Strategic Candidate Exit

The vast majority of research on congressional elections is limited to candidates on the ballot. Scholars of presidential elections have, however, given more attention to early and evolving campaign dynamics, perhaps because of the higher visibility of presidential candidates. Several studies show how the ‘money primary’ in presidential elections matters for success (Adkins and Dowdle Reference Adkins and Dowdle2002; Aldrich Reference Aldrich1980; Goff Reference Goff2005; Mayer Reference Mayer2003; Norrander Reference Norrander2006). Bawn et al. (Reference Bawn, Cohen, Karol, Masket, Noel and Zaller2012) and Cohen et al. (Reference Cohen, Karol, Noel and Zaller2008) have also generated new interest in how endorsements influence nominations, broadening our view of how prenomination activity matters for the choices on the ballot in presidential elections. Data hurdles have left the early action at other levels of office – even congressional office – far more opaque by comparison.Footnote ⁵

Bonica (Reference Bonica2017) and Hassell (Reference Hassell2018) provide two important exceptions to the focus on the ballot and are the only studies of candidate exit at the congressional level. Both highlight the role of money, with Bonica pointing to early money and Hassell focusing on party-connected donors. Bonica finds that, from 2010 to 2014, an early fundraising deficit is associated with candidate exit. Hassell analyzes elections from 2004 to 2014 and shows that candidates who receive more money from national party donors are less likely to drop out. Yet many questions remain because there is no roadmap for studying or even defining dropouts. For example, Hassell defines dropouts as those who raise money in two quarters, whereas Bonica includes those who raise money in the first 90 days of their candidacy.Footnote ⁶

The main concern here is which candidates are most sensitive to early fundraising disparities. Scholars have long cared about how the behaviour of experienced candidates differs from their inexperienced counterparts (Banks and Kiewiet Reference Banks and Kiewiet1989; Canon Reference Canon1993; Carson and Roberts Reference Carson and Roberts2005; Cox and Katz Reference Cox and Katz1996; Hirano and Snyder Reference Hirano and Snyder2019; Jacobson Reference Jacobson1989; Jacobson and Kernell Reference Jacobson and Kernell1983). Experienced candidates are commonly referred to as ‘quality’ candidates because they have run successful campaigns and served in office before. They navigate the electoral environment more strategically than their inexperienced counterparts and are ultimately more likely to win. Experienced candidates are perhaps more attuned to preelection signals of viability as well.

Initial impressions are key, and the first fundraising report is critical for attracting attention and gaining momentum. One of the laws of fundraising is that money begets money (Biersack et al. Reference Biersack, Herrnson and Wilcox1993; Smidt and Christenson Reference Smidt and Christenson2012). Fundraising reports provide a publicly available metric of how candidates stack up in an era where fundraising is a widely used indicator of campaign strength. A variety of political actors, including candidates, donors, journalists, and party elites, monitor campaign dollars and use them to size up a race. Money is objective; candidates either have more or less, and observers are keenly aware of who raises a lot and who raises a little. Candidate behaviour certainly suggests that they care about the money chase.

Changes in the political and electoral landscape provide an important backdrop for why early money has become increasingly salient. The time horizon of congressional elections has shifted, with more of the action happening earlier in the cycle. Primary elections have become more relevant as a growing number of lawmakers represent safe districts. In many districts, the most serious potential or actual competition happens well before November. Technological innovations have also transformed the nature of campaign finance reporting. FEC reports have been publicly available for decades, but with the rise of the Internet, transparency is now coupled with accessibility. Detailed fundraising and spending data can be obtained in seconds. Those who want to follow the money can do so far more easily today.

Fundraising patterns have indeed changed in important ways in recent decades. For one, successful candidates raise money earlier and earlier in the cycle. Among those with prior office experience, 70 per cent raised money by the end of the year before the election in 2020 and 2022, compared to 36 per cent in 1980.Footnote ⁷ In addition, the amount of early money raised by leading fundraisers has skyrocketed. Top fundraisers in competitive open-seat races raised around $100,000 in their first quarter in the 1980s and the 1990s. This amount grew to $200,000 to $300,000 in the 2000s and 2010s and surpassed $300,000 in the 2020s (all in 2021 dollars). Plenty of candidates raise little to no money (Thomsen Reference Thomsen2023), but those who want to win post higher totals today, and they do so earlier in the cycle.

