American politics in the 1930s underwent an electoral realignment, traces of which can still be felt today. The Great Depression was the proximate cause that toppled the Republicans as the dominant party, a status it had enjoyed in national politics since the Civil War, and very comfortably so since the 1890s. Along with the New Deal, the economic crisis installed the Democrats as the newly dominant party in the American electorate for years to come. This is a familiar story, based on surveys conducted long after the events. But as we demonstrate with a "real-time" analysis of party loyalties in the 1930s and 1940s, the Depression and New Deal fell short of giving the Democratic Party a secure lead in the partisan loyalties of ordinary Americans. It was World War II and postwar prosperity that transformed the partisan balance heavily in the Democratic favor. If any particular election may claim a critical role in this electoral realignment, it is the 1948 election. It marks the moment when the Democratic lead in party identification reaches the magnitude that would be recorded in the National Election Studies for decades to come. The Democratic surge in party identification is fueled by the entry into the electorate of a generation that came of age during the 1940s.
We reach these conclusions with the help of survey data on party identification that the Gallup Organization began collecting in early 1937. From that point on until 1952, when the National Election Studies entered the field, nearly 200 opinion surveys were conducted (all but a few by Gallup) that asked about party identification. (1) They comprise a pool of close to half a million respondents. This data set has remained unexplored by students of electoral behavior, except for attention to a scattered survey here or there. Research about party identification as it may have existed in the 1930s or earlier has largely been limited to survey data that were conducted long after the event. The main source, the National Election Studies, did not commence in earnest until 1952. Using such latter-day data to capture attitudes such as party identification 20 years or more in the past requires strong faith in questionable assumptions.
The 1937-52 pool of polls provides an invaluable opportunity to establish the timing of the "New Deal realignment" as it unfolded, the circumstances that generated it, and how much it owed to conversion and generational replacement. Our findings help illuminate the nature and dynamics of historic change in party identification. They also speak to the utility of the realignment concept, which has lost much of its luster in recent years. And finally, the evidence of a major shift of party identification in 1948 should dispel any lingering mystery about Harry Truman's upset victory in the presidential election that year.
We begin with a review of research on the New Deal realignment, noting shortcomings of the evidence for the timing and process of partisan change. Second, we describe the sampling procedure and partisanship measure used by the Gallup polls that provided the bulk of our data; the measure proves a close match for the one used by the National Election Studies. Third, we track aggregate party identification over the course of 1937-52, locating the point in time at which the Democratic lead reaches the magnitude recorded by the National Election Studies beginning in 1952. Fourth, we probe for causal effects of economic conditions and World War II on fluctuations of aggregate partisanship. Fifth, we disaggregate partisanship by birth year to see which age cohort exhibits the strongest movement toward the Democratic Party and how well the various cohorts maintain their partisanship across the years when the polls were conducted (1937-52). Sixth, we probe for a special legacy of World War II on partisanship by examining the effect of wartime service, especially among young veterans. We conclude with a reassessment of the realignment concept. A realignment of partisanship in the electorate requires more than a historic crisis, however traumatic, or a new policy vision, however appealing. It takes compelling success of the ascendant party in mastering historic crises.
The New Deal Realignment
The New Deal realignment stands out as a classic exhibit of this political species. (2) The realignment concept is touted by its advocates for "divid[ing] much of American political history into clearly demarcated 'party-systems eras,' bounded by realigning upheavals from preceding and succeeding eras" (Burnham 1991, 101). The study of realignment has been praised as "one of the most creative, engaging, and influential intellectual enterprises ever undertaken by American political scientists" (Mayhew 2002, 1). At the same time, the realignment concept has been dismissed (by the same author) as a "genre" with little guidance to the last two centuries of American politics: "it is too slippery, too binary, too apocalyptic" (Mayhew 2002, 5,165). The verdict is based on the examination of 14 claims derived from the realignment literature (primarily Beck 1974; Burnham 1970; Key 1955; Sundquist 1973). (3) The 1932 election, widely seen as the critical trigger for the New Deal realignment, does not end up among the top three in American history on these 14 claims, or even ranks highly on many of them (Mayhew 2002, tbl. 7.1). Instead elections not commonly accorded realignment status (such as 1948) fare better than the 1932 election and the other canonical specimens (1860 and 1896) on many of the 14 criteria.
Of greatest concern in this study is a condition embodied in Mayhew's claim #1, which pertains to V. O. Key's definition of a "critical election": a "profound readjustment" in the electorate's "standing decision" that proves "durable" (Key 1955; Key and Munger 1959). (4) We translate this to mean a profound and durable shift in the balance of party identification. Did the New Deal realignment produce such a shift? While studies agree that it did, they differ on the timing and the process of attitude change that led to the shift in aggregate partisanship.
The American Voter made a strong case for the generational type of change (Campbell et al. 1960). According to this view, the partisan balance shifts when a new generation of young voters, in response to a profound event, enters the electorate with a partisan imprint that distinguishes it substantially from the rest of the electorate, which stays put (Beck 1974; Jennings and Niemi 1981). Such a process is congenial to the notion of party identification as a largely immovable attitude. Campbell et al. (1960) consider the Great Depression as the key event that triggered the realignment of party identification. It "swung a heavy proportion of the young electors toward the Democratic Party and gave that party a hold on that generation" still visible in surveys conducted in the 1950s (Campbell et al. 1960, 155). The evidence, however, comes from surveys conducted 20 years after the event (the National Election Studies of the 1950s) and recall of votes cast in the 1930 elections. The intervening years may have distorted memories and changed partisan attitudes based on experiences since the 1930s.
Taking issue with the generational type of partisan change, Erikson and Tedin (1981) made a strong case for conversion. According to this scenario, voters across the ages change party identification as they vote for the opposite party under extraordinary circumstances like the Depression and New Deal. Such an interpretation is compatible with the "revisionist" view that treats party identification as a "running tally" of shortterm evaluations (Fiorina 1981) or presidential approval and economic sentiment (MacKuen, Erikson, and Stimson 1989). Using Literary Digest polls from as far back as 1924, as well as a Gallup Poll from March 1937, Erikson and Tedin (1981) showed that in the 1930s both established and young voters swung toward the Democratic side in party attachments. "By 1936 the realignment was essentially complete. Party identification had become consistent with voting behavior." (Erikson and Tedin 1981, 961). Though using real-time data on party identification, the analysis did not explore how well the conversion held up in the years to come. Did the Democratic lead in party identification at that moment persist long enough to be matched by the first probe of the first National Election Study in 19527
Situated somewhat between generational replacement and conversion is the possibility of mobilization (Andersen 1979; see also Campbell et al. 1960). The mobilization in this case concerns the nonpartisan segment of the electorate that has stayed aloof from electoral politics until a crisis like the Depression rouses it to political life. The decision of Independents to embrace a political party clearly entails a change of political mind, though without abandoning another party. It looks more like a late baptism than an outright conversion from one partisan faith to another. Aside from younger voters, The American Voter also credits older voters who had failed to vote previously with helping the Democrats gain a long-term gain in party identification (Campbell et al. 1960, 153). But the evidence only covers the vote choices of "delayed voters" in 1932 and 1936, not their party identification. And Andersen (1979) relies on a reconstruction of party identification in the 1920s that has met with strong criticism (Niemi, Katz, and Newman 1980).
These issues aside, it is also fair to ask whether the Depression and the New Deal constitute sufficient grounds for a massive realignment of party loyalties in the electorate. However disastrous the economic collapse may have been, is a purely negative experience enough to move the electorate from the party responsible for the calamity to the opposite one? There is no question that the incumbent party will suffer at the polls, but whether the...