Predicting Presidential Appearances During Midterm Elections:
The President and House Candidates, 1982-2010
Rob Mellen Jr.
Mississippi State University
Kathleen Searles
Augusta State University
Abstract
Due to limitations in both time and resources, presidents who wish to assist their co-partisans electoral endeavors must make strategic choices when offering their assistance. Much research has attempted to explain why presidents devote their limited resources during a midterm election, yet we know little about the factors that lead to a presidential visit to a particular congressional district. Our research addresses this gap in the literature by narrowing the focus to the congressional district level. We ask: are the same factors that lead to a presidential visit at the state level operational at the district level? The results suggest they are not. Moreover, we find that while presidents do indeed behave rationally when they make appearances for their co-partisans: visits are more likely to occur when there are multiple higher-level competitive races in a district, and presidents are more likely to go where they are already popular.
Predicting Presidential Appearances During Midterm Elections:
The President and House Candidates, 1982-2010
Thomas Jefferson once said, “Traveling makes a man wiser, but less happy.” Unfortunately, for the happiness of the modern president, traveling is part of the job. Whether visiting a foreign nation or a small town in Ohio, modern presidents spend a lot of time outside of Washington, D.C. However,what factors do presidents consider when choosing where to go when they leave Washington? Presidents, we assume, are rational actors who wish to maximize the return on the investment of their limited resources (Downs, 1957). Scholars differ on what strategic return presidential travel is designed to accomplish. Some scholars argue that presidential travel is self-centered and designed to further the president’s own reelection efforts (Eshbaugh-Soha Nicholson-Crotty, 2003) while others argue it is designed to maximize the number of his co-partisans in Congress(Cohen, Hamman, Krassa, 1991; Hoddie Routh, 2004). We do not view these goals as mutually exclusive, though our own work is theoretically closer to the latter approach. Before addressing the question of what factors presidents consider when deciding where to travel it will be helpful to review each of the theoretical arguments stated above. We begin with a look at the self-centered approach.
Presidential Travel and the Permanent Campaign
As noted previously, some scholars studying presidential travel suggest it is designed to further the president’s own reelection effort and build public support at the state level (Cohen Powell, 2005; Eshbaugh-Soha Nicholson-Crotty, 2003, 2009). Presidential travel then becomes part of what has been labeled the ‘permanent campaign’ (Blumenthal, 1982; Ornstein, 2000),orthe idea that everything a president does from the time he decides to seek the office until the day he leaves office is a part of the campaign. In this vein, Eshbaugh-Soha and Nicholson-Crotty (2003) argue that presidential travel during the first term should be viewed as a ‘dress rehearsal’ in anticipation of the next presidential election. If so, presidents will travel tocompetitive states with many electoral votes, although this logic only applies to a first presidential term due to the 22nd amendment limiting presidents to two terms in office. Indeed, Charnock, McCann, and Tenpas (2009) examined first-term presidential travel from the Eisenhower Administration to the George W. Bush Administration and found that presidents increasingly allocated their trips to large competitive states while neglecting smaller states with fewer electoral votes. Doherty (2007) agrees but suggests that presidential travel has different purposes at different times during a presidential administration. One such purpose may be ‘going local’, which features the presidenttaking his message directly to the people in an effort to create public pressure on Congress to enact the president’s legislative program. President George W. Bush attempted to do this in 2005 with his plan for Social Security privatization, albeit unsuccessfully (Eshbaugh-Soha Peake, 2006). Whether a president is building support for his reelection, his legislative program, or raising funds for his party, each of these scholars suggest that presidential travel isstrategic and self-centered.
