prices respond to changes in factor prices. One application of this approach examines how changes in tax rates affect the price of cigarettes. This approach assumes that the cigarette industry is characterized by constant returns to scale, and by demand curves with a constant price elasticity. Given these conditions, an increase in the tax on cigarettes should cause the price of cigarettes to rise by the exact amount of the tax, if the industry is competitive. In contrast, in a monopolistic industry with constant costs and price elasticity, an increase in the tax would cause prices to rise by more than the tax increase because monopolists set price at a multiple of marginal cost. A fourth procedure examines how changes in output affect prices. In a perfectly competitive industry with constant returns to scale, changes in output generated by shifts in demand would not affect equilibrium price.

Although these methods of estimating market power enjoy widespread use, they have been subject to criticism. One area of concern is that most NEIO studies make strong assumptions regarding the functional form of demand. If these assumptions are inaccurate, the resulting estimates of market power might also be inaccurate. An additional concern is that the NEIO will usually mismeasure market power if firms play a dynamic oligopoly game. This is an especially important concern in the case of whiskey distilling because observers claimed the trust’s price-setting behavior varied greatly over time.

At the industry level, we use three of the four procedures outlined above to estimate market power. First, we consider how regime changes affected price. Because the industry data extend from 1881 through 1898, it is possible to examine prices under three distinct regimes: periods of competition when there were no trusts or pools; periods when a pool effectively restricted output; and periods when the trust dominated the industry. Following Porter, we use periods of competition as a benchmark, and assume price equaled marginal cost in the absence of the trust and pool. Second, we consider how changes in output affected the price of spirits. Because distilling was characterized by constant returns to scale, changes in output would not have affected equilibrium price if the industry were competitive. Third, we consider how changes in factor prices affected the price of spirits. Given the industry’s cost structure, competition predicts that changes in the tax and the price of corn and malt would have generated proportional changes in output prices.

We conduct similar tests at the firm level. In addition, the firm-level data allow us to estimate the demand curve the Whiskey Trust faced, which in itself would be revealing given the paucity of data usually available on the trusts. At both the firm and industry level, we address the concern that the trust’s price-setting behavior varied over time.

4. Industry-Level Data and Analysis

To measure the market power of the Whiskey Trust, we begin by estimating industry-level demand and supply relations. The data are monthly, beginning in 1881 and ending in 1898. The estimation uses 2SLS and computes Newey-West standard errors to control for heteroscedasticity and serial correlation up to two lags. Because all continuous variables are in logs, their coefficients represent elasticities. The coefficients on the dummy variables are semi-elasticity measures. The industry-level analysis extends across three regimes: price wars; pooling; and the trust.

Once the supply and demand equations have been estimated, we use various parameter estimates to draw inferences about market power. We are particularly interested in three sets of estimates. First, we are interested in the coefficient on pool and trust dummies, which provide a direct measure of how much the trust and the pools that preceded it increased the after-tax price of spirits. Second, we are interested in the estimated coefficient on output (B1) in the supply relation. As explained above, because distilling was characterized by constant returns to scale, changes in output would not have affected price if industry behavior corresponded to perfect competition. Competitive pricing therefore implies B1 would be close to zero and statistically insignificant. Third, following Panzar and Rosse, we are interested in the estimated coefficients on factor prices: the price of corn (B2) the price of malt (B3) and the tax (B4). In competitive industries with constant returns to scale and a constant elasticity of demand, factor prices would generate changes in output prices in proportion to their cost shares. Given this, and that the coefficients on factor prices are elasticities, competitive pricing implies:

(11a) B4 = St [roughly 90 percent]

(11b) B2 + B3 = Sc + Sm [roughly 10 percent] and

(11c) B2 + B3 + B4 = 1

where S t, Sm, and Sc are the estimated cost shares from Section 2.b. Monopoly pricing implies:

(12a) B4 > St

(12b) B2 + B3 > Sc + Sm, and

(12c) B2 + B3 + B4 > 1

After controlling for improvements in the operating efficiency of distilleries, we find evidence that the Whiskey Trust generally did not exercise substantial market power. The coefficient on the tax is around .90, and the coefficients on corn and malt are positive though insignificant. Also, the coefficients on malt, corn, and tax sum to 1.00 in one regression and to .96 in another regression; and in formal statistical tests we cannot reject the hypothesis that the coefficients sum to one in both models. Given that these three factors constituted nearly all of marginal cost, this result is consistent with competitive pricing in the absence of pools or trusts. The coefficient on the pool dummy suggests the pool increased the (after-tax) price of spirits 4 percent above cost. Finally, while the average effect of the trust is insignificant, the coefficient on the months when observers claimed the trust charged unusually high prices suggests the trust increased prices by 6 to 7 percent above cost.

