The Two Hindenburg Elections of 1925 and 1932: A Total Reversal of Voter Coalitions

»Die Hindenburg-Wahlen von 1925 und 1932: Eine vollständige Umkehrung der Wähler-Koalitionen«. This article compares the two presidential elections of 1925 and 1932 in an attempt to determine the shifts between these two elections which brought Paul von Hindenburg to power. Although this article does not attempt to add to the historiography of Hindenburg’s election and the subsequent deparliamentarization which has often been thought by historians to have eased Hitler’s transition to power, it attempts to use statistical verification to underline a number of hypotheses generally agreed upon by historians, but which lack substantial evidence. In considering Hindenburg’s election, a number of variables are considered, such as: which parties the Hindenburg voters came from, why Hindenburg was backed rather than his oppositional candidate Wilhelm Marx, and what social background the Hindenburg voters had. Also, the commonly held belief that many of the communist voters fluctuated from the communist party candidate, Ernst Thälmann, to Adolf Hitler is statistically analyzed.

1925 and 1932, are among the most fascinating and historically significant elections of modern German history (see Table 1). They are fascinating for the electoral historian and the generalist alike because of the virtually total reversal between 1925 and 1932 of the voting coalitions that backed and brought to power the aged Field Marshal Paul von Hindenburg. And they are historically extremely significant because it was von Hindenburg who at least encouraged if not sustained the creeping process of departmentalization after 1930, a process that finally brought Hitler into power. It may be readily speculated that another president, e.g., Wilhelm Marx, who as the candidate of the Weimar coalition parties was Hindenburg's chief opponent in 1925, would not so easily have dismissed Reich Chancellor Heinrich Briining in May of 1932. And Marx undoubtedly would not have appointed the right-wing Center party dissident, Franz von Papen, as Briining's successor.
Astonishingly, these two really important Weimar elections have yet to be adequately investigated by electoral historians. An analysis of these elections therefore virtually has to start from scratch. The outcome does not, perhaps, necessarily add something new to what has been assumed by historians about the two Hindenburg elections. The significance of the following analysis lies more in the fact that it provides statistical confirmation for some more or less commonly held but never sufficiently corroborated hypotheses. In the following, I will turn first to the 1925 election in order to find out where-i.e., what parties-the Hindenburg voters came from, and what role was played by the decision of the Catholic Bavarian People's Party (BVP) 225 The Hindenburg Elections 0/1923 and1932 to back Hindenburg instead of the candidate of the Catholic Center Party, Wilhelm Marx. Then I will examine the transition from 1925 to 1932. I will ask what statistical relations may be observed between these two elections and, from a complementary perspective, what social groups supported Hindenburg in 1925 and 1932, respectively. In a third and final step I will try to find out if there really was a significant voter fluctuation between the Communist candidate Ernst Thalmann and Adolf Hitler from the first to the second ballot of 1932, as is so often alleged in contemporary and historical analyses of the collapse of the Weimar Republic. Of course this question cannot be answered directly or beyond any reasonable doubt, since we do not have any methodologically reliable and representative opinion polls for the Weimar period. What we can do, however, is to look first at the statistical relationship between the Hindenburg vote and the vote of other parties and candidates at the level of the i, 200 German counties and cities of that period. The results are statistically sound if we restrict the verbal interpretation of our findings to the territorial, that is, the county, level. Since we are, however, much more interested in individual-level relationships, I will try, in a second-statistically somewhat risky-step, to discern the underlying (but unknown) "true" voting transitions to and from Hindenburg by means of multiple ecological regression analysis. 1 2.1. Some party-vote correlations of the ig25 Hindenburg vote  We thus get five categories with an equal number of counties in each. For each category we assess the percentage of voters won by other parties or candidates than Hindenburg in a series of elections between December 1924 and July 1932. If there is a low percentage of votes for the other parties and candidates in the first quintile (where the Hindenburg vote was lowest) and a growing percentage of votes in the following quintiles (where Hindenburg fared better), we have a positive statistical relationship between the two votes. The correlation coefficient therefore is positive in sign and rather high in magnitude. This is the case for the statistical association between the Hindenburg vote on the one hand, and the vote for the German National  Party (DNVP) or the volkisch-Nazi coalition in the late-1924 parliamentary elections. A positive correlation also exists between the Hindenburg vote and the vote for the joint presidential candidate of the DNVP and the right-liberal DVP on the first ballot of 1925, Karl Jarres. In other words, the higher the Hindenburg vote of 1925 was in a county, the higher, on the average, the DNVP or Jarres vote was in that same county. The opposite applies to candidates who won relatively more votes in the first than in the subsequent Hindenburg quintiles, as is the case with Wilhelm Marx, his close competitor of 1925. The correlation coefficient still is comparatively high, but now of course negative in sign.
We thus find out that German Nationalists and the 1924 coalition of volkisch and national-socialist splinters, as well as Jarres, displayed the same distribution of votes as Hindenburg did: they fared much better, on the average, in counties where Hindenburg was strong than in counties where Hindenburg was weak. For example, in the 165 counties of the first quintile, the Jarres vote amounted to not more than 11.8 percent of the electorate, while in the fifth quintile, the Jarres vote was up to 45.5 percent. In addition, there is a slight, curvilinear relationship between the Hindenburg vote and each of the following: turnout; the vote for the first-ballot candidate of the volkisch Right, Erich von Ludendorff; and, quite unexpectedly, the vote for the firstballot presidential candidate of the Social Democrats, Otto Braun.

