Analyses of the decision to run for the U.S. House are generally based on the self-reported motivations of a sample of candidates or potential candidates. In this article we take a different approach. We use a quasi-experimental design to model the decision of a losing candidate to initiate or reject an immediate repeat match-up against the person to whom they lost in the general election of the previous cycle. Of these potential repeaters we look at "strong" challengers-those who lost but secured more than 40 percent of the vote. We find the decision to repeat to be shaped at least somewhat by evaluations of the candidates chances of winning the party's nomination and her personal desires and abilities. However, when we examine only "strategic" potential repeaters-those who lost in an open seat contest the first time around-we find the decision to repeat to be driven largely by broader political or partisan trends that affect the candidate's evaluation of her chances of winning the general election. This finding confirms analyses of strategic candidates using other data.
Between 1970 and 2000, 670 general elections for the U.S. House of Representatives matched major party candidates who had faced each other in at least one House contest before. In all but one election year since 1976, at least 10 percent of incumbents faced a repeat challenger. Although not exactly frequent, then, such repeat matches are an enduring piece of congressional elections. As a consequence, they have received some scholarly attention.
The principal research question asked about these races is whether the act of repeating assists the challenger the second time around. Ackerman (1957), in his analysis of such races in the early 1950s, for example, found little net advantage accruing from a second challenge. But more recent and comprehensive studies have drawn different conclusions. Squire and Smith (1984), Mack (1998), and Renka (2001) find that from the 1960s through the 1990s some types of repeat challengers are discernibly more successful than first-timers who have similar attributes and have run similar campaigns. In each of these analyses, candidates tend to receive more money and do better in their second bid. This may be a result of the inherent advantages of repeating that allow the repeat challenger to hit the ground running-namely an extant campaign mechanism, previously cultivated contacts, and name recognition.
Many of the repeaters are political novices who neither do well nor seem to be striving for victory. As Canon (1993) argues, no single theory can really explain the motivations of such political amateurs. They seem often drawn to the rematch by an intrinsic dislike of the incumbent or the thrill of the campaign. In addition, as Maisel (1986) suggests, such repeaters may desire to curry favor with their party to improve their visibility in order to run for lower office or simply to make an ideological statement. Consider, for instance, Ted Tyler, the Republican who challenged Democratic Representative Eva clayton in her heavily black North Carolina district between 1992 and 1998. Tyler seems to have been uninterested in the long odds against a Republican victory, and campaigned extensively on an anti-government message. He surely enjoyed it, too, although perhaps not as much as John Buchanan, a Democrat who ran unsuccessfully in Ohio's heavily Republican Tenth District against incumbent Clarence Miller in every election between 1982 and 1990.
Certainly not all repeat challengers fit this description, however. Current or recent House members like Ted Strickland, Mark Neumann, Bill Luther, Walter Capps, James MaIoney, John Kline, and Newt Gingrich were all repeaters. For strong candidates such as these-candidates who have a reasonable chance of winning and who presumably make rational decisions in the'ir attempt to win-repeat matches are worth exploring again and with an approach different from that employed by previous work.
This is because such candidates and the races they run provide a quasi-experimental design within which we can examine why rational individuals declare their candidacy for the U.S. House.1 Obvious systematic approaches to modeling the decision to enter a House race necessarily involve sampling potential candidates and relying on respondents' personal and frank assessments of their own candidacy. Groundbreaking work of this type has been undertaken by the Candidate Emergence Project and Citizen Political Ambition Study.2 It has spawned a rich set of data and interesting findings (Fox and Lawless 2005; Kazee 1994; Maisel and Stone 1997; Maisel, Stone, and Maestas 2001; Maisel, Maestas, and Stone 2005; Stone and Maisel 2003). In this study, however, we present a different set of data by which to test theories of congressional candidacy by comparing repeaters with those who declined to repeat and candidates' initial races with their actual or potential second ones. We utilize data gleaned from direct observations of candidate behavior and the tangible attributes of people and campaigns.
A CATEGORIZATION OF THEORIES OF CONGRESSIONAL CANDIDACY
There are numerous theoretical approaches that allow us to understand a strong candidate's decision to run for the Congress.3 Here, we arrange them into three categories. All of these categories are pan of a broad "rational actor" model of candidacy that assumes potential strong candidates have scarce resources and think in terms of the personal costs and benefits of running (Black 1972; Bianco 1984; Rohde 1979). The categories separate possible motivations by discussing them in terms of the potential candidate!; evaluation of distinctly different phenomena.
