In a well-crafted recent article in this JOURNAL, Robert Fleck argues that voting on the Fair Labor Standards Act (FLSA), the last and among the most contentious pieces of major New Deal legislation, was heavily influenced by regional political differences and by the “southern political system.”¹

This contention is broadly supportive of a view held by several traditional historians, who have argued that political differences between the North and South were a major factor in voting.² According to this view, the considerable support for minimum wages among southern workers, African American leaders, and the general public did not translate into congressional votes for the FLSA due to southern political institutions that effectively disfranchised many of the voters who were predisposed to support the FLSA. In a previous article I argued that, although southern representatives were more likely to vote against the FLSA, this was primarily because of economic differences between the two regions, not political factors. I argued that the opposition to the FLSA was spearheaded by low-wage employers, who were disproportionately concentrated in the South. This note examines the reasons for these conflicting results. It is argued that Fleck's approach to prediction, which evaluates the marginal impact of his independent variables individually, but does not jointly consider the variables that defined the southern political system, is misleading, and a more appropriate approach using his own regression results supports my contention that political differences between North and South were inconsequential to the overall roll-call vote on the FLSA. It is shown that, holding economic factors constant, had the South been politically like the North, the estimated roll-call vote on the FLSA would have been very similar to the actual votes in 1937 and 1938.

The effects of the minimum wage have been extensively analyzed in a wide literature. Three stylized conclusions of this literature form the basis of both Fleck's and my approach to estimating the economic determinates of voting. First, the elasticity of labor demand was sufficiently low that a minimum wage at the level proposed under the FLSA would have resulted in a net redistribution to workers.³

Second, the minimum wage provisions of the act would have reduced the profitability of low-wage firms. Third, the minimum wage provisions would have increased the profitability of high-wage firms that directly competed with covered low-wage firms. The first of these stylized facts implies that workers in both the North and South would have supported the FLSA. The second implies that low-wage firms, which were disproportionately located in the South, would have opposed the FLSA. The third implies that high-wage firms, which were disproportionately in the North, would have supported the FLSA. The evidence from the congressional hearings and from both Fleck's and my analysis of voting is strongly supportive of each of these implications.⁴

Where Fleck and I disagree is in the impact of political factors on legislative voting.

⁵

Specifically, we disagree on whether southern political institutions, which effectively disfranchised many African-Americans and poor whites, groups that disproportionately worked in low-wage jobs and strongly supported the FLSA, were an important factor in southern opposition to the bill. In my article, I used a dummy variable for South (the 11 states of the confederacy, plus Kentucky and Oklahoma) to capture the effect of disfranchisement.⁶

The coefficient on South was small and statistically insignificant, from which I concluded that regional political differences were unimportant after controlling for economic factors and legislator ideology. Fleck improves upon my specification by quantifying the specific political factors that characterized the South. His regressions on voting include Turnout (measured at the district level) and Turnout-Manufacturing Correlation (measured at the state level) as independent variables.⁷

Specifically, he uses probit regressions to estimate relationships of the form:

where ECON is a vector of economic characteristics, and IDEOLOGY is legislators' two-dimensional D-NOMINATE scores.

⁸

His estimates show that increases in both Turnout and Turnout-Manufacturing Correlation had a positive impact on the likelihood that a legislator voted in favor of the FLSA. Furthermore, the coefficients on both variables are statistically significant across a range of model specifications. He then estimates the marginal impact of Turnout, Turnout-Manufacturing Correlation, and the various economic factors. He shows that, all else equal, if the southern mean of Turnout were the same as the northern mean, legislators from the two regions would have had similar predicted probabilities of voting for the FLSA. Furthermore, he shows that an increase in the southern mean of Turnout-Manufacturing Correlation could also have resulted in southern legislators being as likely as northern legislators to vote for the FLSA.⁹

I agree with Fleck that the significance of the coefficients on Turnout and Turnout-Manufacturing Correlation shows that political factors had some influence on voting on the FLSA. Where I disagree with Fleck is in his claim that “the southern political system had an important influence in voting on the FLSA.”¹⁰

