Chapter 11 Congressional Elections Why Some Incumbent Candidates Lose Copyright © 2011. Routledge. All rights reserved. Jamie L. Carson and Carrie P. Eaves In 2008, Joe Knollenberg, the Republican representative from the 9th district of Michigan, lost his bid for a ninth term in the U.S. House. In the previous election Knollenberg had barely defeated his opponent in the general election, winning just 52 percent of the vote. As a result, the Democratic Congressional Campaign Committee targeted him as a vulnerable incumbent in 2008. The Democrats nominated a well-­qualified candidate, Gary Peters, who had served in the Michigan State Senate until he was forced to retire because of state term limits and subsequently went on to serve as the State Lottery Commissioner. Congressman Knollenberg outraised his opponent by more than $1.6 million, yet he still lost his re-­election bid. Similarly, Joe Porter, a Republican Congressman from the Nevada 3rd district, was defeated on Election Day. Porter had served in Congress for six years, yet in 2006 he barely won re-­election by a margin of 48 to 47 percent. As a result, he was also targeted by Democrats and faced stiff competition from the minority leader of the State Senate Dina Titus. Despite a fairly substantial fundraising advantage, Porter was defeated in 2008. Incumbents in the U.S. House of Representatives regularly win re-­election at a rate well over 90 percent (Jacobson 2009). As a result, pundits and political scientists often bemoan the lack of significant competition in congressional elections. Yet, in the two cases described above (as well as in 17 other House races in 2008), the incumbent failed to win re-­election, despite numerous “advantages” associated with the office. More generally, a handful of House and Senate incumbents invariably lose their seats in Congress across individual election cycles. Examining specific trends in these House and Senate elections in which incumbents are defeated could provide vital insight into the declining rates of electoral competition as well as when the rare competitive election might actually emerge. In this chapter, we assess the current state of research on congressional elections, most notably by highlighting the declining rates of electoral competition in House and Senate races over time. In light of the exponential growth in the amount of scholarship on congressional elections during the past few decades, it would be impossible to thoroughly examine all facets of New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. 184   Jamie L. Carson and Carrie P. Eaves this research in a single essay. Instead, we focus on a central puzzle in congressional elections research: why some incumbents lose each election cycle despite the inherent advantages accruing to incumbents. We begin our discussion by addressing one element beyond the direct control of individual candidates—the electoral context. Next, we point to three areas of research that offer clues about why electoral competition is declining as well as when the rare competitive election might occur: the actions of incumbents, the role of challengers, and the impact of money in elections. Finally, we conclude our discussion with specific suggestions about new directions for congressional elections research. Copyright © 2011. Routledge. All rights reserved. Overview of Congressional Elections Research Since the early 1970s, scholars have devoted a significant amount of attention to the topic of congressional elections research (and House elections in particular). To a certain extent, this emphasis grew out of Mayhew’s (1974a) classic observation that we can think of members of Congress as “single-­ minded seekers of re-­election.” Building on this valuable insight, students of congressional politics have spent countless hours investigating a wide variety of important and unanswered questions. For instance, why are incumbents re-­elected at extraordinarily high rates in House (and, to a lesser extent, Senate) elections? Why do some qualified candidates decide to challenge incumbents while many others choose not to? To what extent are the staggering costs of congressional campaigns a contributing factor to incumbency success rates? These and many other similar questions have motivated some of the more interesting examples of cutting-­edge research on congressional elections during the past few decades. Although much has been learned from this research on congressional elections, there are still a number of unanswered questions that, in our view, have received insufficient attention to date. Take, for instance, Figure 11.1, which has often served as the foundation for many of the inquiries pertaining to congressional elections research. Based on the evidence in the figure, a number of discernible patterns are clear. First, during the past 60 years, a growing proportion of incumbents who have sought re-­election in both chambers have been re-­elected. Second, the incumbency re-­election rate in the Senate has been much more variable compared to the House. Third, there is noticeable volatility in the re-­election rates of congressional incumbents, with more noticeable declines in midterm election years. Despite these recognizable trends in House and Senate elections, there is still uncertainty about what explicitly is contributing to each. Another way to think about the lack of electoral competition depicted in Figure 11.1 is to examine the total number of marginal races in these elections. In Figure 11.2, for instance, we plot the total number of U.S. House races from 1948 to 2008 where the incumbent received between 50 and 55 New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Why Some Incumbents Lose Elections   185 100 Percentage reelected 90 80 70 60 50 House Senate 40 1948 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008 Year Figure 11.1 Success rates of incumbents seeking re-election, 1948–2008. 70 60 Number of races Copyright © 2011. Routledge. All rights reserved. percent of the two-­party vote (the traditional definition of marginality). Although there are some noticeable spikes in the number of marginal races in election years such as 1956, 1964, 1974, and 1994, overall the pattern shows a fairly visible decline in marginal races over time. In other words, an increasing number of incumbents appear to be winning by larger margins against their opponents in the past few decades. Mayhew (1974b) first noticed this pattern in discussing the “vanishing marginals” in his classic article and the trend appears to have continued in recent decades, with only a modest 50 40 30 20 10 0 1948 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008 Year Figure 11.2 Total number of marginal U.S. House races, 1948–2008. New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. 186   Jamie L. Carson and Carrie P. Eaves increase in the two most recent elections. Ultimately, the pattern in Figure 11.2 suggests that the majority of House incumbents are far less vulnerable today than they were 40 or 50 years ago. In an attempt to better understand why elections have become less competitive over time, we shift our focus away from the question of why so many incumbents win each year to the equally interesting question of why some incumbents lose each election cycle. By doing this, we hope to offer new insights on specific factors contributing to declining levels of electoral competition in congressional races. Along the way, we also hope to offer differing explanations for incumbency defeats across the House and Senate given the greater volatility associated with Senate elections. We begin our discussion with several contextual factors that influence congressional elections. From there, we shift our attention to the actions of incumbents, the role of challengers, and the impact of money in determining candidate entry patterns and outcomes in House and Senate races. Copyright © 2011. Routledge. All rights reserved. Contextual Factors Although ambition and personal considerations play an important role in determining the level of individual competition in congressional races, macro-­level conditions may be equally important as well. Consider, for instance, national conditions as reflected by the overall state of the economy or levels of presidential approval. If the economy is perceived to be doing well or the incumbent president’s approval levels are high, members of the out-­party may have a much harder time justifying a decision to run for elected office. As such, incumbents may fare better in an upcoming election due to the lower probability of facing an experienced challenger. When national conditions are less favorable, however, incumbents have reason to be concerned about their electoral viability.1 In these circumstances, more qualified candidates are likely to emerge and often have an easier time raising money, which places a greater number of sitting incumbents in increased electoral jeopardy. In addition to national conditions, the nature of each individual congressional race can influence the outcome and level of competition. Open seat elections that lack an incumbent candidate