Gender in American Tobacco Cards 1880–1920: The Role of Coercive Competition

Abstract

This paper adds to the literature on Marxian coercive competition and its negative economic and social outcomes. An historical and econometric analysis of competitive intensity and the portrayal of women in one early form of tobacco advertising is conducted using an original data set. The historical analysis establishes the nature and intensity of competitive relations. Estimation results for a multinomial logit model for various portrayals of women show that a 1% increase in the market share of independent producers caused a 0.35–0.7% and 2.5–4.5% increase in the likelihood that women were included and treated exploitatively in ads in early and late competitive periods, respectively.

Acknowledgments

The idea for this paper originates from the discovery of two 1889 tobacco cards during the deconstruction of my farmhouse. One card from the ‘Leaders’ series (N222) contained a series list where half of the entries were women. This progressive treatment of women generated the idea for this work, but the project focused first on the more dominant exploitive treatment of women. The author is grateful for the research assistance provided by Steve Choi, Ben Stein and Anne Tolsma.

Notes

¹In neoclassical optimization models, sunk costs, whose derivative with respect to a choice variable is zero, would have no impact on decisions.

²Goldstein (1986) refers to this particular dynamic in Marx's work as a micro-macro dialectic.

³Crotty (1993, p. 8) states, ‘But corespective relations are themselves contradictory … they can generate a wide gap between the most advanced technology … and that embedded in the capital stock. Thus, the longer corespective relations are maintained, the greater the incentive for some firm to undermine them.’ Technological change is only one mechanism that can undermine corespective relations. I consider how a dominant cigarette firm's strategy to control related tobacco sectors reignites coercive competition.

⁴The use of optimizing microeconomic models is common in the heterodox literature concerned with the contradictory nature of firm behavior on the macro level. See Goldstein (2006) for a survey.

⁵Advertising with sexual content was not the only risky strategy pursued. Firms implemented imperfect new technology (cigarette rolling machines) that ultimately increased productivity 40-fold in a fledgling industry based on a new product with uncertain demand and thus resulted in an overcapacity crisis (Porter, 1969, p. 70).

⁶One example was the crusades of Anthony Commstock, a US postal inspector and politician. In 1873, Commstock founded the New York Society for the Suppression of Vice and also was responsible for legislation making it illegal to deliver by US mail ‘obscene, lewd, or lascivious’ materials. Cruse (1948, pp. 48, 110–111) dates the first public outcry to 1887–88. He states, ‘One famous “reformer” of that day, Anthony Comstock, organized a Society to suppress the sensuous and immoral cards. Carrie Nation, another zealot, well known in many crusades joined in the fray.’

⁷Smith (1973) finds that the feminist reformers addressed the societal role of women including attacks on male ethos/double standard and limited economic opportunities for women.

⁸The same argument can be advanced for other forms of non-marital sex, including advertising with sexual content directed at men.

⁹This latter position contained both an economic critique and criticism of a male double standard. Smith (1973, p. 51) argued ‘collectively they tended to sympathize with the prostitute or fallen women and condemn the male exploiter or seducer.’

¹⁰‘Many sets of cards featured either photographs or lithographs of buxom young ladies in what must have seemed very daring, if not shocking, costumes. Usually these sets were labeled simply “Actresses’”… Since there was surely little personal identification by the purchaser with the stars, who were usually unnamed, and since actresses were then accorded a low place in the social scale of polite America, it seems clear that such were designed for prurient attraction' (Porter, 1971, p. 35). Additionally, Mitchell (1992, p. 329) refers to the controversy caused by eroticized nudity in art.

¹¹An 1894 letter from W. Duke to J.B. Duke, President of Duke and Sons, a cigarette manufacturing company, discussed below, clearly enumerated these risks.

¹²The Big Four is merely a subset of the Big Six where the two Baltimore-based firms faded from dominance.

¹³Durden (1975, pp. 45–48, 55).

¹⁴Turkish cigarettes were not amenable to machine production. Thus, their wide acceptance reduced the technological advantage of the Trust.

¹⁵These figures combine Duke with the Big Four in the earlier period.

¹⁶US Bureau of Corporations (1909, ch. 23) describes the struggle in this sector and Durden (1975, p. 71) suggests that failure to capture a large share was due to very decentralized and geographically scattered producers and the inability to integrate machine rolling technology.

¹⁷Total percentages can exceed 100% due to the classification of some topics in more than one category.

¹⁸This is the language Forbes & Mitchell (1999) employ to describe the sexual slant of some series. In addition, Cruse (1948, p. 44) states ‘Actresses, pretty women and glamorous beauties formed the principal attraction of the early American and British cigarette cards.’

¹⁹Ranges are provided due to my treatment of the subjective exploitative classification assignments discussed below.

²⁰Cruse (1948, pp. 48, 110–111) notes that the Comstock campaign never gained significant momentum. As a result, the cards ‘caught the public fancy and won remarkable popularity among all classes.’

²¹Durden (1975, pp. 60–61).

²²See Reichert (2003).

²³The New York Public Library's Digital Gallery card collection website http://digitalgallery.nypl.org/nypldigital/explore/dgexplore.cfm?col_id=161 and one additional site http://www1.coe.neu.edu/~dan/z-ns-start/showtime.html were used. Forty percent of series included in this study were viewed in this manner.

²⁴The vast majority of excluded series had four cards or less. Also excluded are the deluge of sports series issued starting around 1910 and numbered from T200–T299. I only include sports series issued as general issues prior to 1910.

²⁵For example a presidential series, N472, lists Grover Cleveland as the last and 22nd president. Given that Cleveland served as the 22nd and 24th presidents, it is likely that this series was issued in the late 1880s.