5. Firm-Level Data and Analysis

We also conduct an analysis at the firm level. Again this involves estimating supply and demand relations. As above the data here are monthly. They extend from 1888 through 1895. We are especially interested in the estimate of the demand curve facing the Whiskey Trust, because this estimate provides a clear and direct measure of the trust’s market power and its ability to raise prices above marginal cost. Simply put, the more elastic the trust’s demand curve, the closer it would have set price to marginal cost. For example, it is easy to show that a firm confronting a demand curve with an elasticity of 2 would set prices 100 percent above marginal cost, while a firm confronting a demand curve with an elasticity of 10 would set prices 11 percent above cost.

We first discuss the results for the supply relation. The supply-side estimates suggest the trust generally exercised little market power, but that it did raise prices for two short periods of time. Because distilling was a constant cost industry, the coefficient on quantity sold should be close to zero if the trust generally behaved as a perfect competitor. Consistent with this prediction, the coefficient on quantity sold is insignificant in all four regressions. The coefficients on factor prices and the tax are also consistent with perfect competition. Given that corn prices, malt prices and the tax constituted (nearly) all of marginal cost, their coefficients would sum to one if the trust behaved as a perfect competitor. Consistent with perfectly-competitive pricing, the estimated cost-shares of malt, corn, and the tax are about .08, .06, and .9 (that is, they sum to roughly one) once we control for periods of unusually high prices. In formal tests at the 5 percent level of significance, we cannot reject the hypothesis that the coefficients on these factors sum to one in any of the four regressions. The coefficients on the high-price period dummies suggest that during these two periods the trust increased the after-tax price of spirits by 7 and 13 percent above cost. These results further highlight the importance of considering the trust’s behavior over time, as well as its average effect.

The demand estimates sharpen, and ultimately help resolve, the question of why the trust increased prices for only short periods of time. The demand estimates indicate the trust faced a downward-sloping demand curve with an elasticity between 4.07 and 8.73. At first glance, these elasticity estimates are surprising. The lower-bound estimate of elasticity (4.07) suggests that if the trust had exercised all of its market power, it would have set prices nearly 33 percent above cost, while the upper-bound estimate (8.73) suggests the trust would have set prices 13 percent above cost (see equation 13). Yet the industry-level estimates, as well as the firm-level supply estimates, suggest the Whiskey Trust increased prices to these levels only occasionally and then only for very short periods of time.

One plausible explanation for this finding is that the trust was reluctant to exploit all of its potential market power for sustained periods because of entry and the threat of entry. Witnesses testifying before the United States Industrial Commission suggested this explanation when they argued:

They may start a trust and buy up a hundred or two hundred distilleries, but the moment they reach out for more than a reasonable profit on the cost of production other plants will be created faster than any trust can buy them up.

To explore this possibility, we conduct two tests. First, we estimate long-run demand curves using quarterly, semi-annual, and annual data. If entry or the threat of entry limited the trust’s ability to raise prices for sustained periods of time, the elasticity of demand in the long run would have been much larger than in the short run. The evidence on this point is ambiguous. When semi-annual and annual data are used, elasticity estimates that are statistically significant are as high as 32, but these results are not robust to alternative econometric specifications. A concern with this approach is that estimating long-run demand elasticities might not pick-up the effects of entry if there was a long lag time between entry and market-power induced price increases, or if the timing of entry was highly variable.

This concern motivates our second test, in which we analyze the monthly data for evidence of large, inward shifts of the demand curve. If market entry gradually eroded the trust's market share, it would have manifested itself in negative demand shocks. Consistent with this prediction, coefficients on year dummies indicate that the trust experienced significant reductions from peak demand during the year 1893.03-1894.02, when quantity demanded would have been 27 percent below its peak ceteris paribus, and year the 1894.03-1895.02, when quantity demanded would have been 37 percent below its peak. (The year dummies suggest demand peaked during the year 1891.03-1892.02.) Coefficients on the half-year dummies yield the identical conclusion.