are much more competitive than the typical incumbent-­contested race (Gaddie and Bullock 2000; Jacobson 2009). Indeed, one could not discuss electoral competition without discussing open seat races. For the most part, races without an incumbent are rare in electoral politics—especially in the Senate. From 1980 to the present, the average number of open seats races in House elections was 43 (out of a possible 435). In 1984 and 1988 there were a mere 27 open seat races; however, in 1992 there were 91 open seat races. Since 2000, the number of open seats on average has declined even further to less than 30 per individual election. Without an incumbent in the race, experienced candidates are more likely to New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Copyright © 2011. Routledge. All rights reserved. Why Some Incumbents Lose Elections   187 emerge since they feel that this may be their best chance to win an election (Banks and Kiewiet 1989; Jacobson 2009). A third contextual factor that may influence aggregate patterns of candidate competition is the timing of a congressional election. Voter turnout fluctuates widely depending on whether or not it is a presidential election year. Midterm elections always lead to decreased voter turnout following the “surge” during the previous presidential election. Without the president at the top of the ballot, political interest among voters wanes. In addition, historically the president’s party does far worse in midterm election years.2 As a result, levels of electoral competition may be affected. Building on the earlier work of Angus Campbell (1960), James Campbell (1991) proposed a general theory of “surge and decline” that seeks to explain how many congressional seats the party of the president loses at the midterm election. According to the theory, the more seats the party of the president picks up during a presidential election, the more seats are at risk during a midterm election. Without the president’s name on the ballot at the midterm, individual congressional candidates become more affected by local conditions, especially if the economy or national conditions have declined in the months preceding the election. Thus, the number of seats the president’s party loses at the midterm (the decline) is a function of the number of seats gained in the previous presidential election (the surge).3 Two recent midterm elections—1998 and 2002—have challenged the conventional view associated with the surge-­and-decline hypothesis. Rather than losing seats at each midterm, the president’s party actually gained a few additional seats in each of these midterm elections. With the midterm losses for Republicans in 2006, it is unclear at the time of writing whether the traditional pattern of surge and decline for the president’s party will continue in the 2010 midterm elections or how severe it might be for Democratic candidates. To a considerable extent, this will depend upon President Obama’s ability to deliver on many of his campaign promises as reflected by his overall approval level, the state of the economy prior to the midterm elections, and the number of experienced Republican challengers willing to run against incumbent Democrats in 2010 (Jacobson 2009). With a congressional redistricting cycle looming on the horizon, the 2010 midterm elections could be the last hurdle the Democrats face to maintaining their majority status in the House for many years to come. The Actions of Incumbents For the past 40 years, students of congressional elections have documented a fairly stable pattern in the context of House and Senate elections: the relative ease with which incumbents seem to get re-­elected. Beginning with the work of scholars who first recognized the underlying advantages of incumbency (see, e.g., Erikson 1971; Mayhew 1974b; Cover 1977; Ferejohn 1977; Fiorina New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Copyright © 2011. Routledge. All rights reserved. 188   Jamie L. Carson and Carrie P. Eaves 1977) and continuing with more recent and innovative attempts to estimate the extent of that advantage in congressional elections (see, e.g., Gelman and King 1990; Cox and Katz 1996, 1999, 2002; Erikson and Palfrey 1998; Ansolabehere, Snyder, and Stewart 2000; Brady, Cogan and Fiorina 2000), social scientists have widely examined both the causes and consequences of this recurring phenomenon. Although there is little doubt that incumbents have a substantial advantage over non-­incumbents in congressional elections, there is considerably less agreement over the sources of the incumbency advantage and the causes of its growth over time. This is unfortunate, since elections go to the very heart of accountability and representation, and an insufficient understanding of how they are conducted as well as their direct and indirect effects limits our ability to generalize from them. Why is the incumbency advantage so conceptually important to our understanding of electoral politics? The simple answer is that public-­ spiritedness and good intensions aside, once a person wins public office, her goal rapidly becomes staying in office. As classically stated by Mayhew (1974a), legislators value re-­election above all other goals. Individuals may enter public office with altruistic intensions, but their proximate goal quickly becomes re-­election. To accomplish their other goals they must be re-­elected, and incumbency is their key advantage (Alford and Brady 1989). To some, the incumbency advantage is indicative of an electorate that is, by and large, content with the performance of its elected representatives. To many others, however, it raises potential concerns about the issue of democratic account­ ability, often resulting in increased demands for campaign finance reform or enactment of term limits for incumbents—issues that have received a widespread amount of attention since the early 1990s. On the whole, concern about democratic accountability has sparked an interest among students of congressional elections to investigate the extent to which legislative behavior has discernible electoral consequences. While some scholars have focused on the electoral rewards associated with distributive benefits (see, e.g., Bickers and Stein 1994, 1996; Alvarez and Saving 1997; Sellers 1997), others have examined the potential advantages linked to casework (Alford and Hibbing 1981; Serra and Moon 1994) and incumbents’ roll call voting records (Erikson 1971; Wright 1978; Jacobson 1993). Although far from conclusive, these studies have offered evidence that incumbents typically receive marginal benefits from these and other legislative activities. This, of course, has been a source of concern among journalists, pundits, and scholars alike who question whether potential advantages that accrue to incumbents make it impractical for quality challengers to mount an effective campaign. After all, how can we expect legislators to be responsive to their constituents if incumbents have little reason to believe that they will be defeated in the upcoming election? To be sure, congressional incumbents spend a considerable amount of time engaging in activities that they anticipate will cultivate a sense of loyalty New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Copyright © 2011. Routledge. All rights reserved. Why Some Incumbents Lose Elections   189 with their constituents and improve their chances of re-­election. Indeed, Mayhew (1974a) finds that legislators split their time between advertising, credit claiming, and position taking in an effort to facilitate their re-­election goal.4 Fenno (1978) observes that incumbents cultivate a personal “home style” with their constituents in an attempt to foster a sense of trust and loyalty among prospective voters. Fiorina (1989) argues that Congress is responsible for the creation of a Washington establishment that allows individual legislators to frequently engage in pork barrel and casework activities, which brings them favor with their constituents. Incumbents also raise large sums of money with the expectation that it will enhance their prospects for re-­election by discouraging the emergence of experienced, prospective challengers (Abramowitz 1989; Box-­Steffensmeier 1996). As the preceding studies suggest, there is little doubt that legislators believe that their actions and behavior impact upon their electoral fortunes. Given that voters directly elect legislators, most (if not all) recognize the importance and necessity of being responsive to their constituents. Since they also possess a finite amount of time and scarce resources, we should not expect them to act frivolously with regard to the allocation of these limited resources. Moreover, the extent to which we evaluate an incumbent as “marginal” or “safe” is largely contingent upon the activities that the legislator engages in while in office. Incumbents who appear “out of touch” with their constituencies will generally have a much harder time getting re-­elected than those legislators who attempt to maintain a connection with their voters. As Mayhew (1974a: 37) explains, “what characterizes ‘safe’ representatives is not that they are beyond electoral reach, but that their efforts are very likely to bring them uninterrupted electoral success.” In light of the numerous advantages accruing to incumbents, why are some incumbents routinely defeated at the polls? Although there is little consensus in response to this question, there is some evidence to suggest that incumbents may not always be responsive enough to their constituents. As noted above, legislators strongly believe that their actions in office assist them in their re-­election efforts. As such, if they fail to provide distributive benefits to their districts in the form of casework or pork barrel projects, there is the underlying perception that they are more likely to be defeated (Serra and Moon 1994; Sellers 1997). In addition, incumbents who focus more on goals such as policy or power within the chamber (see, e.g., Fenno 1973) may undermine their own chances of re-­election if they forget what got them elected in the first place. Another factor that may contribute to greater incumbency defeats is increased loyalty to one’s political party. In any given Congress, numerous controversial and divisive issues come before the House and Senate chamber, often requiring individual legislators to express positions on legislation that may be unpopular with a large segment of their constituencies. If members of Congress represent moderate constituencies and are perceived as voting too New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. 190   Jamie L. Carson and Carrie P. Eaves often with their party on controversial legislation, then there is an increased chance that their voters will choose to vote for someone else in an upcoming election (Carson et al. 2010). One recent example of this is the health care reform bill that passed the House on March 22, 2010. Many “Blue Dog” Democrats who represented moderate districts were reluctant to vote “yes” on the legislation given the strong possibility of electoral repercussions in the fall. While it is unclear as of writing what the electoral effects will be on this highly visible vote, it is interesting to note that nearly all of the defections were by moderate members of the Democratic coalition. One final point about incumbency defeats that should be made is that the actual number of losses might in fact be greater if incumbents did not behave strategically while in office. In any given election year, a few vulnerable incumbents decide not to seek re-­election due to their precarious electoral circumstances. Although these strategic retirements are often couched in terms of legislators wanting to “spend more time with their families,” in many respects these decisions are routinely made to avoid risking electoral defeat (Hall and Van Houweling 1995; Carson 2005). Indeed, the fact that incumbents do choose to retire strategically instead of risk the possibility of electoral defeat tends to inflate the extent to which incumbents actually are (or are perceived to be) re-­elected (Cox and Katz 2002). Copyright © 2011. Routledge. All rights reserved. The Role of Challengers In certain respects, focusing on the behavior of incumbents while in office can only tell us so much about candidate competition and electoral outcomes. Beyond legislators serving in the House and Senate, we also need to consider the role of their opponents, specifically in terms of who decides to challenge incumbents in a given election year. Although most of the attention in the incumbency advantage literature during the 1970s focused on the role of the incumbents, this began to change in the next decade. Starting with some early work by Jacobson (1980) and Jacobson and Kernell (1981) regarding strategic politicians, a shift in emphasis in congressional elections research began to occur. In this and much of the subsequent research on congressional elections, attention was given not just to incumbents and their legislative behavior, but to the role of challengers in explaining electoral outcomes. This development represented an important step in the congressional elections literature as scholars since then have regularly considered the effects of candidate experience when evaluating House and Senate elections. Candidate Emergence In shifting attention away from an exclusive emphasis on incumbents, Hinckley (1980) was among the first to observe that scholars of congressional elections were neglecting an important piece of the electoral equation. She argued New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Copyright © 2011. Routledge. All rights reserved. Why Some Incumbents Lose Elections   191 that the incumbency advantage “may be less a matter of what incumbents do right than of what challengers do wrong—or not at all” (442). By only considering the actions of incumbents, scholars had neglected the other key player in congressional elections—the candidate running against an incumbent. Indeed, the great variation among candidates who decide to challenge incumbents could also drive varying levels of competition in congressional elections. To systematically evaluate the impact of candidates, one must first ask, “Who runs for Congress?” Neither campaigning nor serving in Congress are simple tasks. Yet Kazee (1980) found that many candidates run for Congress solely because they enjoy campaigning for office. Subsequently, Kazee (1983) noticed that the presence of an incumbent candidate could serve as a deterrent for potential challengers. In follow-­up studies of ambition and candidate emergence, a number of scholars have opted to employ a case study method examining who decides to emerge in a given congressional district (see, e.g., Fowler and McClure 1989; Kazee 1994). Ultimately, the study of candidate emergence can be especially challenging due to the inherent difficulty in identifying those potential candidates who opted not to run for elective office. In an effort to overcome this potential limitation, Maisel and Stone (1997) first proposed an innovative approach to this question by studying the behavior of state legislators through their Candidate Emergence Project. Climbing up the political ladder of ambition, Maisel and Stone recognized that many career politicians move from state legislatures to the U.S. House. As a result, studying the emergence decisions of state legislators lends itself well to the study of the emergence of congressional candidates. Through their analysis of state legislators, Maisel and Stone find that potential candidates are most strongly influenced by their perceptions of success. Those state legislators who believe that they have a strong chance of winning a seat in the House of Representatives are the ones most likely to emerge as a candidate. More recently, Maestas et al. (2006) show how the costs and benefits of running for and holding elective office can influence candidate entry decisions among state legislators running for the House. For a somewhat different perspective on candidate emergence, Fox and Lawless (2005) examine the lack of female candidates running for and serving in Congress. They evaluate this question by surveying individuals working in careers that are regarded as yielding the highest eligibility pool for political candidates (i.e., law, business, education, and political activism). Fox and Lawless find that women are significantly less likely to demonstrate the appropriate political ambition given the lack of significant role models. Women in their survey were less likely to be recruited to run for office and less likely to consider a candidacy for elective office. Ultimately, their work highlights the need to consider individual-­level factors such as gender differences as well as the national political climate typically studied in analyses of candidate emergence (on this point, see also Lawless and Fox 2005). New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. 