The first of these phenomena is the candidates assessment of her chances of winning the general election. There is a set of arguments that suggest strong challengers evaluate their chances of winning the seat in the context of the fundamental national political and economic environment (Jacobson and Kernell 1981), for example. Whether the upcoming election is a midterm or is in a presidential year will likely play a role in the decision to run, with potential candidates from the incumbent presidents party often choosing not to run in what are generally inhospitable midterm years (Abramowitz 1985; Campbell 1993). Conditions in the next election can also be gauged through the lens of public perceptions of the parties and the performance of the national economy. Frequently candidates will make decisions accordingly (Alesina, Londregan, and Rosenthal 1993; Bianco 1984; Heller and Norputh 1994; Jacobson and Kernell 1981; Jacobson and Kemell 1990; Tufte 1975). Moreover, any changes the district has undergone in the reapportionment process that takes place every decade may also be a factor in the assessment of general election outcomes (Gelman and King 1994; Hetherington, Larsen, and Globetti 2003; Maisel, Maestas, and Stone 2004). Losing candidates may be expected to run again when these political and economic factors are tilted against the incumbent to a greater degree than they were in the previous election.
second, in order to wage another challenge, aspiring repeaters must first win their party's nomination. Measurable factors that could affect the potential repeater!; assessment of the chances of winning her party's nomination include formal rules, enacted within each state or by each party, that govern filing and primary types and dates (Ansolabehere and Gerber 1996; Wnghton and Squire 1997). These rules or dates may benefit certain types of candidates while hurting others. In addition, the pool of viable and competitive candidates within a party may influence a candidate's chance of renomination (Ehrenhalt 1991; Krasno and Green 1988; Stone and Maisel 2003; Romero 2004). These factors have little or nothing directly to do with candidates' views about the likely outcome of the general election. As Banks and Kiewiet (1989) argue, in some instances candidates may even believe their chance of winning the primary will increase as their chance of winning the general decreases. Stone and Maisel (2003) have provided empirical evidence to show that this is the case.
Third, we expect candidates also to assess the effect that their personal characteristics, including the desire to win and political skills, will have on the campaign. Despite being notoriously nebulous and idiosyncratic, it is these kinds of motivations that the Candidate Emergence Project is admirably trying to conceptualize and measure. Personal characteristics include a variety of traits-some more easily observable than others-like ideological conviction, the ability to grasp issues, a dedication to service, the personal opportunity costs of a campaign, and a willingness to speak in public (Maisel, Stone, and Maestes 2001). Perhaps the most studied is progressive ambition-the notion that careerist office holders make their way, generally one rung at a time, up the political ladder (Fowler 1993: 54-60; Fowler and McClure 1989; Schlesinger 1966). Related to this is the potential candidate!; evaluation of how much she desires a House seat at any particular time. Progressively ambitious candidates are more likely to want it.
DATA AND HYPOTHESES
The theories spawn a series of specific hypotheses that congressional scholars have tested using a variety of different candidates at many different times. As stated above, we test a number of these hypotheses about congressional candidacy by examining a data set of individuals who could initiate a repetition of a previous general election race against a major party incumbent because they had lost to this person in the first race. Our central goal is to test the hypotheses and compare their performances in model specifications using two different sets of potential repeaters that we define below-"strong" candidates and a subset called "strategic" candidates. Fully explaining the decision to repeat is not our primary aim.
Our data obviously allow us to understand more about why strong candidates repeat. But they also provide a quasiexperiment in which we can draw conclusions about the decision to run for the House by potentially strong candidates more generally-that is, the data can help us understand the decision to run of potential first timers as well.
The decision to repeat is useful in this regard for four reasons. First, for those who do decide to repeat, we can compare the first and second races. If we assume rational behavior, we should expect a strong candidate to have the expectation that one or a combination of things in the second race will be different, and better, for her. With this in mind, atypical characteristics of an initial race and/or discernible differences between the timing of the first and potentially second elections can help us understand the decision to run.
Second, we can compare those who repeat with those who do not. Repeaters, for example, may share distinctive personal attributes. Third, using potential repeaters means we have a fixed universe of cases. Because it is easy to observe whether candidates from a previous election chose to run again or not, we do not need to become involved in the uncertain task of identifying potential candidates from the general population. Fourth and finally of course, modeling the decision to repeat allows us to hold a challengers opponent constant.4 Holding the opponent static makes it easier to isolate other potential reasons for a repeat.