Some ecological regression estimates of the "true" voter fluctuations to and from Paul von Hindenburg in ig2$
It would be quite hazardous to interpret these findings in terms of individual or group relationships-to assume, that is, that all or most Hindenburg voters were necessarily former Jarres and DNVP voters. So-called ecological fallacies, such as the erroneous assumption that the relationships of one level of analysis would be equivalent to the other, could (but by no means necessarily must) result from such a tacit assumption of congruence. 3 To get somewhat better estimates of voter fluctuations, one has to take into consideration the development of the other parties or candidates as well. This is done by multiple ecological regression analysis-a powerful but somewhat dangerous statistical technique that bases its estimates on rather "strong" distributional premises such as linearity, non-contextuality of relationships, etc. Only if these premises are met by the data (which we cannot fully know) can the estimates of ecological regression equations be inter- Cell entries: Transition probabilities, estimated by multiple regression analysis; county level data. "From* = percentage of pany electorate switching to Hindenburg; To" = percentage of Hindenburg voters switching to one of the indicated parties. preted as "true" individual level fluctuations. If not, they still represent a good aggregate level estimate of the statistical relationship between the development of the Hindenburg vote and the vote for other parties and candidates. Since we cannot completely know if all assumptions of the method are really met, we should restrict our interpretation of the findings to differences of magnitude. 4 Tables 3 and 4 report some ecological regression estimates of the voter fluctuations to and from Hindenburg. The cell entries represent percentages. The first column of numbers of Table 3 informs about the transition probability of the December 1924 Reichstag voters from the parties indicated at the left of the table to Hindenburg. According to these estimates, between three-quarters and four-fifths of all rightwing voters (i.e., NSFB, DNVP, DVP, and various splinter parties) of December 1924 seem to have supported their joint second-ballot presidential candidate, Hindenburg. From the other parties and the nonvoters, only a rather insignificant minority seems to have voted for Hindenburg. The flux of voters from the various candidates of  Cell entries: Transition probabilities, estimated by multiple ecological regression analysis. If we assume for the moment that we can fully trust our ecological regression estimates, about every sixth Hindenburg voter of 1925 voted DNVP in 1928, and every fourth seems to have voted SPD. The latter result, which at first sight looks quite contraintuitive, was probably due to the influx of new voters into the SPD in 1928-voters who, according to other ecological regression findings, seem to have voted DNVP in 1924 (and consequently Hindenburg in 1925) and probably defected to the Nazis after 1928. The current findings suggest that about 20 percent of former Hindenburg partisans voted for the Nazis in 1930-a number that accounts for almost half of the NSDAP electorate of that year. Other parties, with the quite plausible exception of the splinter parties, were not able to gain substantial numbers of Hindenburg's 1925 voters.

A closer look at the contribution of former Held and Thalmann voters to the electoral success of the Hindenburg ticket
In some accounts of the 1925 presidential election, Hindenburg's second-ballot victory is attributed either to the refusal of the   for Bavaria, while Table 6 displays transition probabilities. The effects of the BVP's recommendation for Hindenburg are clearly discernible. On the first ballot, Jarres got only 6 percent of the eligible voters in that 20 percent of the Bavarian counties where the BVP vote of the previous December was highest; the overall correlation coefficient is rather strong and negative in sign (-0.74). By contrast, Hindenburg was able to collect 40 percent of the electorate in the heaviest BVP (and first-ballot Held) precincts. By the same token, Marx won many fewer votes here than might have been expected. Table 6 indicates that approximately 60 percent of the first-ballot Held partisans followed their party's recommendation and voted for Hindenburg on the second ballot, compared to only about 20 percent who switched to Marx. This would indeed imply that about half a million votes could be attributed to the BVP's unfortunate recommendation. In the light of Hindenburg's past political record, the BVP's electoral policy may be characterized as shortsighted if not frivolous.

