Why Americans Split Their Tickets: Campaigns, Competition, and Divided Government by Barry C. Burden, David C. Kimball
|Reviewed By:||James E. Campbell|
|Reviewed in:||The Journal of Politics|
|Date accepted online:||11/01/2005|
|Published in print:||Volume 66, Issue 4, Pages 1304-1347|
Although most Americans cast straight party tickets, about 15% of voters between 1952 and 1964 and about 25% of voters between 1968 and 1992 split their presidential and House votes. Barry C. Burden and David C. Kimball ask why these voters split their tickets. Their answer is that the lack of competition in congressional elections and, to a lesser extent, popular presidential candidates drive ticket splitting. Faced with a well-financed congressional incumbent with moderate issue positions, a number of voters are pulled away from their party in the congressional contest (if the seat is contested at all) to cast a split ticket (80). A popular presidential candidate of the opposite party can also induce a split ticket (85). Ticket splitting increased since the 1960s largely as the result of the conjunction of the decline in local congressional competition and the gradual realignment most evident in the South. With the completion of the staggered Republican realignment producing a more competitive party system nationally, partisanship was restored nearly to prior levels, and ticket splitting declined to around 18% in the 1996 and 2000 elections.
Burden and Kimball's analysis focuses largely on estimated ticket-splitting rates at the congressional district and state levels (for their Senate analysis). These are derived using Gary King's ecological inference technique. The authors present careful validity checks of these intricately processed estimates against National Election Survey (NES) data (uncorrected for the actual presidential and House vote divisions) and find a close correspondence. While they take comfort in this correspondence, this may argue instead for relying on NES data more heavily since there are not so many district contextual variables in Burden and Kimball's models that there would be a low N problem for district types (e.g., there are plenty of respondents in Democratic incumbent districts), and ticket splitting is ultimately an individual act best assessed with individual-level data. Moreover, because of the huge turnout disparities among districts, treating districts as equal cases amounts to treating voters very unequally and may lead to erroneous findings (as it often has in evaluations of electoral system bias).
Despite this problem and perhaps an underemphasis of presidential sources of ticket splitting (the national presidential vote division is strongly correlated with the rate of Republican-presidential and Democratic-congressional ticket splitting from 1952 to 2000 (r = .78)), Burden and Kimball basically have it right regarding the causes of ticket splitting, why it increased after the 1960s, and why it declined in the 1990s. They also have it right regarding what has
Burden and Kimball extend their analysis (Chapter 5) to divided voting between elections and explore the basis for midterm seat losses for the president's party. This is the least successful portion of the book. Although they raise an interesting point about party identification change between elections, their claim that midterm seat losses are based largely on vote defections by partisans of the losing presidential party rather than on a turnout differential (with the winning presidential party's partisans turning out in higher numbers in the on-year race) does not stand up. Contrary to their estimates (113), an analysis of individual level NES data (corrected for the actual vote divisions) indicates that party identifiers of the party winning the presidency are a significantly larger portion of presidential than midterm electorates. Republicans comprised about four percentage points more of the 1984 and 1988 on-year electorates than of the 1986 and 1990 midterm electorates, and Democrats made up about three percentage points more of the 1992 electorate when they won the White House than of the 1994 midterm electorate. As with the ticket-splitting analysis, one wonders why Burden and Kimball prefer the complicated guesswork of ecologically inferred data to more direct and reliable measurements. Similarly, defection rates for on-year and midterm electorates from 1956 to 1994 (corrected to the actual national vote and using current party identification to ensure comparability of defection rates across election years) are
These reservations aside, Burden and Kimball's principal observations are well taken: ticket splitting is caused to a great extent by the uncompetitiveness of local congressional politics and not by maneuverings to achieve moderation through policy balancing.