Experienced candidates have long been known to make different entry decisions. The question here is instead how their exit decisions differ. Two main hypotheses are examined in the sections below. The first is that experienced candidates who fail to make fundraising inroads are expected to be more likely to drop out because of what a weaker haul conveys. The second addresses this relationship over time. Early money has attracted more attention as competition has shifted earlier in the cycle and as relevant political actors have turned to reports to infer viability. Experienced candidates who struggle to fundraise are expected to be more likely to drop out in the post-2000 era because of the heightened salience of early money.

It is not at all obvious that experienced candidates will make different exit decisions. Experienced candidates do not enter races lightly. Indeed, the foundation of the strategic candidate entry framework is that experienced candidates are more knowledgeable about their electoral prospects prior to their entry, which explains why they select the races they do. They might similarly have a better sense of how much money they can raise or decide that the opportunity is too good to pass up. Moreover, Hassell (Reference Hassell2018) finds that experienced House and Senate candidates are either less likely to drop out than inexperienced candidates or equally likely to do so in races where there are no disparities in experience.Footnote ⁸ It is thus not a foregone conclusion that patterns of candidate exit will differ in the ways outlined above.Footnote ⁹

Others might instead suggest that the argument is wholly consistent with the strategic candidate entry framework. Candidates likely do drop out because they think they will lose. However, the focus on exit differs from the vast majority of studies on whether experienced candidates run for office and why they run when they do. Scholars have given little attention to how experienced candidates navigate the preprimary period after they decide to enter. Dropouts provide a window into the information that the most serious and strategic candidates use to weigh the costs and benefits of staying in the race. Entry and exit decisions alike are important because high-quality candidates are the lynchpin of democratic government.

In sum, the main expectations pertain to why dropout decisions are likely to vary across candidates and over time. An additional point is that we must look earlier in the cycle – and beyond the ballot – to understand how money matters in different ways across candidates. The assumption that underlies much of the research on strategic candidate entry is that candidates run when they think they can win, but less attention has been given to why the calculus of candidacy might change and how the early choices of donors are associated with candidate behaviour. Today, we would overlook important reasons why the ballot looks the way it does if we failed to consider the significance of early fundraising and its implications for candidate exit.

Data

The analysis focuses on dropouts in U.S. House races from 1980 to 2022. The main obstacle to studying dropouts is data collection, because it is difficult to construct samples of those who initiate a candidacy but withdraw before the election. Here, dropouts include those who filed to run with the Federal Election Commission and raised money but did not appear on the primary ballot.Footnote ¹⁰ One advantage of this measure is that these individuals have taken a costly and public step of running for office, and it captures the more serious dropouts in the preprimary period.Footnote ¹¹ It excludes others who initiated a candidacy in another way but did not appear on the ballot, though the importance of money for viability makes fundraising an appropriate starting point.

I draw on two datasets in the analyses below.Footnote ¹² First, I collected the full sample of on-ballot primary candidates from 1980 to 2022 from the America Votes series, the FEC website, and the State Board of Elections’ websites. Second, I used FEC data to identify the candidates who raised money. The dropouts are those in the FEC dataset but not the on-ballot dataset.Footnote ¹³ There are nearly 26,000 non-incumbents in the sample. The dataset includes all quarterly and preelection receipts raised by each candidate within the election cycle.Footnote ¹⁴ It also includes whether the individual held previous elected office, the most commonly used measure of candidate ‘quality’ (Jacobson Reference Jacobson1989).Footnote ¹⁵

Between 1980 and 2022, a total of 2,283 candidates filed with the FEC and raised money but were not on the primary ballot.Footnote ¹⁶ The number of dropouts has increased dramatically over time, ranging from a low of 22 in 1990 to a high of 331 in 2022 (or from 2 per cent of non-incumbents on the ballot in 1994 to 19 per cent in 2022). This increase further illustrates the role of fundraising in campaigns today compared to the 1980s and 1990s, when potential candidates could test the waters in less public ways. Dropouts are also more likely to have served in a lower-level office: 28 per cent of dropouts are experienced candidates, versus 22 per cent of non-incumbents on the ballot (p < 0.01). Of the 649 experienced dropouts, 253 were in open seats, compared to 228 and 168 in challenger-party and incumbent-contested primaries, respectively.