An increase in the southern Turnout-Manufacturing Correlation would have made the South even less politically similar to the North than was actually the case in 1937. In effect what Fleck has shown is that a legislator from a hypothetical state with southern economic characteristics and neither southern nor northern political characteristics would have been as likely to have voted for the FLSA as a legislator from a state with northern economic and political characteristics. It does not follow that actual political differences between the North and South had a major impact on roll-call voting. For the southern political system to have played a major role in voting on the FLSA, it would have to have been the case that a legislator from a hypothetical state with southern economic characteristics and northern political characteristics would have been considerably more likely to vote for the FLSA than a legislator with southern economic and political characteristics.

Fortunately, it is possible to test the influence of political factors and the southern political system on voting on the FLSA using Fleck's regression results. Fleck estimates the effect of political and economic characteristics and ideology on the two main votes in the House: first, the final vote to pass the bill on 24 May 1938 and secondly, the vote to recommit the bill which was passed on 17 December 1937. His estimating technique is probit. He estimates the marginal impact of each of the political and economic variables individually on the probability that a legislator votes in favor of the FLSA, but does not evaluate the combined effect of political and economic characteristics. Specifically, he does not construct estimates of the effect of the southern political system; i.e., consider the difference in roll-call voting depending on whether the values of political variables are evaluated at the southern or northern mean values, holding constant the values of all economic variables. In addition he does not evaluate the joint importance of economic characteristics to voting. This can be done using the following formula to estimate the probability that a legislator with a specific set of political and economic characteristics would have voted in favor of the FLSA¹¹

Table 1 shows the estimated probabilities of voting in favor of the FLSA using Fleck's regressions and the mean values of the independent variables from Fleck's data appendix. The columns follow Fleck's four model specifications. The rows show predicted values under various assumptions about economic and political characteristics. The first two rows show the predicted values under the assumptions that both economic and political characteristics are at their respective northern and southern means. The third and fourth rows show the predicted values if either economic or political characteristics are at the southern mean and the other is at the northern mean.¹²

It is evident from Table 1 that had the South been economically like the North, the model predicts that there would have been little southern opposition to the bill and that southern voting would have been virtually the same as northern voting. On the other hand, had the South been politically like the North, the model predicts that voting on the bill would have been split in much the same way as it actually was in 1937 and 1938. In two of the four specifications, a representative from a southern state that was politically like the North is estimated to have been less likely to vote for the bill than had the state been politically like the South.

Table 2 shows the estimated probabilities of voting against recommitting the FLSA.¹³

As with Table 1, there is strong evidence that economic factors affected voting against recommitting, but no evidence that political factors affected voting. In each specification a representative from a southern state with economic variables at the northern mean is predicted to have been about as likely as a northern representative to oppose recommitting. On the other hand, a representative from a southern state with political variables at the northern mean is predicted to have been about as likely as a southern representative from a state with political variables at the southern mean to oppose recommitting. In three of the four specifications a representative from a southern state that was politically like the North is estimated to have been less likely to vote for the bill than had the state been politically like the South.

In conclusion, this note examines Robert Fleck's claim that the southern political system was an important determinate of voting on the FLSA. To do so I have constructed a simple counterfactual estimate of the probabilities that legislators with northern and southern political and economic characteristics would have voted against recommitting the bill in December 1937 and voted in favor of passing the bill in May 1938. It is shown that regional differences in political characteristics had little effect on the voting; after holding economic factors constant, a hypothetical representative from a state with southern political characteristics is estimated to have had approximately the same probability of voting for the bill (or against recommitting) as one from a state with northern political characteristics. This suggests a much narrower role for political factors than the one implied by Fleck. District-level differences in turnout and state-level differences in the turnout-manufacturing correlations may have influenced the votes of individual representatives, but do not explain the vast difference in North-South voting on the FLSA. On the other hand, North-South differences in economic characteristics fully explain regional differences in voting.