Maximizing Co-Partisans
A second theoretical approach to the study of presidential travel suggests that the president’s strategy is to maximize the number of co-partisans in the next Congress. After all, more co-partisans in Congress should enhance the prospect of legislative success for the president’s agenda (Bond Fleisher, 1990; Edwards, 1990). Presidents may seek to increase the number of their co-partisans in Congress by assisting their electoral efforts through use ofthe ‘bully pulpit’ and the ‘power to persuade’ (Neustadt,1990). One way a president may do this is through direct interaction with citizens at public appearances with candidates for the U.S. Senate, U.S. House of Representatives, or state governor (Sellers Denton, 2006). However, given scheduling constraints, the president cannot possibly travel everywhere. Thus, most researchers argue that presidential trips are strategic in their timing, location, and type of visit (Doherty, 2007, 2010). Doherty finds that presidents frequently travel as their party’s ‘fundraiser in chief’ to aida co-partisan’s effortto raise cash for a congressional campaign. Herrnson and Morris (2007) suggest that President George W. Bush’s fundraising trips on behalf of Republican congressional candidates helps to explain the historic Republican victories in the 2002 midterm congressional election. That brings us to the focal point of our own research.
Presidential Travel During Midterm Elections
Some researchers studyingpresidential travel focus on travel specificallyduring the midterm election year, as do we. Lang,Rottinghaus,and Peters (2011) codify presidential travel during midterm election years into three categories: 1) fundraisers; 2) campaign rallies and; 3) ‘virtual’ visits. We focus our attention on the second of these categories, campaign rallies, which take place during the roughly two-month period prior to a midterm election. The literature in this area dividesroughly into one of two themes: the effects of midterm presidential travel (Cohen, Hamman, Krassa, 1991; Herrnson Morris, 2007; Keele, Fogarty, Stimson, 2004; Vaughn, 2004) or the predictors of midterm travel (Eshbaugh-Soha Nicholson-Crotty, 2009; Hoddie Routh, 2004).Our work falls into the latter category but it is helpful to review the theoretical foundations of the former category first.
The Effects of Midterm Presidential Travel
Scholars focusing on presidential efforts to assist their co-partisans disagree about whether or not such efforts actually have the intended effect. Cohen et al. (1991) examined presidential efforts on behalf of Senate candidates from 1966-1986 and found that presidential visits may help a candidate’s fundraising effortsand exert an influence on a candidate’s vote share under specific conditions.[1] The authors argue that the effects of a presidential visit are likely to be most pronounced in the most competitive races. In such races, Cohen and colleagues argue, a campaign appearance by the president may convert a loss to a victory for the president’s co-partisan. Similarly, Keele et al. (2004) examine presidential travel in 2002 with the goal of determining whether President Bush’s efforts on behalf of his co-partisans led to just such effects. The authors suggest that presidential visits were not all that important in 2002 as President Bush spent most of his time in states where he was already quite popular and his co-partisans were likely to win anyway.In fact, the authors conclude that Republicans whom Bush visited underperformed on the two-party vote share when compared with Republicans who did not receive a visit. Herrnson and Morris (2007) also examine President Bush’s efforts in 2002 and reach the opposite conclusion, though their data includes fundraising visits as well as rally appearances.
The Predictors of Presidential Midterm Travel
The second theme found in the midterm visits literature moves beyond the question of what effect presidential travel has on the fortunes of presidential co-partisans by examining the motivation for presidential visits. Vaughn (2004) looks at whether the president ‘rewards’ a senator who supports his legislative agenda with a visit or ‘punishes’ a senator who fails to support him by withholding a visit. Neither hypothesis is supported by the data, though with only six observations generalizations are very difficult. More recently, Bryan and McClurg (2009) examine President Bush’s campaign visits on behalf of Senatecandidates in 2002 and 2006. They posit that presidential appearances are likely to be a function of several variables, including previous legislative support, competitiveness of the race, incumbency, and presidential popularity (nationally and within the particular state in the most recent presidential election). The authors find, as expected, Bush’s visits in both 2002 and 2006 were tilted toward the most competitive races and that he was more likely to appear for an incumbent than for a challenger. The authors also suggest that previous legislative support played little role in determining which states Bush visited in 2002 but the opposite was true in 2006. Essentially, Bush ventured into unfriendly territory more often when he was popular nationally than he did once his national popularity diminished. Finally, Hoddie and Routh (2004) engage in a state-level analysis of presidential visits from 1954 to 1998 and find that presidential travel in midterm congressional elections is strategically designed with a focus on states with the highest combination of competitive races for Senate, House, and Governor. In direct contrast to the permanent campaign literature, the authors suggest that the president’s primary concern appears to be securing the election of as many of his co-partisans as possible to individual statehouses and the U.S. Congress.