The data set we use consists of all potential repeaters who can be considered strong challengers and who ran the first time from 1978 to 2000. We identify strong candidates as those who received at least 40 percent of the vote in the initial general election. This may seem a somewhat arbitrary definition, but absent the ability to understand the precise motivations of candidates we think it represents the best way in which to pinpoint credible challengers with an opportunity to win a rematch with the victor in the first race. These individuals are most likely to fit within the basic rational actor model. Forty percent is also the cutoff used by Canon (1993) and Jacobson (1997) to identify candidates who are not political amateurs. Moreover, expanding our definition of strong challengers to those who received in excess of 35 percent of the vote did not change our substantive findings at all. These individuals represent the universe of cases-strong challengers who are potential repeaters.
We excluded a number of potential cases. We jettisoned, for example, observations where the incumbent retired or died-except those where the incumbent retired or died after the filing deadline or, where we could ascertain, the repeat challenger's declaration of candidacy.5 Because of its all-party primary, potential challengers from Louisiana are excluded, and special elections are not counted as first races-a repeat in a subsequent general is therefore considered a first, and not a second, race. We include only general elections where both candidates are from the two major parties. If there are successive repeat matches, data are from the race immediately prior to the repeat, not the first contest. The data do not include cases in which, as far as we could tell, the potential repeater was placed in a district different from the incumbent after redistricting. The data set has 1,314 cases of which 240 are considered repeaters.
We are also modeling the behavior of individual candidates using aggregate data. This might suggest a problem of ecological inference. But we utilize direct observations of the behavior of individuals. Conclusions based upon general patterns of behavior are not inferred from indirect observations.
From the theoretical work on congressional candidacy discussed above, we derive a series of testable hypotheses: Hypotheses testing general election theories:
H^sub 1^: Strong candidates of the president's party are less likely to repeat in a midterm.
The "iron law" of midterms-that the presidents party always loses seats in such elections-was only broken in 1998 and 2002. We should therefore expect decisions to repeat to be highly sensitive to the electoral calendar and the party of the presidency. To test this hypothesis we code cases in which the potential repeater was of the president's party and the upcoming election was a midterm -1. We code cases where the potential repeater was of the party not in control of the White House and the upcoming election was a midterm 1. Such individuals should expect a favorable year for House candidates of their party across the country. All cases in which the approaching election is in a presidential year are coded 0. We call this variable midterm.
H^sub 2a^: Strong candidates of the president's party are less likely to repeat if the economy deteriorates after the first race.
H^sub 2b^: Strong candidates are more likely to repeat if the economy deteriorates after the first race.
A number of scholars believe congressional elections are largely unaffected by the economy (Alesina and Rosenthal 1995; Erikson 1990). Others have suggested that the electoral performance of and declarations of candidacy by members of the president's party are determined by economic conditions (Tufte 1978; Jacobson 1989). Here, potential repeaters of the president's party will be encouraged to run again by a strong economy, those of the opposition discouraged. More recent research, however, shows that economic conditions impact the success of incumbents, regardless of party, even more (Grier and McGarrity 1998, 2002). In this case, potential repeaters may feel they can use a weak economy to their advantage in a general election.
We test these two assertions using the change in real disposable income (RDI) from the fourth quarter of the year of the initial match-up to the fourth quarter of the next year-roughly the time when the decision to run again must be made. RDI has been used to measure the impact of economic conditions on congressional elections many times before (Born 1986; Jacobson 1989). For H^sub 2a^ the value is multiplied by -1 if the potential repeater is not of the presidents party. For H^sub 2b^, we just use the raw change in RDI. The first variable is labeled presidential economy, the second variable incumbent economy.
H^sub 3^: Strong candidates are more likely to repeat if the district is, between races, redrawn unfavorably for the incumbent.
Repeating should occur when the new district is drawn favorably for a candidate of the potential repeaters party. To test this hypothesis, we use a variable favorable redistricting that is, for those districts that are redrawn, essentially the difference between the partisan leanings of the district after and prior to reapportionment as revealed by its presidential vote.6
Hypotheses testing party nomination theories:
H^sub 4^: Strong candidates are more likely to repeat if they are Republicans.
This hypothesis is derived from the notion that, over much of the 1978-to-2002 period, Republicans had a harder time recruiting quality candidates for the House than have Democrats (Bond, Fleisher, and Talbert 1997). This was the result of having a smaller national pool of such individuals (Ehrenhalt 1991). As a result, Republican candidates generally faced fewer opponents in primaries. We test the hypothesis using a dummy variable challenger party that is coded 1 for Republicans and O for Democrats.
H^sub 5^: Strong candidates are more likely to repeat if they are subject to an open primary.