The next section examines early money and strategic candidate exit across this four-decade period. The dependent variable is coded 1 if the individual raised money but dropped out before the primary election and zero if they were on the ballot. The main independent variables are the candidate’s share of early money raised and whether they held prior elected office. I use the amount in the candidate’s first report as a share of the leading early fundraiser’s total. The candidate in the race with the highest value in their first report thus has a value of one.Footnote ¹⁷ This measure captures my interest in the candidate’s first fundraising impression and their relative strength at the race level. I include interactions between fundraising, prior office experience, and time period (1980-98 and 2000-22) to test the expectations outlined above.Footnote ¹⁸

I control for several factors that affect competition and entry decisions. First, I account for seat type, district partisanship, and the state party rules governing preprimary endorsements (Canon Reference Canon1993; Hirano and Snyder Reference Hirano and Snyder2019; Jewell and Morehouse Reference Jewell and Morehouse2001).Footnote ¹⁹ Each model includes a binary variable for open-seat and challenger-party races, with incumbent-contested races as the baseline. Jacobson’s (Reference Jacobson2015) presidential vote share data are used to measure district partisanship. I include indicators for competitive and safe districts, with unfavourable districts as the baseline. In addition, the number of state legislators in a state may matter for the supply of potential candidates. I also control for the month of the primary election, the number of on-ballot candidates, gender, and party. State and year fixed effects are included in all of the models.Footnote ²⁰

The Decision to Drop Out

The results are presented in Table 1.Footnote ²¹ In Models 2 and 3, the sample is divided by time period; Model 4 includes an interaction between time period and early money. Across models, candidates who raise more early money are less likely to drop out, which conforms to Bonica’s (Reference Bonica2017) and Hassell’s (Reference Hassell2018) findings. In addition, experienced candidates are more likely to exit than inexperienced candidates, on average.

Table 1. Early Money, Experience, and Candidate Exit

Note: Results are from logistic regressions from 1980 to 2022. Standard errors are clustered at the race level. The dependent variable is whether the candidate dropped out of the race. ^†p < 0.10, *p < 0.05, **p < 0.01.

The main consideration here is whether candidates respond differently to early fundraising disparities. Experienced candidates are expected to be acutely aware of their position in the financial horse race. The frontloading of competition and the increased reliance on early money as a signal may mean that early money matters more for candidate exit today than in the pre-2000 era. The results are in line with expectations. Across models, the likelihood of dropping out decreases for experienced candidates who raise more early money relative to their competitors. In addition, the interaction between early money and the post-2000 era is negative in Model 4, indicating that the association is even starker in the latter half of this period.

Figure 1 plots the predicted probability of exit in open seats for experienced and inexperienced candidates. The pre- and post-2000 periods are shown in the left and right graphs, respectively. For experienced candidates who raise 20 per cent of the top fundraiser’s receipts, the likelihood of dropping out increases from 7 per cent in the pre-2000 era to 24 per cent in the post-2000 era. These values are 4 and 15 per cent, respectively, for inexperienced candidates. Put differently, for experienced candidates, a shift from raising zero per cent of the top fundraiser’s receipts to leading the race decreases the likelihood of exit by 7 percentage points in the pre-2000 period but by 27 points in the post-2000 period. For inexperienced candidates, a similar shift decreases the likelihood of exit by 2 points in the pre-2000 period and by 11 points in the post-2000 period. Experienced candidates are more sensitive to fundraising disparities across this period, but they are even more sensitive in the current money-driven era.

Figure 1. Probability of Dropping Out, By Early Money and Experience.

Note: Predicted values are generated from Model 4 in Table 1.

An additional consideration is whether the results are instead reflective of changes in primary election calendars or in primary election filing deadlines. I draw on Boatright’s (Reference Boatright2013) primary elections data to examine this possibility further. Some states have moved their primary dates and filing deadlines earlier in the cycle, and the question is whether dropout patterns differ across groups. In Table 2, the sample is split by candidates in states where the primary election month did and did not change over time (Models 1 and 2) and in states where the primary ballot filing deadline did and did not move earlier in the cycle (Models 3 and 4).Footnote ²²

Table 2. Changes in Primary Election Dates and Filing Deadlines

Note: Results are from logistic regressions from 1980 to 2022. For Models 3 and 4, the data are from 1984 to 2022. Standard errors are clustered at the race level. The dependent variable is whether the candidate dropped out of the race. ^†p < 0.10, *p < 0.05, **p < 0.01.

