The politics of beer: analysis of the congressional votes on the beer bill of 1933

Specifically, a section of the Democratic Party’s platform said, “we favor immediate modification of the Volstead Act; to legalize the manufacture and sale of beer and other beverages of such alcoholic content as is permissible under the Constitution and to provide therefrom a proper and needed revenue.” http://​www.​presidency.​ucsb.​edu/​ws/​?​pid=​29595.

Three-point-two percent alcohol by weight (ABW) is around 4% alcohol by volume (ABV). That was the ABV of a typical American lager beer prior to Prohibition. For a modern comparison, Coors Light and Bud Light are 4.2% ABV today.

April 7 is currently known as “National Beer Day” in the United States in homage to beer’s legalization on that date in 1933.

Barley output data for each state are from the 1930 Census of Agriculture. Hops data are from Table 2 of “An Economic Study of the Hop Industry in Oregon,” Agricultural Experiment Station, Oregon State Agricultural College, Station Bulletin 288, June 1931.

Barrels of beer produced in each state are from Cherrington (1915).

Importantly, a number of senators abstained from voting on both the March 16th bill and on the amended bill on March 20th. Therefore, those abstentions are excluded from the analysis. Specifically, there were 22 abstentions for the March 16th vote and 15 abstentions for the March 20th vote. Interestingly, there do not appear to be any clear patterns that emerge. Minnesota was the only state who’s entire delegation abstained from both votes, while the entire state delegations of Kentucky, South Carolina, and Wyoming abstained for the March 20th vote. An evaluation of the abstention patterns based on the median value of each of the main variables of interest indicates that only four of the abstentions occurring on the March 20th vote were from states with above the median number of brewing interests in 1932. Otherwise, there appears to be a fairly even split.

These data are available at legacy.voteview.com/dwnomin.htm. A much more thorough discussion of these scores and the construction of the index is laid out in Poole and Rosenthal (2007).

These data are from the Association of Religion Data Archives, http://​thearda.​com/​Archive/​ChState.​asp.

While a somewhat more informative analysis could include county-level data, which is more readily available for some of the variables, including religious affiliation, data on the beer industry and agricultural production are available only at the state level. Furthermore, given the prevalence of heavily gerrymandered House districts from the 73rd Congress (see Martis 1982, p. 167 for a detailed breakdown), even the use of county-level data would still require a significant amount of subjectivity on the researcher’s part to decide how best to match counties to particular districts. For these reasons, we continue to employ state-level data in the analysis. However, we did consider one specification that evaluated each individual Representative’s vote as a dependent variable using a probit model with those variables that were available at the district level (party affiliation, DW-NOMINATE scores and a dummy variable for Southern states) as explanatory variables. The results conform to what we find using state-level data—members of the Democratic Party were more likely to vote “yes”, while personal conservatism and individuals from the South were more likely to vote “no”. All variables, except the first dimension DW-NOMINATE scores were statistically significant at conventional levels. These results are not reported in the interest of space, but are available upon request.

We use the 12th day of the month as the cutoff so that, for example, if a state legalized on April 12 it received a zero, but if between April 13 and May 12, it received a one. We tried alternative cutoff dates such as the 15th and the 20th of the month, and the results are not qualitatively different throughout the study.

We also consider these specifications excluding the DW-NOMINATE score, since this may not perfectly reflect state-level legislative preferences. Overall, the results are qualitatively the same, save for the coefficient on per capita barrels of beer produced, which shifts from being marginally significant to insignificant. Given the comparable results between the specifications, we do not report these in the interest of space. However, they are available upon request.

We also considered negative binomial regressions for robustness. Overall, there are only marginal differences to report. Most importantly, each of the results for preexisting brewing interests maintain the same sign coefficients and are statistically significant. While these results have been excluded for the sake of space, they are available upon request.