State laws and party rules dealing with primary types should also affect the decision of strong candidates to repeat. Potential repeaters should be more successful in areas where parties have less control over primary outcomes (Kanthak and Morton 2001). In open primaries, the repeater is likely to be the front runner, mainly because of the residual resources of the previous campaign-such as a staff, name recognition, support networks, and access to contributors. On the other hand, in closed primaries where parties exercise more control over their nominees, parties may not necessarily support repeat candidacies because such individuals have tarnished their luster in the previous loss and there may be more attractive alternatives. In other words, defeated candidates are the de facto front-runners for the nomination after their first bid, and will only lose this status if political elites pass a negative judgment upon their previous effort to win the seat. The variable primary type is used to test this hypothesis. It is coded 0 for closed primaries, 1 for semi-open or modified open primaries, and 2 for open and blanket primaries. The data are from Bott (1990).
H^sub 6^: Strong candidates are more likely to repeat if the filing deadline is early.
Filing and primary dates may also have an effect on the decision to run (Ansolabehere and Gerber 1996; Wrighton and Squire 1997). In states where there is an early filing deadline, potential candidates have less time to decide whether or not to run. The relatively quick resumption of campaigning ought to benefit an individual who ran but lost in the previous cycle. She will have a campaign in place, extant relationships with donors and party officials, and instant name recognition. These are important resources in the attempt to capture the nomination and make winning the primary more likely. A variable filing is the number of months between the filing deadline and the general election.7
Hypotheses testing personal theories:
H^sub 7^: Strong candidates who have previously held elective office are more likely to repeat.
Consistent with the theory of progressive ambition (Schlesinger 1966), we should expect those who have held lower elective office to be less disconcerted by electoral defeat and more willing to exploit the value other scholars say have accrued by running in the first race (Mack 1998; Renka 2001; Squire and Smith 1984). These individuals are more likely to be careerists and have a greater desire to be in the House. We code potential repeaters who have held elective office previously 1 in a variable termed experience, those who have not 0.8 It measures the potential repeaters evaluation of her desire to win a House seat at this time.
H^sub 8^: Strong candidates who spent relatively little money in the first race are more likely to repeat.
If, compared with similar candidates in the first election, the candidate was able to spend relatively little money, she might be convinced that she could obtain more for the second race and hence do better. Spending relatively little money is indicative of a poor campaign. Concomitantly having lost after being able to spend a lot of money may persuade a potential challenger that she cannot win even after running a strong campaign. Here we are getting, the best we think we can, at a potential repeater's personal assessment of her performance in the first race and expectations about it in the second. This has not been considered in previous literature, but absent systematic data about the personal evaluation of the first campaign-the amount of time the candidate spent campaigning and the quality of the decisions she made, for example-money spent provides a nice proxy of the potential repeater's evaluation of her performance. The variable money spent is the percentage the money spent in the first race is of the mean for all candidates in the potential repeater's category that year. We look at six separate categories of candidate, Democrats and Republicans in each of incumbents, challengers, and individuals running for open seats. A Democrat who challenged an incumbent in the first race, then, has the total money she spent compared with the mean of all Democratic challengers for that year.
It is necessary for us to control for a number of factors that are likely to affect the decision to repeat. First, we use a dummy variable previous repeater if the challenger has repeated against the incumbent before (if this is so, the case is coded 1). Repeating previously is likely to diminish the chance of doing so again as strong repeaters learn that defeating their opponent is very difficult. second, we code cases 1 if the challenger was the incumbent in the first race-for this variable, previous incumbent, we believe such potential repeaters will decide not to run having just lost their seat, often after many years in the House. Third, we use a variable that identifies the first race as an open seat contest-open seats are coded 1. Incumbents have enormous advantages in House elections (Gelman and King 1990; King and Gelman 1991; Levitt and Wolfram 1997) and, like moths to a flame, better challengers are drawn to open seats (Jacobson 1997: 77-78; Cox and Katz 1996). Fourth, we control for the seniority of the incumbent. If strong candidates do oppose incumbents, they look for vulnerable ones (Hersch and MacDougall 1994; Jacobson 1997: 35-37). The more senior an incumbent, the more likely she has access to electoral resources that assist with reelection (Herrera and Yawn 1999). Seniority is also indicative of how difficult it will be to remove the member (Jacobson 1997: 77-78; Cox and Katz 1996). The variable incumbent seniority is the number of years from the first race that the incumbent had been in the House continuously. Finally, we use a measure that is the vote difference in percentage points between the potential challenger and the incumbent in the first race. The thinking here is the closer the first race, the more likely a repeat.
As a first cut at our data, let us consider all strong potential repeaters. Table 1-in which the dichotomous dependent variable is coded 1 if the strong challenger decides to repeat against the current incumbent, 0 if not-shows the results of a logit model in which all cases and independent variables are included. Standard errors are robust and clustered by the year of the first race because observations within the same year are not necessarily independent of one another.