The patterns are consistent with those above. Experienced candidates who struggle to raise money are more likely to drop out in states where the primary calendar both did and did not change over time (Models 1 and 2). They are also more likely to drop out in states where the primary ballot filing deadline did and did not move earlier in the cycle (Models 3 and 4). The interaction between early fundraising disparities and post-2000 is negative across models as well. The findings are suggestive of a new temporality of congressional elections where candidacies are benchmarked to FEC reporting timelines and quarterly deadlines rather than state primary election dates or state filing deadlines, but this question is beyond the scope here.

With respect to the control variables, the probability of exit is higher in open seats, challenger-party primaries, and in favourable partisan districts (Hirano and Snyder Reference Hirano and Snyder2019; Stone and Maisel Reference Stone and Maisel2003). The likelihood of dropping out decreases as the number of primary candidates increases. Women candidates are less likely to exit the race as well, though this result is driven by Democratic women. The latter result differs from that in Niven’s (Reference Niven2006) analysis of Florida state legislators in 2000 and 2002, which may reflect changes over time in the entry, support, and success of women candidates, particularly on the Democratic side (Crowder-Meyer and Cooperman Reference Crowder-Meyer and Cooperman2018; Teele et al. Reference Teele, Kalla and Rosenbluth2018; Thomsen Reference Thomsen, Dodd, Oppenheimer and Evans2021; Thomsen and Swers Reference Thomsen and Swers2017).

In sum, the results provide support for the argument that experienced candidates use early money as a signal of viability in different ways than inexperienced candidates. They are even more responsive to resource disparities in the contemporary money-driven era. What is more, the soaring demands of early fundraising mean that top fundraisers today need to raise far more money to take the top spot. A long line of work has tracked the entry of experienced candidates, but exit decisions have important consequences for the trajectory of races and the choices on the ballot as well.

Early Money and Career Incentives

This section leverages variation in career incentives to explore whether electoral considerations instead drive the exit of experienced candidates. The analyses are limited to the dropout decisions of nearly 800 sitting state legislators who ran for the U.S. House from 2000 to 2016.Footnote ²³ State legislators face different opportunity costs, and they vary on a number of dimensions that influence progressive ambition. One factor in particular is whether they are up for re-election to their state legislative seat. I use Klarner’s (Reference Klarner2018) dataset of state legislative elections and Fouirnaies and Hall’s (Reference Fouirnaies and Hall2022) data on term-limited legislators to code whether sitting state legislators are or are not up for re-election in the election cycle of their congressional campaign.

An initial look at what state legislators did instead shows that far more dropouts ran for or remained in the state legislature than those who stayed in the U.S. House race (73 v. 30 per cent, respectively; p < 0.01). Unsurprisingly, more dropouts were up for re-election and risked losing their state legislative seat: 66 per cent of dropouts were up for re-election, compared to 53 per cent of those who stayed in the congressional race (p < 0.01). Another difference is that state legislators who stayed in the U.S. House race were more likely to be term-limited than dropouts (20 v. 10 per cent; p < 0.01) and thus did not risk losing their seat because they were ineligible to run again. A similar percentage of dropouts and non-dropouts were in the middle of their state legislative term and did not risk losing their seat (24 v. 27 per cent; not significant).

State legislators who risk losing their seats clearly have a different set of incentives than those who do not. However, I also examine how re-election status and the risk of seat loss are associated with dropping out across fundraising levels. While those who raise less money might be sensitive to losing their state legislative seat, re-election status might have a limited impact on exit decisions when candidates fare better in early fundraising. I ran the same models as the previous section and included a binary variable for whether state legislators were up for re-election. The results are provided in Table 3. Model 1 includes all sitting state legislators, and I split the sample into those who raised less than the median of $110,000 in their first quarter and those who raised more in Models 2 and 3, respectively.