192   Jamie L. Carson and Carrie P. Eaves Copyright © 2011. Routledge. All rights reserved. Candidate Quality Not all candidates who decide to challenge incumbents are created equal. Some bring more name recognition, campaign skill, and election experience than others. In their classic study of congressional elections, Jacobson and Kernell (1981) examine aggregate patterns of candidate emergence to identify whether or not challengers behave strategically. Through an analysis of congressional elections in the 1970s, they find that experienced candidates (those who have previously held an elected office) are more likely to run when national and partisan tides favor their candidacy. As a result, Jacobson and Kernell conclude that experienced challengers wait until the conditions are the most advantageous before deciding to emerge (on this point, see also Carson (2005) for comparisons across House and Senate elections). Individual-­level strategic decisions can combine to produce a major swing in the partisan composition of Congress even in the absence of a significant partisan shift in the electorate. As such, strategic decisions by congressional candidates, based on factors such as the perceived likelihood of victory, the apparent value of the seat to them, and the opportunity costs of running for office both reflect and enhance national partisan swings. Consequently, which candidates run, and when, is critical to the interpersonal and partisan dynamics that shape the legislative process in the U.S. Congress. National parties exhibit strategic behavior as well, providing campaign resources such as campaign appearances by party leaders to quality candidates running in competitive elections to increase their prospects for victory. Ultimately, a greater number of experienced candidates running for elective office can mean the difference between victory and defeat for incumbent legislators in vulnerable districts. Jacobson (1989) provides additional support for his initial findings with Kernell using an expanded dataset that includes elections from 1946 to 1986. Although experienced challengers are most likely to emerge when there is an open seat (no incumbent seeking re-­election), there is only a small percentage of open seats in any given election. In addition, if an incumbent appeared particularly vulnerable in their previous race or was involved in a political scandal, then a quality challenger is more likely to emerge. We can see this during the 2008 election when Nancy Boyda (D-­KS) was targeted by the National Republican Congressional Committee as a vulnerable incumbent after winning her previous election by a narrow margin of 51 to 47 percent. As a result, two strong challengers stood in the Republican primary: State Treasurer Lynn Jenkins and former Congressman Jim Ryun. Jenkins emerged as the winner of the congressional primary and was able to narrowly defeat the incumbent. Other scholars have developed alternative measures of candidate quality (see, e.g., Bond, Covington, and Fleisher 1985; Krasno and Green 1988), yet Jacobson’s dichotomous measure has become widely used throughout the New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Copyright © 2011. Routledge. All rights reserved. Why Some Incumbents Lose Elections   193 study of House elections. Certainly the somewhat blunt measure is not without potential flaws, as there are occasionally successful candidates who win elections with no previous elective experience (along with numerous quality candidates who are unable to unseat incumbent legislators). Yet the relative parsimony and ease of replication have made Jacobson’s measure of previous elected experience a reliable measure in the study of House elections. In their study of candidate competition, Banks and Kiewiet (1989) focus on a somewhat different question: given that most incumbents face some type of challenge each year, why do candidates without political experience elect to challenge incumbent members of Congress? Using primary elections data from 1980 to 1984, they find that the motivations of weaker, inexperienced candidates are similar to experienced challengers—both want to win. Weak candidates recognize that they are more likely to gain their party’s nomination when an incumbent is present, than when there is an open seat. Although they will most likely lose in the general election, these candidates can increase their name recognition and electoral skill. If they wait until an incumbent retires, the primary field will most likely be filled with many experienced candidates, making it more difficult for them to receive their party’s nomination. While research on strategic politicians has offered valuable insight into House elections, relatively few scholars have considered the same phenomenon in the context of the Senate. For their part, Senate scholars who have studied candidate quality have been reluctant to adopt the dichotomous measure of previous elective experience. Lublin (1994) studies strategic politicians in the Senate from 1952 to 1990 and finds that members of the House of Representatives are the most successful Senate challengers. As a result, he proposes ranking previous elected experience of Senate candidates with members of the House of Representatives ranked above governors and other statewide officials. One would expect a governor to be a stronger Senate challenger having already represented the entire state, while members of the House are only familiar with a portion of the Senate constituency. Nevertheless, Lublin argues that voters most likely feel that service in the national government requires a different skill set than state-­level elected office. In terms of his analysis, Lublin finds that candidates running for the Senate take into account similar factors as their House counterparts, such as economic, national, and partisan tides. While most scholars agree that not all political candidates are created equal, they have to date been unable to identify what specific characteristics make an experienced candidate more likely to win. Is it their greater name recognition among the voters that gives them an electoral advantage? Are they simply better politicians, possessing certain skills making them more of an expert in the art of electioneering? Perhaps candidates who have previously won elected office better understand how to manage a campaign and utilize New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. 194   Jamie L. Carson and Carrie P. Eaves their resources such as time and money (on this point, see Steen 2006). Absent additional empirical work in this area, it would be difficult to accurately assess which of these competing explanations is most helpful in addressing these important questions. Moreover, the weights attached to each of these possible explanations have direct implications for assessing the health of representation and electoral accountability in the United States. Copyright © 2011. Routledge. All rights reserved. Money in Congressional Elections In 2008, Ric Keller, a four-­term Republican from Florida’s 8th district, was outspent by his opponent to the tune of over $1.4 million dollars. Keller himself raised $1,168,536 for his unsuccessful re-­election bid. However, challengers need not outspend their opponents by millions of dollars to unseat an incumbent member of Congress. Of the 19 incumbents who lost their seats in 2008, 11 outspent their opponents (Abramson, Aldrich, and Rohde 2010). Not surprising to most political observers, the average cost of campaigning for Congress has steadily risen each year. In 2006, for example, the average Senate campaign cost over eight million dollars while the average race for a House seat hovered at just under one million dollars (Jacobson 2009). Candidates for political office must have access to, or be able to raise, substantial sums of money just to be competitive in any congressional election. With the passage of the Federal Election Campaign Act in the early 1970s, candidates for political office were required to disclose the amount of money raised and spent in their political campaigns. As such, it became easier to collect and analyze this information. Jacobson (1978, 1980) was among the first to systematically examine the nature of the relationship between campaign spending and election outcomes in congressional races. Indeed, Jacobson suggested that campaign spending may not have the same effects for incumbents and challengers. Based on his analyses of congressional elections from the 1970s, Jacobson (1978, 1980) found that the amount of money spent by challengers has a much more pronounced effect on election outcomes than does spending by incumbents. In a widely cited study, Jacobson (1978: 469) asserted that “the more incumbents spend, the worse they do.” The reason for this somewhat counterintuitive claim, Jacobson (1978: 469) argues, is that incumbents “raise and spend money in direct proportion to the magnitude of the electoral threat posed by the challenger, but this reactive spending fails to offset the progress made by the challenger that inspires it in the first place.” Challengers, in contrast, get more “bang for the buck” since they are starting with significantly lower levels of name recognition than incumbents. Although surprising at first, Jacobson contends that this anomaly may best be understood in terms of the quality of the challenger and the ­perceived closeness of the race. Incumbents who run against a quality New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Copyright © 2011. Routledge. All rights reserved. Why Some Incumbents Lose Elections   195 c­ andidate in an election often have to raise and spend additional money in their campaign, but additional spending does not yield much in the way of further advantages since incumbents often start off with relatively high levels of name recognition. Moreover, high-­quality or experienced candidates are usually adept at raising money themselves, which makes it more challenging for an incumbent to defeat them. Incumbents who perceive themselves to be in electoral danger can and do typically raise and spend more money than do incumbents who perceive themselves to be relatively safe in a particular election. Therefore, on average, incumbents who win by a large electoral margin spend relatively small amounts of money when compared to incumbents in close