There are some interesting findings. Not surprisingly, the controls all perform well. Indeed, it is clear that collectively they provide most of the model's predictive power. More importantly for our analysis, none of the variables used to examine the assessment of a candidates chances to win the general election have statistically significant coefficients and we must therefore reject the related hypotheses. Note also that although the primary type and filing variables do not have statistically significant coefficients, the challenger party variable does, and it performs as hypothesized. Traditionally, Republicans are more likely to repeat. Indeed, when dichotomous variables are given their modal value and all continuous variables are held at their means, the predicted probability of a Republican repeating is 21.3 percent, for a Democrat it is 15.1 percent. We attribute the difference to the traditionally smaller pool of potential Republican candidates.
For the most part, however, the hypotheses used to see whether these candidates run again for personal reasons are corroborated. Those who raise relatively little money the first go around seem to think their campaign can be better in a repeat match. Holding all the other variables at the levels we describe above, the predicted probability of a challenger who spent 10 percent as much as a similarly situated candidate in the year of the first race repeating is 30.2 percent (the 10 percent figure is approximately one standard deviation below the money spent variables mean). For one who spent 3.1 times more than the candidates she is compared to (a figure that is approximately one standard deviation above the variable's mean) the predicted probability of a repeat is 15 percent.
The money spent variable's performance can be interpreted differently, however. Note that the experience variable behaves counter-intuitively. This, in turn, suggests that experienced potential repeaters are more deterred by defeat than their inexperienced counterparts and that holding previous elected office does not compel them to rush to try to win a House seat. It may be that politically experienced potential repeaters are more realistic about their prospects after they are defeated the first time, while those who have not held political office before either do not understand or do not care about the meaning of the initial defeat and the difficulties of unseating congressional incumbents. Similarly, it may be the case that those who can only spend little money in the first race have attributes and outlooks like those of inexperienced challengers. Indeed, as if to confirm that many potential repeaters do not care about their prospects in the general, note that the seniority control has a positive coefficient. It shows that challengers are more likely to repeat against presumably entrenched senior incumbents.
The aggregate data presented in Table 1 may obscure some important patterns, however. Therefore in the model specification shown in Table 2, we analyze a subset of the population of strong potential repeaters who we might expect to make their decisions to run again in a different way. This subset is what we call "strategic" and is those 304 potential repeaters who ran for an open seat in the first race. Their decision to run for an open seat and win their party's nomination in the first instance is also presumably indicative of their political acumen (Jacobson and Kernell 1981).9 As such, the decision of these candidates to repeat should be more likely to corroborate hypotheses related to an assessment of general election outcomes (Alesina, Londregan and Rosenthal 1993; Bianco 1984; Heller and Norputh 1994; Jacobson and Kernell 1981, 1990.
The results of this model are considerably different to that of strong challengers more broadly defined.10 It does seem that the decision of strategic challengers to repeat is driven more by an assessment of their chances of winning the general election than an evaluation of personal qualities or primary election outcomes. The midterm and incumbent economy variable have statistically significant coefficients with predicted signs. Indeed, in substantive terms, their impacts are quite large. The predicted probability of a strategic candidate repeating when RDI is -1.3 percent (roughly one standard deviation below its mean) and all other variables are held at their usual levels is 8.7 percent; when it is at 3.7 percent (and approximately one standard deviation above its mean) the predicted probability is 4 percent. The predicted probability of a strategic challenger of the president's party repeating in a midterm is 3 percent; it is 10 percent for a strategic challenger of the party that does not occupy the White House.
One other variable has a statistically significant coefficient. The primary type variable's performance suggests that strategic challengers who have closed primaries are more likely to repeat than those that have open ones. This is counter to our hypothesis. It suggests, perhaps, that party activists recognize the inherent advantages the potential repeater can take into a general. Neither of the variables that measure personal characteristics performs robustly, however.11
The results in Table 2 are consistent with much of the work done on strategic candidacies (Alesina, Londregan, and Rosenthal 1993; Bianco 1984; Heller and Norputh 1994; Jacobson and Kernell 1981, 1990; Tufte 1975). The decision of strategic candidates to repeat is shaped greatly by an evaluation of the individual's chances of winning the general election. This makes these candidates very different from strong candidates more generally who do not respond to assessments of general election outcomes but, to a much greater degree, base their decision to repeat on an analysis of their personal attributes and chances of winning their party's nomination.
We have examined the motivations for strong challengers to initiate a repeat campaign against the candidate to whom they lost two years previously. This exercise, we argue, is not only one in testing hypotheses about why such individuals repeat but sheds some light on why people decide to run for the House in the first place. It does so as a parsimonious and systematic quasi-experiment that assists in our understanding of what is an extremely complex process.