²⁶When reported dates result in a range of dates, I use the midpoint of the range for ranges that are 5 years or less. For larger ranges, a date is not assigned unless additional information is available.

²⁷Competition is a dynamic process that is unlikely to be captured by static/threshold determinations of industry structure. The strategies of non-dominant competitors, their likely success and the response of the dominant firms is what determines whether coercive competition exists. This can only be determined via an historical analysis. While an industry might be classified as concentrated via a share threshold, it is quite possible that the competitive strategies of challengers can sufficiently erode market share by enough to set off coercive competition without crossing below the concentration threshold characterizing market structure.

²⁸The vast majority of series (118/125 or 94.4%) classified as exploitive contain an exploited women in every card. In the case of progressive treatment, the median proportion of women in a series thus treated is 0.08, while the mode is 0.02. Thus, the proportion of women progressively treated has little variation with values close to zero. Progressive treatment in this sense is token progressive treatment. Finally, 63% of series classified as neutral contain 100% women all treated the same. When men are included, the proportion varies some, but is typically close to 50%. In a regression analysis, this variation would mostly be explained by the inclusion of men in a series. Thus, little would be added from considering the proportion of women as the dependent variable.

²⁹The ability to normalize coefficients on the basis of one outcome underlies this advantage.

³⁰In the former, while twice the parameters are estimated (in a three choice model), the potential sample size is 534. In the latter, the potential sample sizes by treatment are either 125 (neutral), or 16 (progressive) or 144 (exploitive).

³¹For example, a series that featured the architecture of famous buildings and landmarks by its subject matter would not include human subjects.

³²These cases are associated with series numbered N91, N135, N140, N281, T76, T104, T115, T116, and T117.

³³While burlesque had its start as ‘travesty’ in the 1830s, by 1880 it had evolved into a form of entertainment that had more sexual content.

³⁴The formal derivation of these outcomes from a three-equation structural model is available upon request from the author.

³⁵Given that the progressive treatment of women occurred respectively in only 3.1% (16 series) of series and that women were ignored in only 4.1% (21) of series, these combinations significantly mitigate the zero cell/specification problem.

³⁶For convenience, the results when BOXEVENT and BBATTEND are included are not presented. These variables were found to be statistically insignificant using a likelihood ratio test.

³⁷See Mitchell (1992).

³⁸This interpretation of market share does not consider potential declines in dominant firms' share, profits and advertising as a result of increases in rivals' share. Dominant firms in the early period and the Trust in the latter period faced an over capacity/over investment crisis and as a result had considerable amounts of irreversible investment sunk into their market positions. This necessitated that their illiquid capital be defended and the primary defense mechanism was aggressive advertizing (Porter, 1969, p. 70).

³⁹Even in the Trust era, many of the ATC's advertising decisions, besides general guidelines, were made via decentralized decision making on the plant, brand and tobacco type basis.

⁴⁰I do not conduct a formal (Hausman) test for the endogeneity of ISH and Subject due to the well known limitations of that test and the practical implementation problems associated with my application. In general, these tests have low power. The large number of subjects makes a general test for the endogeneity of subject choice impractical. While a test for the endogeneity of PRETTY that comprises 45% of all subject matter and is the main source of exploitive treatment is feasible, there are still practical limitations. The dichotomous nature of PRETTY unnecessarily complicates these tests and the small variation in such variables makes interpretation of results more difficult. Finally, the interactive nature of the ISHE and ISHL variables (market share times a period dummy) further complicates the specification of equations that either generate residuals or fitted values for potentially endogenous variables used in variants of endogeneity tests. In contrast, I rely on theorizing concerning endogeneities.

⁴¹It should be noted that the final equation is derived from the over-fitted equation through the use of a series of likelihood ratio tests for the joint hypothesis that a subset of variable coefficients are simultaneously equal to zero.

A fuller specification than the over-fit equation reported was also estimated. It included two additional exogenous determinants of subject choice: BBATTEND and BOXEVENT. These variables are excluded here because they added very little to the analysis. They were statistically insignificant and did not alter the qualitative nature of the results. Given the limited number of sports issues and the smaller number of baseball and boxing specific issues and that it is unlikely that card issuers used sporting series to avoid having to include women because they targeted an audience that likely would respond to sexual content (SPORTS and PRETTY subjects are complements), these variables are not relevant. In contrast, the popularity of animal and nature series and the use of such series as an alternative to the inclusion of women better justifies the inclusion of the NSOC variable (ANNAT and PRETTY choices are substitutes).

It should be noted that LR tests for the exclusion of variables use the equation with BBATTEND and BOXEVENT (not reported in Table 4) as the base equation. Log L for these equations are 320.36 and 320.12 respectively for the TREATW and TREATW1 equations.

⁴²At least part of the difference in the magnitude of these effects across competitive periods is expected due to differences in the level of shares in the two periods (the discontinuity discussed above).

⁴³A test for the sum of the ORIENT coefficients in the neutral/progressive and exploitive cases reveals that the sum is significantly greater than both zero and 0.763 at the 5% level of significance.

⁴⁴The percentage of shares issued to Kimball during the formation of the Trust was, along with Goodwin, the smallest, 10% in comparison to the 30% of shares given to each of AG and Duke (Durden, 1975, p. 57). This is indicative of Kimball's weaker position within the top five firms. Kinney received an intermediate distribution of 20%.

⁴⁵The number of societies in existence is taken from Environmental History Timeline 1810–1890 and 1890–1920 see www.radford.edu/wkovarik/envhist/4industrial.htmland/5progressive.html. Accessed 22 February 2010.