races. Thus, Jacobson concludes that increased spending can definitely help the challenger, but is almost always associated with lower vote shares for the incumbent.5 Beyond examining the effects of money in determining electoral outcomes, students of congressional politics have also examined the deterrent effects of incumbent fundraising. Most notably, some scholars believe that incumbents accumulate large sums of money in their war chests in an attempt to “scare off  ” potential challengers from running in a given election (Epstein and Zemsky 1995; Jacobson 1990a, 2009). Indeed, using a variety of different measures, several scholars have found that the amount of money raised by an incumbent does serve as a potential deterrent to candidate entry in congressional races (see, e.g., Goldenberg, Traugott, and Baumgartner 1986; Goidel and Gross 1994; Hersch and McDougall 1994; Box-­Steffensmeier 1996; Carson 2003). Nevertheless, not everyone who studies congressional elections agrees with this conclusion about the deterrent effect of money. Several different students of legislative politics (see, e.g., Krasno and Green 1988; Epstein and Zemsky 1995; Milyo 1998; Ansolabehere and Snyder 2000; Goodliffe 2001, 2007) have found that the amount of money raised by an incumbent does not serve as an effective deterrent to entry of strong or experienced challengers in congressional races.6 Regardless of which side one falls with respect to these divergent findings, nearly everyone agrees that money is a critical component of electoral success. Moreover, and as noted earlier, there is little doubt that congressional elections, like presidential elections, have become much more expensive in recent decades. Growing disparities in spending patterns by incumbents and challengers raise obvious concerns about representation and democratic accountability in the context of House and Senate elections. Indeed, some view the lack of financial competitiveness in congressional races as a contributing factor in the decline of electoral competition during the post-­World War II era (see, e.g., Jacobson and Kernell 1981; Abramowitz 1989, 1991). At the same time, how one views this important debate has important policy implications given recent discussions about campaign finance reform efforts in political races. In light of this, we hope to see additional work examine and resolve these important questions. New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. 196   Jamie L. Carson and Carrie P. Eaves Copyright © 2011. Routledge. All rights reserved. Why Incumbents Lose So why, ultimately, do some incumbents lose their bids for re-­election? Unfortunately for those incumbent members of the House and Senate planning their upcoming campaigns, there is no simple answer. Incumbents work diligently to cultivate a positive brand name for themselves at home in their district. Members of Congress who are unable to effectively advertise, credit-­claim, or take positions may be the most electorally vulnerable (Mayhew 1974a). Yet another possibility is that incumbents lose or do worse than expected because they face highly qualified candidates with previous elected experience (Jacobson 1989). If incumbents are unable to ward off or deter a quality challenger from emerging, the race will likely be much more expensive. Unfortunately for incumbents, spending by challengers has been found to be more effective than incumbent spending as challengers are able to get more “bang for their buck” in congressional races (Jacobson 1980). Quality challengers able to raise large sums of money are likely to be the most competitive opposition an incumbent will face. Indeed, Jacobson (2009) finds that strong challengers with greater financial resources are 20 percent more likely to defeat incumbents than challengers lacking political experience. In some cases national tides, outside of their control, may simply act against incumbents. This was clearly the case in 2006, when public approval of President Bush hovered at around 40 percent and Americans had grown weary of the ongoing wars in both Iraq and Afghanistan. Of the 22 incumbents who lost in 2006, all but one were Republicans (Jacobson 2007b). In many cases, party organizations target vulnerable incumbents who won their previous election by a small margin. The national parties then work to recruit qualified candidates and help raise large sums of money to unseat these incumbents. A final problem largely outside the control of incumbents is redistricting. Bauer and Hibbing (1989) argue that incumbents are most likely to lose only when they face a scandal or their district boundaries have been dramatically altered. Although redistricting is a concern that incumbents must deal with only every ten years, when it does occur, it may be a contributing factor to a small number of incumbent defeats. Ultimately, while the vast majority of incumbents win re-­election handily, they are not invincible as congressional candidates. Incumbents are not guaranteed certain victory in any given election. In the words of Mayhew (1974a: 37), When we say “Congressman Smith is unbeatable,” we do not mean that there is nothing he could do that would lose him his seat. Rather we mean, “Congressman Smith is unbeatable as long as he continues to do the things he is doing.” New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Why Some Incumbents Lose Elections   197 If incumbents stray too far from the desires of their constituents or do not work diligently enough to provide the voters with distributive benefits or bureaucratic assistance, then an incumbent can expect stiff competition in the next election. However, the large majority of incumbents appear to be aware of how their actions can influence electoral outcomes and most work diligently to maintain their seat in Congress. Copyright © 2011. Routledge. All rights reserved. New Directions in Congressional Elections Research As noted earlier, research on congressional elections has made remarkable progress since the early 1970s when scholars first observed that incumbents seemed to be winning re-­election at much higher rates than in the past (Erikson 1971; Mayhew 1974b). Even with the decades of ensuing research, however, we believe several important areas have been overlooked in terms of further enhancing our understanding of congressional elections. To begin with, it is not an oversight on our part that most of the discussion in this chapter has focused almost exclusively on House elections; rather, this is reflective of the current state of the congressional elections literature. On the whole, we know a great deal about House elections but comparatively less about Senate elections as reflected in the asymmetry in the scholarship across the two chambers of Congress. Many research questions, which have been systematically examined in the context of the House, have been virtually ignored in the upper chamber. In fact, and until very recently, much of the extant research on the U.S. Congress simply ignored the bicameral nature of the legislature and focused solely on the House. This is unfortunate, as a single Senate election has a far greater chance of being pivotal than a typical House election given the smaller size of the upper chamber. In addition, with the longer terms of office for U.S. Senators (six years compared to two for their counterparts in the House), the dynamics of the electoral connection may be significantly altered for legislators in the upper chamber. The increased length of time between elections may give senators greater leeway in the legislative process compared to legislators in the House, or it may make them less responsive except in the periods leading up to an election. To systematically examine these types of issues in greater detail, we propose additional research be carried out on Senate elections. In particular, we would like to see new analyses that examine fundamentally important questions such as the incumbency advantage, the impact of challenger quality, and the effects of campaign fundraising in the context of Senate elections. This is certainly not to say that these questions have never been addressed (see, e.g., Abramowitz and Segal 1993; Krasno 1994). Distinct institutional differences between the two chambers may lead to significant variation in elections to both the House and Senate. For instance, a Senate election offers New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Copyright © 2011. Routledge. All rights reserved. 