Our findings generally reinforce those detected by other scholars of congressional candidacy. We find that strategic candidates-individuals we identify in this study as those whose initial run was for an open seat-assume their repeat candidacy based largely upon their evaluation of their chances of winning the general election. By contrast, candidates whom we theorize should act rationally but did not undertake their first run for the House in an open seat contest-that is, all other strong candidates-base their decision to at least some degree upon an assessment of their own personal qualities and, to a lesser extent, their chances of winning their party's nomination.
We do not claim to have explained definitively the decision to repeat, however. This is because candidacy decisions are often shaped by intangible and often whimsical motivations. How a repeat candidacy will affect career and family is just one such question candidates must ask of themselves. There are, as Gary Stolbert (2003) notes, costs to running-"personal, physical, emotional, and financial"-that are large but difficult to quantify). These are similar to the personal factors we have shown to play an especially important role in the decisions of inexperienced and non-strategic candidates to run again. We hope the work emanating from the Candidate Emergence Project will continue to further our understanding of how personal factors shape the decision to run.
NOTE: Previous versions of this article were delivered as papers to the 2001 annual meeting of the Midwest Political Science Association and the 2002 annual meeting of the American Political Science Association. We would like to thank Paul Gronke, Paul Herrnson, and Kimberly Maslin-Wicks for their comments and suggestions and Cheng Blanco and Sydney Beveridge for research assistance. Gary Jacobson kindly provided us with data on candidates' political experience and Jeffrey Milyo provided candidate data from similar research.
1 We are not the first to use repeat matches with this methodology. Steven D. Levitt (1994) did so in an article about campaign spending.
2 Caddie (2003) provides nine case studies of the decision to run from across the country. He examines, however, entry into state and local races.
3 Fowler (1993: 41-71), for example, suggests there are five theoretical "traditions" on the subject. The first or "sociological' tradition examines the social standing of potential candidates, the "psychological" tradition assesses personality traits, the "process" tradition focuses on opportunity structures, the "rational actor" tradition looks at the personal costs and benefits of running, and the "rule-based" tradition analyzes the effect of political institutions.
4 The opponent may be held constant, but we do realize he or she will have changed. If the first race was for an open seat, for example, the opponent is now an incumbent. If she was an incumbent to begin with, she is now more senior. We enter variables in our model to control for such effects. For the first, we utilize the dummy open seat, for the second we use a variable incumbent seniority that is the number of years the incumbent has served.
5 It is hard to tell in such cases whether or not the potential repeater's decision to run or not to run is based upon the presence of the incumbent. We did include some repeaters where the incumbent had retired or died before the filing deadline because in these cases we discovered decisions to take on the incumbent before the retirement or death intervened. These cases were identified from our reading of district commentaries in The Almanac of American Politics and primary election results in America Votes.
6 The score is actually calculated as (d ^sub t+1^ - n) - (d ^sub t-1^ - n) where d = the last Democratic presidential candidate's two-party vote in the district, n is that vote nationally, t+1 is the district in the upcoming election, and t-1 is the district in the first race. The score for Republican potential repeaters is made negative. For times and districts where there is no redrawing, the variable's value is 0.
7 It could be argued that the date of the primary is a better measure, since it determines the length of time contestants for the out party's nomination have to campaign. Early primaries provide potential repeaters with great advantages, their opponents must accrue support more quickly. Not surprisingly, however, primary dates and filing deadlines are highly correlated.
8 Gary Jacobson kindly provided these data.
9 We could have also analyzed the subset of potential repeaters who had been defeated as incumbents the first time around. Indeed, there are several noteworthy cases in our data of previous incumbents running again, and often winning back their seat in a rematch. In 1996, for instance, three Democrats won back seats they had lost in 1994. This might suggest that former incumbents can be more strategic about choosing rematches. Only 14 of our 188 defeated incumbents ran again, however. In addition, a logit model with rare-event corrected estimates revealed that, with the exception of vote difference, none of the variables used in the model of strategic challengers were statistically significant.
10 Because there is no variance in the previous repeater, previous incumbent, and incumbent seniority variables they are omitted from the analysis of strategic potential repeaters.
11 Separating the data on strong repeaters into various multi-year snapshots does not provide us with substantively different results. These candidates run again more for personal than political reasons whatever the period. In snapshots of years post-1984, however, we find the favorable redistricting and presidential economy coefficients to be statistically significant and of the predicted sign.
Abramowitz, Alan I. 1985. "Economic Conditions, Presidential Popularity, and Voting Behavior in Midterm Congressional Elections." Journal of Politics 47: 31-43.