Table 3. Dropout Decisions Among Sitting State Legislators, By Re-Election Status and Early Money

Note: Results are from logistic regressions from 2000 to 2016. The dependent variable is whether the candidate dropped out. ^†p < 0.10, *p < 0.05, **p < 0.01.

Similar to the bivariate patterns, the results for the full sample in Model 1 indicate that those who are up for re-election are more likely to drop out than those who are not. However, when the sample is divided into state legislators who raised more and less than the median sitting state legislator, it is clear that the relationship between re-election status and dropout decisions is driven by those who struggle to raise money. Among state legislators who raise less, the predicted probability of dropping out is 29 per cent for those who are up for re-election, compared to 19 per cent for those who are not. Among those who fare better in fundraising, the relationship is not significant. The predicted probability of dropping out in this sample is 2 per cent for both those who are and are not up for re-election.

An additional consideration is whether state legislators who drop out are of lower quality than those who remain in the race. Indeed, experienced candidates raise more money than inexperienced candidates in part because they are higher-quality candidates and more likely to win. Most measures of quality are binary indicators of whether the candidate has held any elected office; additional measures of relevant office experience include those who have held state legislative office (Hirano and Snyder Reference Hirano and Snyder2019). While it is difficult to measure quality within state legislators, I use Bucchianeri et al.’s (Reference Bucchianeri, Volden and Wiseman2025) state legislative effectiveness scores (SLES) to examine the effectiveness of dropouts and on-ballot candidates. State legislative effectiveness scores follow Volden and Wiseman’s (Reference Volden and Wiseman2014) measures of effectiveness (LES) at the congressional level. The LES is a comprehensive measure combining fifteen metrics of the bills each member sponsors, how far they move through the lawmaking process, and their relative substantive significance.

Among sitting state legislators who run for Congress, dropouts are no less effective than those who remain in the race. The median SLES score for dropouts is 1.03, compared to 0.99 for those on the ballot (not significant). Nor are dropouts different from on-ballot candidates on a host of SLES metrics, including their current SLES or lagged SLES values. Dropouts even have slightly higher scores ‘relative to expectations’ (2.12 v. 1.96; p < 0.05), but this may reflect the fact that on-ballot candidates are campaigning for another office during the session. Overall, legislative effectiveness scores are similar for dropouts and on-ballot candidates, and there is little evidence that dropouts are less effective lawmakers.

Nor do a host of other metrics suggest that sitting state legislators who drop out are of lower quality. Eleven per cent of dropouts are party leaders, compared to 15 per cent of on-ballot candidates. Thirty-three per cent of both dropouts and on-ballot candidates are committee chairs. Forty-six per cent of dropouts are on power committees, compared to 52 per cent of on-ballot candidates. Dropouts and on-ballot candidates have been in office for an average of 3.6 and 3.7 terms, respectively. In terms of electoral strength, 19 per cent of dropouts and 26 per cent of on-ballot candidates were unopposed in their prior general election. None of these differences is statistically significant. I am unable to compare these metrics in the full sample, but at least among sitting state legislators, re-election incentives and legislator productivity appear to play a smaller role in exit decisions than early money.

More generally, the results suggest that our understanding of strategic candidate behaviour is diminished by a ballot-centred lens. Candidates do not just enter; they also update their expectations in the first months of their campaign and decide to exit or stay in the race. The preprimary period has received less attention at the congressional level, but it comes with its own uncertainty that candidates must navigate. One key part of this uncertainty is tied to fundraising, and the utility of running for office depends in part on post-entry fundraising success. The dynamics that unfold in the critical months before the election have important consequences for changes in perceived viability and ultimately the choices that voters have.

Implications for Competition

The final task is to put dropout decisions in a broader context and consider the implications for competition. The first question is how the number of experienced candidates on the ballot is associated with the quality of primary competition. If additional experienced candidates do not result in closer races, their exit is a less obvious concern. I draw on both vote share and fundraising measures of competition, where primaries are coded as competitive if the top vote-getter and top fundraiser received less than 57.5 per cent of votes and receipts, respectively (Hirano and Snyder Reference Hirano and Snyder2019; Thomsen Reference Thomsen2023). I include indicator variables for the number of experienced candidates, ranging from zero to four or more.Footnote ²⁴ The sample is limited to primaries without an incumbent in safe or competitive districts. I include binary variables for open seats and safe districts and an interaction between the two; challenger-party primaries and competitive districts are the baseline.