198   Jamie L. Carson and Carrie P. Eaves an incumbent the rare opportunity to watch someone else run a campaign for the same office with the same constituents. At some point during a senator’s term, her other colleague from her home state will be forced to run for re-­ election. As such, senators may alter their campaign behavior based on the campaigns and elections of their counterparts. Unique facets of the Senate such as these merit more scholarly attention. Another area ripe for additional research with respect to congressional elections is greater historical analyses of congressional elections. Most research on House elections artificially limits its sample to congressional elections from 1946 to the present, owing to the availability of these data. However, there is a wealth of congressional elections data prior to that time that is readily available for study. Turning to the study of historical election trends can offer additional leverage on questions relevant to the modern Congress that are more difficult to address in the modern era. For example, Carson, Engstrom, and Roberts (2007) find strong evidence that the incumbency advantage existed in House elections as far back as the late nineteenth century when many of the advantages naturally accruing to incumbents such as the perquisites of office simply did not exist. Studying congressional elections over longer time periods will allow for the development of more dynamic and generalizable theories of congressional elections. A third area where additional research on House and Senate elections could be beneficial is greater emphasis on primary elections. Political scientists know very little about the actions of candidates in primaries and voters’ decisions in congressional primaries. In many congressional districts dominated by one party, results from the general election make it appear as if there is little or no competition for the seat. However, there may be a great deal of competition at the primary level that is simply being overlooked, since relatively few scholars study these types of elections (for exceptions, see Banks and Kiewiet 1989; Galderisi, Ezra, and Lyons 2001). By ignoring this important stage of the electoral process, we may be missing a great deal of existing competition in congressional elections. Understanding when and under what conditions a candidate emerges in primary elections as well as what factors lead to a candidate’s success in a field crowded with fellow partisans is essential. One final area we would like to see scholars devote additional effort to in future research is trying to understand why certain candidates are more likely to run, and ultimately, to win. We know that candidates with previous elected experience perform better in elections to the House of Representatives than do their inexperienced counterparts, yet we cannot readily identify why this is the case. What is it about these quality challengers that makes them better candidates in these elections? Are they more successful fundraisers, having already built a base of support in previous electoral campaigns? Or is it simply that these individuals have greater familiarity with the voters, which obviously translates into increased voter recognition on Election Day? New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. Why Some Incumbents Lose Elections   199 Another possibility is that experienced challengers are better at running and managing a political campaign and, thus, are better suited to compete in congressional elections. Without additional research on House and Senate elections, however, we cannot adequately address these important questions, which are fundamental to our understanding of enhanced accountability, competition, and representation in our electoral system. Conclusion Competition is at the heart of electoral research. As noted from the outset of this chapter, the exceedingly high re-­election rate of incumbent members of Congress (especially in the House of Representatives) suggests to many that the accountability and responsibility mechanism has been severed between constituents and their elected representatives. One should keep in mind, however, that simply because most elections are not competitive does not mean that electoral accountability does not exist. It may be that for the most part voters are content or happy with their current members of Congress. Furthermore, the advantages accruing to incumbents are not insurmountable for the most experienced and qualified candidates. When incumbents do stray from the interests of their constituents, or take their re-­election for granted, they can be defeated. Scholars of both House and Senate elections should continue to seek new and innovative ways to understand competition and accountability in our electoral process. Copyright © 2011. Routledge. All rights reserved. Notes 1. Two recent examples of such elections are 1994 and 2006. The first was worse for Democratic candidates due to the unpopularity of President Clinton, while the latter was unfavorable for Republicans due to growing displeasure with President Bush and the ongoing Iraq War. 2. Since the Civil War, the president’s party has lost seats in the House of Representatives in all but three midterm elections. 3. Several additional scholars offer various offshoots of the surge-­and-decline theory also suggesting that the party of the president tends to lose seats during the midterm. See, for instance, Tufte (1975) and Kernell (1977) on negative voting and the economy, Fiorina (1981) in connection with retrospective voting, and Oppenheimer, Stimson, and Waterman (1986) on the exposure thesis. 4. For evidence supporting each of these contentions, see Cover and Brumberg (1982), Alford and Hibbing (1981), and Bovitz and Carson (2006) respectively. 5. This counterintuitive finding has come under intense criticism by congressional scholars who disagree over how to properly specify empirical models incorporating incumbent and challenger spending. For views contrary to those of Jacobson in the context of House elections, see Green and Krasno (1988), Ansolabehere and Gerber (1994), Levitt (1994), and Erikson and Palfrey (1998). For contrary evidence from Senate elections, see Gerber (1998). 6. Given the prevailing trends in other areas of congressional elections research, it should not come as a surprise that most of the research on the effects of money and elections has been examined using House elections. New Directions in Campaigns and Elections, edited by Stephen K. Medvic, Routledge, 2011. ProQuest Ebook Central, http://ebookcentral.proquest.com/lib/utah/detail.action?docID=668302. Created from utah on 2020-02-05 09:42:15. The Role of Partisan Gerrymandering in U.S. Elections Eric McGhee Public Policy Institute of California mcghee@ppic.org Abstract The literature on partisan bias in U.S. redistricting has emphasized the role of incidental or accidental causes like incumbency and political geography, while questioning the importance of deliberate gerrymandering. I reconsider this settled understanding with a wider range of data, a robust difference-in-differences design, and a new measure of bias called the “efficiency gap” that measures the differential wasted votes of a gerrymander directly. I show that the accepted notion of a null effect for gerrymandering was appropriate for an earlier time period, but that gerrymandering has played a much larger and more durable role in the last two rounds of redistricting. I conclude that it is no longer certain that gerrymandering will play a modest role in the future. Electronic copy available at: https://ssrn.com/abstract=3341696 Existing research on partisan advantage in U.S. elections has emphasized the role of incidental or accidental bias in the modern redistricting era. Factors like incumbency and a state’s underlying political geography have been considered among the most significant factors driving partisan advantage. Gerrymandering, or the deliberate effort to craft the lines for greater advantage, has been found to be less significant. I reconsider this settled understanding with a new measure of partisan advantage called the “efficiency gap”, a broader time series, and data from both U.S. House and state legislative elections. I confirm that there is some role for incidental advantage, and that the first thirty years of the one person, one vote era did in fact feature weak average effects from partisan control. But the last two redistricting cycles have seen an increase in the role of deliberate manipulation independent of these other causes. At the same time, observed partisan advantages have become more durable through the life of the redistricting plan. These results are understandable since many of the conditions that would have led to stronger gerrymandering in earlier redistricting cycles were absent. The incentive to gerrymander will be higher when a party can be more certain of the gerrymander’s effect and when control of the legislative chamber is realistically at stake. Powerful new computing technology and greater predictability of partisan support have made it easier to anticipate a gerrymander’s effect, and in the case of the U.S. House of Representatives control of the chamber is up for grabs in a way that was not the case in 1970s and 1980s. I conclude that given the trends in American political life, it is no longer safe to assume that future partisan manipulation will be modest. Background and Literature Single member district elections like those common to the United States are susceptible to partisan manipulation through the distribution of wasted votes. Votes in excess of those necessary to win are wasted because they do not contribute to victory, as are all those cast for a losing candidate. If one party