Ackerman, Donald H. 1957. "Significance of Congressional Races with Identical Candidates in Successive Elections." Midwest Journal of Political Science 1: 173-80.
Alesina, Alberto, John Londregan, and Howard Rosenthal. 1993. "A Model of the Political Economy of the United States." American Political Science Review 87: 12-33.
Alesina, Alberto, and Howard Rosenthal. 1995. Partisan Politics, Divided Government, and the Economy. New York: Cambridge University Press.
Ansolabehere, Stephen,, and Alan Gerber. 1996. "The Effects of Filing Fees and Petition Requirements on U.S. House Elections." Legislative Studies Quarterly 21: 249-64.
Banks, Jeffrey S. and D. Roderick Kiewiet. 1989. "Explaining Patterns of Candidate Competition in Congressional Elections." American Journal of Political Science 33: 997-1015.
Bianco, William T. 1984. "Strategic Decisions on Candidacy in U.S. Congressional Districts." Legislative Studies Quarterly 9: 351-64.
Black, Gordon S. 1972. "A Theory of Political Ambition: Career Choices and the Role of Structural Incentives." American Political Science Review 66: 144-59.
Bond, Jon R., Richard Fleisher, and Jeffrey C. Talbert. 1997. "Partisan Differences in Candidate Quality in Open Seat House Races." Political Research Quarterly 50: 281-99.
Born, Richard. 1986. "Strategic Politicians and Unresponsive Voters." American Political Science Review 80: 599-612.
Bott, Alexander J. 1990. Handbook of United States Election Laws and Practices. New York: Greenwood Press.
Campbell, James. E. 1993. The Presidential Pulse of Congressional Elections. Lexington: University of Kentucky Press.
Canon, David. 1993. "Sacrificial Lambs or Strategic Politicians? Political Amateurs in the United States Congress." American Journal of Political Science 37: 1119-41.
Cox, Gary W, and Jonathan N. Katz. 1996. "Why Did the Incumbency Advantage in U.S. House Elections Grow?" American journal of Political Science 40: 478-97.
Ehrenhalt, Alan. 1991. The United States of Ambition: Politicians, Power, and the Pursuit of Office. New York: Times Books.
Erikson, Robert S. 1990. "Economic Conditions and the Congressional Vote: A Review of the Macrolevel Evidence." American Journal of Political Science 34: 373-99.
Fowler, Linda L. 1993. Candidates, Congress, and the American Democracy. Ann Arbor: University of Michigan Press.
Fowler, Linda L., and Robert D. McClure. 1989. Political Ambition: Who Decides to Run for Congress. New Haven, CT: Yale University Press.
Fox, Richard L., and Jennifer L. Lawless. 2005. "To Run or Not to Run for Office: Explaining Nascent Political Ambition." American Journal of Political Science 49 (3): 659-76.
Gaddie, Ronald Keith. 2003. Born to Run: Origins of the Political Career. Lanham, MD: Rowman and Littlefield.
Gelman, Andrew, and Gary King. 1990. "Estimating Incumbency Advantage Without Bias." American Journal of Political Science 34: 1142-64.
Grier, Kevin B., and Joseph R McGarrity. 1998. "The Effect of Macroeconomic Fluctuations on the Electoral Fortunes of House Incumbents." Journal of Law and Economics 41: 143-61.
_____. 2002. "Presidential Party, Incumbency, and the Effects of Economic Fluctuations on House Elections, 1916-1996." Public Choice 110: 143-62.
Heller, H. Brandon, and Helmut Norputh. 1994. "Let the Good Times Roll: The Economic Expectations of U.S. Voters." American Journal of Political Science 38: 625-50.
Herrera, Richard, and Michael Yawn. 1999. "The Emergence of the Personal Vote." Journal of Politics 61: 136-50.
Hersch, Philip L., and Gerald M. MacDougall. 1994. "Campaign War Chests as a Barrier to Entry in Congressional Races." Economic Inquiry 32: 630-41.
Hetherington, Marc J., Brace A. Larsen, and Suzanne Globetti. 2003. "The Redistricting Cycle and Strategic Candidate Decisions in U.S. House Races." Journal of Politics 65: 1221-34.
Jacobson, Gary C. 1989. "Strategic Politicians and the Dynamics of U.S. House Elections, 1946-1986." American Political Science Review 83: 773-93.
_____. 1997. The Politics of Congressional Elections, 4th ed. New York: Longman.
Jacobson, Gary, and Samuel Kernell. 1981. Strategy and Choice in Congressional Elections. New Haven, CT: Yale University Press.
