President George H. W. Bush famously employed a veto strategy in the 101st and 102nd Congress to project strength and defend "his party and his powers" (Mullins and Wildavsky 1992, 36). This strategy served President Bush rather well until he suffered a serious defeat at the end of his time in office. On September 17, 1992, the House of Representatives adopted the conference report on S. 12, the Cable Television Consumer Protection and Competition Act of 1992, by a margin of 280-128 (CQ Almanac 1992). This supposedly "veto-proof' margin barely surpassed the two-thirds supermajority needed to override the president's veto. President Bush vetoed the bill on October 3, 1992. Bush's veto set up a highly public confrontation between Congress and the president in the midst of his reelection campaign. The White House "stressed loyalty to the president as the main reason why Republicans should switch" their votes on the cable bill and vote to sustain the president's veto (CQ Almanac 1992). In the end, the White House lost this argument as the House overrode the president's veto by a margin of 308-114. On perhaps the highest profile veto of Bush's presidency, zero members of the House switched their votes to support his position while 38 members either switched their votes to defect from the president's coalition or abstained on the final passage vote and then voted against the president during the veto override attempt. This crushing legislative defeat helped cripple Bush's already weak campaign for reelection. President Bush lost both the cable battle in October and the electoral war in November.
Political science currently offers two primary explanations for member voting behavior on presidential agenda items. The two dominant arguments assert that members respond to the president based on (1) party (Bond and Fleisher 1980, 1984, and 1990; Edwards 1980, 1989), (2) ideology (Bond and Fleisher 1990; Krehbiel 1998; Poole and Rosenthal 1997). Some scholars combine party and ideology to argue that American politics takes place in a "four-party" system composed of liberal Democrats, conservative Democrats, liberal Republicans, and conservative Republicans (Burns 1963; Bond and Fleisher 1990; Fleisher and Bond 1996). Neither party nor ideology fully accounts for member behavior on the Cable Television Consumer Protection and Competition Act of 1992. First, Democrats held 270 seats in the 102nd Congress (61%), and therefore needed Republican assistance to override President Bush's veto. Second, ideological explanations do not accurately describe member behavior on S. 12 in 1992. Of the 38 members who changed their votes to defect from President Bush's coalition or abstained on the final passage vote and then voted for S. 12 during the veto override attempt, only four come from the veto pivot quintile. In fact, the members who changed their votes on S. 12 came from all over the ideological spectrum as captured by NOMINATE scores. In this article, I argue that to more accurately predict and explain member behavior on veto override attempts, political scientists must consider how constituencies and electoral incentives affect members' votes.
The high stakes politics surrounding veto override attempts offer an ideal opportunity to study the president's influence in Congress. A bill's content is identical on both final passage and override votes, yet some members of Congress vote differently on these two roll-calls. Cameron (2000) examines veto power at the macro-level and finds that the veto's existence as an institutional rule can cause Congress to strategically modify legislation both in anticipation of a potential veto and after the president issues a veto. I build on these findings by looking at the micro-level and examining the behavior of the individual members of Congress, whose support the president needs to successfully veto legislation.
Krehbiel (1998, 163) is one of few to study the veto's effect at the individual level. He finds that "postwar presidents do seem to exert a positive influence over individual legislators in the veto arena." He divides members of Congress into quintiles based on their NOMINATE scores and focuses on legislators in the veto pivot quintile. Krehbiel finds that, on average, presidents can gain one vote of every eight from these legislators. My analysis goes beyond the focus on ideology and "pivotal politics" alone to investigate how constituencies and electoral incentives influence legislators to switch their votes on veto override attempts.
The pivotal politics argument suggests that about 12% (Krehbiel's one out of every eight) of the members within the veto pivot quintile switch their votes to join the president's coalition on veto override attempts. I build upon this finding with an argument that explains and predicts which members inside the veto pivot will be among the 12% who join the president's coalition. Further, many legislators outside of the veto pivot quintile change their voting behavior as well. What motivates these members? It cannot be simply ideological considerations. Something else must be at work. I attempt to offer a more comprehensive explanation of vote switching during veto override attempts through the development and testing of two hypotheses based on constituency and electoral factors: the Constituency Hypothesis and the Vulnerability Hypothesis. I find that the president's ability to win members' support on override attempts is affected by party, ideology, members' electoral vulnerability, and the president's strength in members' constituencies. Presidential strength, which measures the president's relative strength in the constituency of each member of Congress, is strongly related to members' decisions to switch their votes on veto override attempts.
Why Go Beyond Pivotal Politics?
In Pivotal Politics, Krehbiel (1998) develops an elegant model to describe how enacting coalitions form. This model argues that members at and around the veto pivot are critical because legislative coalitions must include these members to override the president's veto. Krehbiel divides members into quintiles based on their NOMINATE scores and investigates whether members in some quintiles behave differently than others on votes in a veto chain. His analyses demonstrate that members in the veto pivot change their votes on veto override attempts more often than other members. The pivotal politics model is widely accepted today due to its elegance, rigorous testing, and great predictive power.
While the pivotal politics model admirably explains the legislative process using a parsimonious set of assumptions, it neglects some important aspects in the formation and composition of legislative coalitions. Relying solely on the arguments of Pivotal Politics, one is unable to explain a considerable portion of member voting behavior on override attempts. At the aggregate level in both chambers of Congress, members outside of the veto pivot quintile comprise the majority of members who switch their voting behavior on override attempts.
Figure 1 takes all members of the House who switched their votes between final passage and the veto override during the time period of 1973 to 2011 and displays the relative percentage of members who come from each ideological quintile. Members from the veto pivot quintile make up just 26.9% of all House members who switch their votes on override attempts. In fact, slightly more members (27.2%) from what Krehbiel calls the extreme opponents quintile change their votes. Focusing solely on the veto pivot quintile in the House would cause one to miss nearly three-quarters of the members who switch their vote! Excluding these members seriously undermines attempts to explain vote switching on veto overrides. This result holds true when I disaggregate the data and look separately at members who switch votes to either join, or defect from, the president's coalition on veto override attempts.
An equal percentage of members from the veto pivot quintile and the extreme opponents quintile change their votes to join the president's legislative coalition on the override attempt. Members from each quintile comprise 40.8% of presidential joiners (see Figure 1). Furthermore, members from the veto pivot quintile make up just 22% of all members who defect from the president's coalition. These findings suggest that political scientists must incorporate more than ideological proximity in their attempts to explain vote switching on veto overrides.
Ideology clearly affects members' decisions to switch votes on veto override attempts. Members from the veto pivot quintile do comprise a disproportionate share of those who switch their votes on override attempts, especially those who switch their votes to join the president's coalition. At the same time, there appears to be much more at work than ideology alone. In the aggregate, members from the veto pivot quintile do not comprise a majority of the members who switch their votes between final passage and the veto override attempt. (1) In fact, members from the veto pivot quintile do not even comprise a clear plurality of presidential joiners or presidential defectors in the aggregate. A more complete explanation of member behavior on veto override attempts is needed for two reasons. First, members from the veto pivot quintile are not a majority of the members who switch their votes on veto override attempts. Second, not all members in the veto pivot quintile change their voting behavior on override attempts.
Partisan differences help explain why ideology does not capture all of the variance in member behavior on override attempts. Almost 20% of presidential party members become "presidential joiners" by voting for the bill on final passage and then switching votes to oppose the bill on the override attempt (see Table 1). Less than 2% of opposition party members switch votes to become presidential joiners. On the other...