has more such wasted votes than the other, it is at a disadvantage because its support does not translate as efficiently into representation. Gerrymandering, or the deliberate maldistribution of wasted votes, is often called “packing” and “cracking” because one party’s supporters are “packed” in a small number of districts the party will be guaranteed to win, and the rest are “cracked” across a larger number of districts the party will lose by relatively narrow margins. Although purposeful packing and cracking can give one party an advantage, it is not the only way such an advantage can emerge. Anything that produces the same pattern of support across districts will produce the same advantage. Supporters of a relatively urban party will be naturally packed in dense central cities where it would be difficult to draw a district with a balanced outcome. Likewise, efforts to increase minority representation, such as through the Voting Rights Act, might create districts that are overwhelmingly Democratic, producing yet another packing effect. Even the incumbency advantage can produce an effect that mimics a gerrymander. At first blush it might appear that having more incumbents would make a party look disadvantaged, since each incumbent would win by a larger Electronic copy available at: https://ssrn.com/abstract=3341696 margin and so would appear to waste more votes. But this is only true for incumbents who would still win without the advantage. For those who would otherwise lose, the incumbency advantage is the difference between a few extra votes wasted on a winning effort and a large number wasted on a losing one. On balance, then, having more incumbents is likely to shift bias in a party’s favor. The Supreme Court’s “one person one vote” jurisprudence of the 1960s created a larger opening for gerrymandering than before by forcing virtually every state to redraw its districts every 10 years. Prior to those decisions, it was common in many states to leave districts in place rather than struggle through the political fight required to change them (Cox and Katz 2002). Yet early research in this modern redistricting era uncovered very little role for gerrymandering. Unified partisan control of the process produced some advantage, but the results were modest and inconsistent (Squire 1985; Niemi and Jackman 1991; Gelman and King 1994; Niemi and Abramowitz 1994; Bullock 1982; Butler and Cain 1992). One prominent study even concluded that partisan gerrymandering made elections fairer to both parties by undoing bias that had accumulated through other causes over the previous decade (Gelman and King 1994). Scholars have offered three main explanations for this finding. The first is legal: court precedent and legislation such as the Voting Rights Act have tied legislators’ hands to the point where they have little discretion to secure greater advantage (Gelman and King 1994; Bullock 1982). By this logic, the Supreme Court’s redistricting cases may have forced a reason to redistrict more often, but at the cost of the flexibility needed to produce a partisan advantage. The second explanation is political. Within any redistricting party there is a potential tension between the party’s desire to maximize seats and the desire of individual incumbents to be reelected. The logic of packing and cracking dictates that a party should win its seats by the narrowest acceptable margins to avoid wasting voters that could be used to win another seat elsewhere. But incumbents would naturally prefer to win by broader margins as protection against adverse partisan tides. This conflict between the party and the individual can constrain the seat gains that might otherwise occur (Cain 1984). Indeed, after concluding that partisan gerrymandering played only a limited role, the literature turned to incumbent protection and reduced competition as the more likely consequences of the redistricting process (Basehart and Comer 1991; Abramowitz et al. 2006b, 2006a; McCarty et al. 2006; McDonald 2006). The final explanation is weak party loyalty in the electorate itself. This argument sees parties as ready to gerrymander, but unable to do so effectively because voters are not reliably partisan enough. For example, Born (1985) provides evidence that the plans drawn for the 1980s cycle were intended to advantage the party drawing the lines but often did not deliver on their expectations. In place of gerrymandering, scholars have proposed several alternative causes of partisan advantage. At least one study has emphasized the role of incumbency, and the growing incumbency advantage, in the disappearance of a proRepublican bias in congressional elections in the 1960s and early 1970s (King and Gelman 1991). The effort to produce more representation for minorities under the Voting Rights Act has also been discussed. Districts where minority voters are Electronic copy available at: https://ssrn.com/abstract=3341696 numerous enough to be able to elect their preferred candidates are likely to be strongly Democratic, creating the possibility of more wasted votes for that party (Stephanopolous 2016).1 Several studies confirmed this effect, arguing it cost Democrats around ten congressional seats and two to sixteen state house seats in each of the ten southern states (Brace et al. 1987; Hill 1995; Lublin 1997; Canon 1999; Lublin and Voss 2003; though see also Petrocik and Desposato 1998). Perhaps the most common explanation in recent years has been the simplest one: that partisan bias reflects a state’s underlying political geography (Chen and Rodden 2013; Chen and Cottrell 2016). Because Democrats tend to be concentrated in central cities, it can make it difficult to draw compact districts that will prevent Democratic votes from being wasted in these areas. Rural areas, by contrast, have been modestly but not strongly Republican, making it easy to draw districts that given Republicans the smaller majorities that a classic gerrymander would require. If a redistricting party chooses to adhere to traditional geographic constraints and draw districts that are compact and keep cities, counties, and communities of interest whole, the natural results could easily be a larger Republican advantage without any intent to produce one. Theory and Measurement The conclusions that have been expressed in the literature for the time period analyzed are not necessarily incorrect. Nonetheless, there are two important reasons to doubt the general conclusion that gerrymandering is unimportant in all times and places. One concerns the motivation behind gerrymandering, and the second concerns its measurement. In terms of motivation, it is important to understand the conditions under which a party has the strongest incentive to gerrymander. As noted above, a gerrymandering party must be able to reliably identify its own supporters or the gerrymander itself will not turn out as expected. Weak party loyalty might also drain some of the urgency from a gerrymandering effort. This is especially true if voters are disloyal partisans because they support incumbents. Why should a majority party fight a pitched battle to aggressively gerrymander, when the incumbency advantage will ensure the party holds the majority for the foreseeable future anyway? Independent of these dynamics, gerrymandering is also more compelling when party control of the chamber is at issue. To be sure, parties with strong majorities might still want to pad their numbers, especially if there are supermajority requirements for important legislation. But because the policy and power gains of outright chamber control are greater, the incentive to preserve control is likely to be greater as well. This set of incentives makes the likelihood of finding a gerrymandering effect conditional, not absolute. The conditions today make finding an effect much more likely than several decades ago when most of the earlier gerrymandering studies 1 See Thornburg v. Gingles, 478 U.S. 30 (1986) (setting forth the framework for establishing racial vote dilution under Section 2 of the Voting Rights Act). Electronic copy available at: https://ssrn.com/abstract=3341696 were conducted. Party loyalty in the electorate is demonstrably and dramatically higher, increasing both the predictive power of a gerrymander and a party’s desire to conduct one in the first place. Adding to this predictive power is new computer mapping technology that allows more draft plans to be drawn more quickly and with greater precision. What is more, control of the House of Representatives is legitimately at stake in a way that it was not in the 1970s and 1980s redistricting cycles, increasing the potential gains from gerrymandered congressional plans. Consistent with this idea, McGann et al. (2016) find some evidence of stronger gerrymandering effects in more recent years with similar metrics to those used in the earlier work. Apart from the reasons why gerrymandering might be more common now than in the past, there are also potential measurement problems with the existing literature on partisan advantage. Remarkably little time is spent defining and justifying the measures of partisan