_____. 1990. "National Forces in the 1986 U.S. House Elections." Legislative Studies Quarterly 15: 65-87.
Kanthak, Kristin, and Rebecca B. Morton. 2001. "The Effects of Primary Systems on Congressional Elections." In Peter Galderisi, and Mike Lyons, eds., Congressional Primaries and the Politics of Representation. Lanham, MD: Rowman and Littlefield.
Kazee, Thomas A. 1994. "The Emergence of Congressional Candidates." In Thomas A. Kazee, ed., Who Runs for Congress? Ambition, Context and Candidate Emergence. Washington DC: Congressional Quarterly Press.
King, Gary, and Andrew Gelman. 1991. "Systematic Consequences of Incumbency Advantage in United States House Elections." American Journal o/Political Science 35: 110-38.
Krasno, Jonathan, and Donald Green. 1988. "Preempting Quality Challengers in House Elections." Journal of Politics 50: 920-36.
Levitt, Steven. 1994. "Using Repeat Challengers to Estimate the Effect of Campaign Spending on Election Outcomes in the United States House. "Journal of Political Economy 102: 777-98.
Levitt, Steven, and Catherine D. Wolfram. 1997. "Decomposing the Sources of Incumbency in the U.S. House." Legislative Studies Quarterly 22: 45-60.
Mack, W R. 1998. "Repeat Challengers: Are They the Best Challengers Around?" American Politics Quarterly 26: 308-43.
Maisel, L. Sandy. 1986. From Obscurity to Oblivion: Running in the Congressional Primary. Knoxville: University of Tennessee Press.
Maisel, L. Sandy, Cherie D. Maestas, Walter J. Stone. 2005. "Candidate Emergence in 2002: The Impact of Redistricting on Potential Candidates' Decisions." In Thomas E. Mann and Bruce Cain, eds., Party Lines: Competition, Partisanship and Congressional Redistncting. Washington, DC: Brookings Institution Press.
Maisel, L. Sandy, Cherie Maestas, and Walter J. Stone. 2004. "Candidate Emergence in 2002: The Impact of Redistricting on Potential Candidates' Decisions." Paper presented to the Brookings Institution/Institute of Governmental Studies Conference on Competition, Partisanship, and Congressional Redistricting, Washington, DC.
Maisel, L. Sandy, and Walter J. Stone. 1997. "Determinants of Candidate Emergence in U.S. House Elections: An Exploratory Study." Legislative Studies Quarterly 22: 79-96.
Maisel, L. Sandy, Walter J. Stone, and Cherie Maestas. 2001. "Quality Challengers to Congressional Incumbents: Can Better Candidates Be Found?" In Paul S. Herrnson, ed., Playing Hardball: Campaigning for the U.S. Congress. Upper Saddle River, NJ: Prentice Hall.
Mayhew, David R. 1986. Placing Parties in American Politics: Organization, Electoral Setting, and Government Activity in the Twentieth Century. Princeton, NJ: Princeton University Press.
Renka, Russell D. 2001. "Elections with Rematched Opponents in the U.S. House of Representatives, 1970-1998." Paper presented at the annual meeting of the Midwest Political Science Association, Chicago.
Rohde, David W 1979. "Risk-Bearing and Progressive Ambition: The case of Members of the United States House of Representatives." American Journal of Political Science 23: 1-26.
Romero, David. 2004. "The Prospects-Based Dynamics of the House Candidacy Decision." American Politics Research 32: 119-41.
Schlesinger, Joseph A. 1966. Ambition and Politics: Political Careers in the United States. Chicago: Rand McNally
Squire, Peverill, and Eric R.A.N. Smith. 1984. "Repeat Challengers in Congressional Elections." American Politics Quarterly 12: 51-70.
Stolberg, Sheryl Gay. 2003. "Wooed for Congress, Fewer Will Say 'I Do'." New York Times, October 14, A22.
Stone, Walter J., and L. Sandy Maisel. 2003. "The Not So Simple Calculus of Winning: Potential U.S. House Candidates' Nomination and General Election Prospects." Journal of Politics 65: 951-77.
Tufte, Edward R. 1975. "Determinants of Outcomes of Midterm Congressional Elections." American Political Science Review 69: 812-26.
_____. 1978. Political Control of the Economy. Princeton, NJ: Princeton University Press.
Wrighton, J. Mark, and Peverill Squire. 1997. "Uncontested Seats and Electoral Competition for the U.S. House of Representatives Over Time." Journal of Politics 59: 452-68.
ANDREW J. TAYLOR, NORTH CAROLINA STATE UNIVERSITY
ROBERT G. BOATRIGHT, CLARK UNIVERSITY
Received: November 15, 2004
Accepted for Publication: December 13, 2004