The results are shown in Table 4. Models 1 and 2 include the full sample, and Models 3 and 4 are limited to primaries with at least one experienced candidate on the ballot. Across models, primaries with more experienced candidates on the ballot are far more likely to be competitive. There is a strong association between the number of experienced candidates and primary competition. The size of the coefficients increases with each experienced candidate on the ballot. Open seats are also more likely to be competitive, which is consistent with a long line of research. District partisanship plays a smaller role in part because the sample is limited to safe and competitive districts where candidates have a shot at winning the general election.

Table 4. Experienced Candidates on the Ballot and Primary Competition

Note: Results are from logistic regressions from 1980 to 2022. The dependent variable is whether the primary is competitive with vote share and fundraising measures of competition. Models 1 and 2 include primaries without an incumbent in safe or competitive districts, and Models 3 and 4 are limited to primaries with at least one experienced candidate on the ballot. ^†p < 0.10, *p < 0.05, **p < 0.01.

Predicted values for open-seat primaries are presented in Figure 2. With both vote share and fundraising measures of primary competition (in the left and right graphs, respectively), each additional experienced candidate is associated with an increase in the likelihood of a competitive race. The largest jump between values is from one to two experienced candidates on the ballot. With the vote share measure (left graph), the likelihood the race is competitive increases from 38 per cent when there is one experienced candidate on the ballot to 80 per cent when there are two. With the receipt share measure (right graph), the likelihood the race is competitive increases from 27 per cent when there is one experienced candidate on the ballot to 61 per cent when there are two.

Figure 2. Likelihood of Competitive Primary by Number of Experienced On-Ballot Candidates.

Note: Predicted values are generated from Models 1 and 2 in Table 4.

A related question is what the ballot looked like in races where an experienced candidate dropped out. Figure 3 shows the number of experienced candidates on the ballot in primaries with at least one experienced dropout. The sample is again limited to safe or competitive districts where candidates could win the general election. The left graph includes races with no same-party incumbent (challenger-party and open seats), and the right graph only includes open seats. In primaries with an experienced dropout, there was either zero or one experienced candidate on the ballot in 63 per cent of primaries with no same-party incumbent. Even in the best-case scenarios of electoral competition – open-seat races in safe or competitive districts – 46 per cent of primaries with an experienced dropout had either zero or one experienced candidate on the ballot.Footnote ²⁵

Figure 3. Number of Experienced Candidates on the Ballot in Primaries with an Experienced Dropout.

Note: The graphs show the number of experienced candidates on the ballot in primaries with an experienced dropout. The y-axis is the percentage of primaries that fall in each category and thus total 100 per cent. The left graph includes all primaries with no incumbent, and the right graph is further limited to open seats. The sample is limited to safe and competitive districts.

It is impossible to know whether outcomes would have been different had experienced dropouts stayed in, but some races would have likely been closer. There were 180 primaries where a non-incumbent won the general election and had an experienced dropout in the primary. For eighty-four of the winners, there was either zero or one experienced candidate on the primary ballot; in these races, 40 per cent of the primaries were close with the vote-share measure. In the remaining ninety-six races with two or more experienced candidates on the primary ballot, 89 per cent were close. If we value competition, the presence of experienced candidates is a clear boon. And unsurprisingly, 79 per cent of these districts elected an experienced candidate when there were two or more experienced candidates on the primary ballot, compared to 58 per cent of districts when there were zero or one.

Some might instead cite the value of dropouts for winnowing the field, similar to presidential elections. But the race for Congress is not at all like that for the presidency. Lawmakers face relatively few close elections in their careers. Of all House members in office from 1980 to 2018, nearly one-fourth did not face a single competitive primary or general election during their tenure in office. Another 30 per cent had a total of one, with 90 per cent of those being their first election as non-incumbents. Constituents may be satisfied with their representatives, so this is not inherently problematic, but others have similarly suggested that competition is good when lawmakers are selected initially (Hirano and Snyder Reference Hirano and Snyder2019). At some point, either prior to or during an officeholder’s tenure, competition for votes is essential for democratic government.