THE 1932 HINDENBURG ELECTION
It is well known that Hindenburg's presidential record was far better until 1930, or even March 1932, than many liberal and socialist commentators had expected. In 1932, the Weimar coalition parties even regarded the Field Marshal as the only chance to keep Hitler and the NSDAP from power. Thus, at the age of 85, much against his own 1925 Hindenburg Vote intention (he would have preferred either to head a right-wing ticket or to be the "nonpolitical" candidate of the whole people), Hindenburg changed political camps in regard to the political parties supporting him. 7 It is interesting to explore the voters' reactions to this change of coalitions and to investigate the parallels and differences between the Hindenburg electorates of 1925 and 1932.
Another look at Table 2 shows that Hindenburg's electoral success in 1932 was highest in those counties where he was least successful in 1925. Hitler, on the other hand, was able to draw much more electoral support in the old Hindenburg strongholds than Hindenburg himself (see also Chart 2). If the transition probabilities in Table 4 Table 7, reveal the radical rearrangement undergone by the Hindenburg voting coalition. In 1925, the Hindenburg vote was lower in predominantly Catholic, in urban, industrialized districts, and in regions where unemployment was above average. By contrast, the Hindenburg vote of 1932 increased with the number of Catholics and self-employed in the district. And Hitler's constituency of 1932, like Hindenburg's of 1925, was located in predominantly Protestant counties, in rural areas, and in districts with lower than average unemployment rates. 8 The information presented in Table 7 is bivariate in character: only two variables are compared at one time. The real world, however, is  (Table 8). In order to construct such a "tree," we first divide the 831 county units of the Reich into three subgroups according to the percentage of Catholics living in these counties (religious denomination is by far the most important predictor of the Hitler and Hindenburg vote in 193 2!). For these three subgroups of counties, we calculate the average percentage of Hindenburg, Hitler, and Marx voters. In the next step the three denominational county classes are then divided according to their degree of urbanization. Again the average percentage of Hindenburg, Hitler, and Marx voters is calculated for each of the resulting six groups. We thus find, for example, that the Hindenburg vote was far below average in rural Catholic areas in 1925 (23 percent); in 1932, however, Hindenburg was able in these very same counties to mobilize 59 percent of the eligible voters, while Adolf Hitler was able to win only 19 percent of the electorate in this branch of our tree. In the next and final step, the resulting six county classes are again divided into three sub-classes each, according to the prevalent economic sector, so that we are now looking at 18 different county categories which are socially and politically more homogeneous than the less differentiated branches of the tree above this last level. We then determine the share of the vote in each of the eighteen branches for the three main contenders of the two elections under consideration. 9 While space constraints prohibit a detailed description, one can readily see from the "tree" that the Hindenburg voting coalition underwent a radical change: the distribution of Hindenburg votes in 1932 is much closer to that of the Marx vote of 1925 than to the first Hindenburg vote. Likewise, the Hitler vote of 1932 closely matches the Hindenburg vote of 1925: in those socially defined subgroups where Hindenburg's showing was strong in 1925, Hitler gathered an above-average share of the votes in 1932, and vice versa. From this perspective, the conservative and right-wing voter coalition that 9. Analogous "trees" for all major parties and Weimar elections are presented in Jurgen W. Falter et al., Wahlen und Abstimmungen in der Weimarer Republik, 194-203.
brought Hindenburg into power in the first Weimar presidential election may indeed be described as the harbinger of the electoral triumphs of the NSDAP of 1932 and 1933. It therefore may be interpreted as the first effective gathering of the antirepublican forces that would later bring the Weimar Republic to an end. It is often suggested that the increase in Hitler's constituency (about 2 million votes) during the second ballot of the 1932 presidential election may have been largely due to defections from the Communist leader Ernst Thalmann, who lost about 1.2 million votes. This hypothesis, which is based mostly on local impressionistic evidence (the proverbial Communist tavern which changed colors overnight), is rooted in the widespread conviction that ultimately the totalitarian extremes were not so terribly far apart and that the step from the Communists to the Nazis was much more readily taken than ideology or propaganda might lead one to expect. This idea of the proximity of the extremes finds additional theoretical endorsement in the conviction that many, if not most, of Hitler's and Thalmann's followers were unpolitical, socially uprooted products of mass society, so-called protest voters who could easily be seduced by unrealistic promises and who therefore fell prey to the totalitarian temptations of the time. 11 However, little quantitative evidence has ever been provided that would either prove or disprove this transition hypothesis.
In Table 9A, the statistical relationship between the percentage point change of the Thalmann and Hitler vote between the first and second ballot of the 1932 presidential election is scrutinized. Again, quintiles and correlation coefficients are examined. In contrast to Tables 2 and 5, however, we are now looking at so-called change variables, i.e., percentage-point differences of the vote between the first and second ballot. What we find is a near-perfect independence of the develop- gression analysis of the percentage-point change of the Hitler vote, we detect a small but significant positive relationship: the increase of the Hitler vote from March to April 1932, was somewhat higher where the Thalmann vote was above average in the first ballot or declined more strongly from the first to the second ballot, as the standardized regression coefficients of the following two equations show: The ecological regression estimates amount to about 13 percent of former Thalmann voters switching to Adolf Hitler in the second ballot (see Table 9B). This would imply that almost 30 percent of the new Hitler voters would indeed have been former Thalmann followers. Another 20 percent of the Thalmann supporters seem to have abstained during the second round of the presidential election. But according to the same ecological regression findings, the vast majority of the new Hitler voters of April 1932-about 60 percent-were former Duesterberg supporters. Again, the voter fluctuation seems to have been more complex and differentiated than is normally assumed. If these ecological regression estimates are correct, then there were a few hundred thousand first-round Thalmann voters who joined the ranks of the Hitler coalition. While they made up neither a majority of Thalmann defectors nor of new Hitler recruits, they are sufficiently numerous to give credence to the local events and personal experiences reported in the biographical literature.