The Congressional District Level and Redefining Competitive
Each of the preceding analyses provides us with some interesting insights about how and where presidents devote their limited resources during a midterm election. Yet, they are not designed to tell us much about the criteria a president may consider when making the decision to campaign in a specific congressional district. Our research builds upon the knowledge produced by previous studies and allows us to address this gap in the literature by narrowing the focus to the congressional district level. Are the same factors that lead to a presidential visit at the state level operational at the district level?
Specifically, we examine presidential rally appearances for co-partisans at the congressional district level. We define a rally appearance as one in which the president makes a public appearance and delivers a speech on behalf of a co-partisan seeking election to the House of Representatives. We only include rally appearances specifically labeled as an appearance on behalf of a specific candidate by the Public Papers of the Presidents of the United States as catalogued by the American Presidency Project (Wooley Peters, 2012). Our search identified 88 such appearances by presidents between 1982 and 2010.
The primary question we seek to answer is why does a president appear at a campaign rally in one district but not in another district? To make this determination we advance a series of district-centered hypotheses. Like several of the preceding studies, we expect race competitiveness to be a key factor. However, these studiesuse a statewide measure that defines competitiveness using a post-election standard employed by Mayhew (1974), which labels a race as competitive if the winner received fifty-five percent or less of the popular vote in the previous election. We take issue with that definition because declaring a race competitive ex post facto may obfuscate our ability to predict presidential campaign appearances by including races that became competitive due to presidential involvement. It may also lead to the exclusion of races that turned out to be non-competitive due to presidential involvement. As such, we operationalize competitiveness using CQ Weekly’s pre-election forecast for each midterm election included in this paper.[2] The forecasts rely on an independent analysis of each race, including district level polling and insider information. Since a president makes his decision to make an appearance before the outcome of the race is known, we think this standard makes more sense than one which relies on the results of the election (or the one prior) to label the race as competitive.
A District Driven Approach
Congressional districts vary greatly in their degree of safety for one party or the other. Infact, in most congressional midterm elections only about ten percent of the seats in the House are considered competitive. The rest either strongly favor one party or the other, or are considered safe for one of the parties. Safe districts account for well over half of the 435 seats in the House of Representatives each election cycle. Given the relative safety of most seats, we expect that presidents will focus their midterm efforts on those districts most likely to be competitive in the upcoming midterm election. These districts hold out the greatest prospect for a return on the investment of time and energy by a president as he gains little, if anything, by scheduling an appearance in a district his party is already likely to win (or lose). It is the uncertainty of winning (or losing) that drives a president to appear in a competitive congressional district. Stated as a hypothesis:
H1: Presidential campaign appearances during midterm elections are more likely to be allocated to competitive districts than to districts that are not competitive.
Though we expect to find that presidents choose to appear in competitive rather than uncompetitive races, not all competitive races are likely to draw the attention of a president. Congressional races may be competitive for a variety of reasons ranging from candidate quality, a scandal involving a current member of Congress, or even a huge investment in campaign spending by a non-party organization. None of these factors may have anything to do with the president. Yet, the president will have either won or lost the popular vote for president in the district two years earlier. We expect presidents to prefer campaign appearances in friendly territory. This preference will shape the universe of districts that a president is willing to visit. Thus:
H2: Presidential visits during a midterm election are more likely to occur in competitive districts in which the president is more popular than in competitive districts in which the president’s opponent is more popular.