advantage that are employed. The most common and most thoroughly defended metric is “symmetry” (Grofman and King 2007). Symmetry measures whether parties with equal vote shares have equal seat shares. This is accomplished either by shifting the distribution of vote shares until both parties have 50 percent of the vote, or until the vote shares of the two parties are reversed.2 In the first case, both parties should have half the seats under the counterfactual; in the second case, the counterfactual majority party should have exactly the same seat share as the majority party had with the actual results. Beyond symmetry, studies have used the “swing ratio”—the change in seat share divided by the change in vote share—for the change before and after redistricting (Niemi and Jackman 1991; Abramowitz 1983). Some have even used the competitiveness (the “responsiveness”) of a plan as signs of a gerrymander’s effect (Cox and Katz 2002). Recent work has also emphasized the significant limitations of these measures for capturing the concept of bias in redistricting (McGhee 2014). Partisan advantage from redistricting is largely about winning more seats without winning more votes: in other words, increasing one’s seat share conditional on vote share. Both symmetry and responsiveness fail to capture this basic idea in many circumstances. Symmetry is especially problematic for uncompetitive states because in such states the symmetry counterfactual is implausible, offering information about seat share for an electoral scenario that will never happen. The swing ratio comes much closer, but without an absolute threshold between a system favoring one party and a system favoring the other, a high ratio might just indicate a competitive plan rather than a biased one. Moreover, the swing ratio studies have tended to use just a single election before and after redistricting. This risks conflating random shocks in any given election with a plan’s longer-term consequences. In this paper I rely instead on a new measure of packing and cracking called the “efficiency gap” (EG). The EG measures the partisan balance of wasted votes directly by tallying up the total number for each party, taking the difference 2 For example, if one party receives 60 percent of the vote, symmetry would involve shifting the distribution until the other party received 60 percent of the vote instead. Electronic copy available at: https://ssrn.com/abstract=3341696 between them, and dividing by the total votes cast (McGhee 2014; Stephanopolous and McGhee 2015). This calculation produces a measure that is itself a difference between seat share and vote share in a given year: specifically, it implies that for every one percent of the vote above 50 percent, a party should get an extra two percent of the seats. Thus, the metric inherently embodies the required notion of seat share conditional on vote share without counterfactuals or calculations of change between elections. This offers a definition of partisan advantage that is consistent for a wide range of states and redistricting plans. I calculate the efficiency gap for state legislative and congressional elections from 1972 through either 2014 (in the case of state legislatures) or 2016 (in the case of the House of Representatives). I limit the data in two ways. First, I restrict the state legislative analysis to lower houses and to states with single-member district elections, and I am forced to drop some states in the early portion of the study period due to missing data. This leaves me with between 23 and 39 states for analysis, depending on the election year (early years feature fewer states than later ones, and the maximum number of states is achieved by 1992). Second, I limit my analysis of the U.S. House to states with at least seven congressional districts to ensure the analysis is focused on states where gerrymandering is likely to produce the greatest benefit to the gerrymandering party. This leaves between 21 and 24 states in each election year for the congressional calculation. For the most part, the EG is a simple calculation based on the observed election results for any given year. However, uncontested seats present special issues. An uncontested seat paints a misleading picture of wasted votes, suggesting no wasted votes for the losing party and too many wasted votes for the winning party. The outcome would have been much closer (though still likely uncompetitive) if the seat had been contested. To obtain a more accurate picture, I impute missing values for these seats with a hierarchical model that regresses the outcome in contested races on incumbency and presidential vote in the district, with random terms for state and year. Because the EG applies to such a wide range of circumstances, it is wellsuited for systematic analysis. There is no need to limit the analysis to competitive states, as is recommended with symmetry (Gelman and King 1994). (Indeed, the fact that McGann, et al. do not limit their symmetry analysis to competitive states raises questions about their conclusions (McGann et al. 2016).) By contrast, with the EG I can examine all years and states where gerrymandering is likely to be an issue, and treat competition as an analytical variable instead. This makes my analysis of gerrymandering the most comprehensive, systematic, and robust to date. Analysis Gerrymandering should only occur when one party controls the entire redistricting process. As such, my goal is to identify how unified partisan control of Electronic copy available at: https://ssrn.com/abstract=3341696 Table 1. Efficiency gaps by redistricting cycle and party control, 1972-2016 State Legislative Congressional Republican Neither Democratic Republican Neither Democratic 1970s -0.056 -0.002 0.056 0.021 0.005 0.020 1980s -0.026 -0.007 0.044 0.017 -0.003 0.039 1990s -0.068 -0.030 0.044 -- -0.011 -0.020 2000s -0.049 -0.017 0.000 -0.097 -0.035 -0.004 2010s -0.078 -0.025 0.016 -0.131 -0.027 0.086 Note: Cell entries are average efficiency gaps for plans under the unified control of each specified party. Negative values favor Republicans; positive values favor Democrats. the process affects a plan’s efficiency gap. My first take at this question can be found in Table 1, which shows the average efficiency gap under plans drawn exclusively by Republicans or Democrats and those drawn by the courts, a redistricting commission, or under divided partisan control (“neither”). Negative values indicate a bias in favor of Republicans, and positive values favor Democrats. For the most part, state legislative elections show exactly the differences that would be expected if gerrymandering were important. Without exception, the average Democratic plan is more Democratic than the average Republican plans, and the average of all other plans falls in between. The largest differences are found in the 1970s and the 1990s, where the average Democratic and Republican plans differ by 11.2%. Congressional plans, by contrast, show far less consistency. There is no discernable pattern from the 1970s through the 1990s. Substantial partisan differences then emerge in the 2000s, and the plans drawn in the most recent redistricting cycle have a larger partisan difference (21.7%) than is seen in any other redistricting cycle or chamber examined here. These simple means do not take into account all of the reasons other than gerrymandering for the observed differences. The differences might be due to deliberate manipulation of the lines, or they might have emerged from any of the other factors discussed above: shifts in political geography, incumbency advantages, or minority representation. My first take at this problem is a simple difference-indifferences model with a control for the lagged EG (Table 2). The state fixed effects in this model capture political geography or minority representation in the broadest sense: enduring differences between the states over this entire time period from 1972 through the present. The year fixed effects give us a sense of how much bias has shifted on average in favor of either party from one year to another, likely due to average changes in the incumbents in each party or national shifts in political Electronic copy available at: https://ssrn.com/abstract=3341696 Table 2. Explaining the efficiency gap in state legislative and congressional elections, 1972-2016 State Legislative (1) (2) Congressional (1) (2) Party control 0.021*** (0.003) 0.015*** (0.004) 0.029*** (0.007) 0.019** (0.007) Lagged EG 0.374*** (0.035) 0.141** (0.042) 0.278*** (0.044) 0.108# (0.056) Black representation -- -0.194* (0.082) 0.023 (0.089) Latino representation -- -0.118# (0.071) -0.163 (0.153) Urbanization -- -0.119 (0.088) -0.363# (0.193) Incumbent balance -- 0.105*** (0.013) 0.089*** (0.020) Intercept State fixed effects? Year fixed effects? RMSE Adjusted R2 N 0.003 (0.015) 0.065 (0.049) -0.040# (0.022) 0.166 (0.110) Yes Yes Yes Yes Yes Yes Yes Yes 0.037 0.639 745 0.034 0.675 681 0.072 0.389 487 0.071 0.410 465 Note: Cell entries are ordinary least squares coefficients and standard errors, calculated in R 3.4.0. *** p
Purchase answer to see full attachment