Presidential Visits and Open Seats
So far, weexpectpresidential visits to be concentrated in congressional districts that are both competitive and were won by the president in the last election. There are still likely to be many more districts than a president is able to visit during the course of a single election campaign so it is necessary to identify other factors that influence a president’s strategic choice to travel. One potential criterion is for presidents to allocatevisits to competitive districts with an open seat. One reason a president might choose to campaign in an open seat district rather than for an incumbent is the incumbency advantage identified by David Mayhew (1974). Incumbents enjoy exceptionally high reelection rates, even in down years for their party. In any given midterm, 90-95% of incumbents seeking reelection are successful, though incumbents in competitive districts enjoy a slightly lower success rate. Even in the “wave” elections of 1994 and 2006, in excess of 90% of incumbents seeking reelection to the House were successful. Open seat districts tend to be much more competitive than races featuring an incumbent seeking reelection (Jacobson, 1987). Rather, candidates seeking election to open seats do not enjoy the incumbency advantages a sitting member of Congress does, resulting in a greater likelihood of a more competitive race, and the possibility of the seat switching parties. Since a president needs support in Congress from likeminded members, it seems logical that a president would direct his campaign efforts during a midterm election towards those districts most vulnerable for his party, those with an open seat. Perhaps an appearance by the president can provide a co-partisan seeking election in an open seat district with some of the advantages typically enjoyed by incumbents, such as name recognition and superior fundraising capability. We hypothesize that:
H3: Presidential visits are more likely to occur in a competitive district with an open seat than in a competitive district in which an incumbent is seeking re-election.
Senatorial and Gubernatorial Contests
An additional factor that might contribute to a president’s strategic decision to visit is the presence of one or more higher-level races on the same ballot. Competitive congressional districts in states that also have a competitive Senate or gubernatorial contest, may be more likely to draw the interest of the president than a single competitive house race might. Stated as a series of hypotheses:
H4: A president is more likely to visit a competitive district featuring a competitive Senate contest than a competitive district that does not have a competitive Senate contest.
H5: A president is more likely to visit a competitive district that has a competitive gubernatorial contest than a competitive district that does not have a competitive gubernatorial contest.
Data and Methods
Variables
The dependent variable is a presidential visit to the district in a midterm election year. To test the hypotheses in this paper we examined CQ Weekly’s midterm election forecasts for each midterm election between 1982 and 2010and recorded their rating for each congressional district[3]. CQ Weekly rates each district contest as either a toss-up, leaning Democratic, leaning Republican, likely Democrat or Republican, or safe for the party currently holding that seat. CQ operationalizes race ratings based on several independent criteria including the strength of the candidates, partisan makeup of the district population, district level polling, and expert evaluation of the race. We coded a race as competitive if CQ classified the race as a toss-up or leaning to one of the parties. This coding produced a data set containing 422races ranked as competitive and 3058 races ranked as uncompetitive. Over the 28-year period covered in this study, approximately 12.4% of all races for the United States House of Representatives were rated as competitive. In line with our first hypothesis (H1), the primary independent variable of interest is competitiveness of the congressional district, which we coded ‘1’ if leaning towards the opposing party, ‘2’ if a toss-up, and ‘3’ if leaning towards the president’s party. Category ‘1’, leaning towards the opposing party, is omitted as the baseline category. This independent variable improves upon previous work that uses binary competitive variables, resulting in a loss of specificity (Cohen et al., 1991). Additionally, the competitive independent variable lends nuance to our understanding of strategic presidential visit allocation: rather than examine a safe seat versus an unsafe seat, we are able to investigate how gradations of competitiveness might affect a president’s strategic decision to visit.[4] To capture the effects of popularity of the president in the district (H2), following in the footsteps of Cohen et al. (1991), we include a measure of the difference between presidential vote share and opposition vote share in the previous presidential election. Additional independent variables coded for each congressional district included whether or notthe incumbent in the district sought reelection (H3), and whether or not the state featured a competitive race for the United States Senate (H4), state governor (H5), or both. We relieduponthe same CQ Weekly pre-election forecasts to determine whether a Senate or gubernatorial race was competitive as we did for categorizing the competitiveness of House races.