Experience, institutions, and candidate emergence: the political career returns to state legislative service

Abstract

More than half of the current members of the US Congress served in their state legislature prior to holding federal office. We quantify the relationship between state legislative service and career progression to Congress. Using close elections for exogenous assignment of political experience across otherwise similar candidates, we show that serving in the state legislature more than doubles an individual's probability of eventually contesting a Congressional seat relative to a similar candidate who lost in a comparable election; it also doubles the individual politician's probability of eventually winning a Congressional seat. State legislatures thus create national politicians out of otherwise marginal political entrants. We then show that the effect of state legislative service on career progression is larger in more professionalized legislatures, highlighting the role of institutions in facilitating political career progression. Our results hold important implications for representation and accountability, and confirm that prevailing institutions can affect political selection via career progression.

Political selection is the process by which individuals become candidates for office, ultimately shaping the pool of elected officials. A large literature examines individual factors that affect candidate emergence, finding relationships with ambition (Maestas et al., 2006), gender (Fox and Lawless, 2014; Brown et al., 2019; Bernhard and Benedictis-Kessner, 2021), personality (Dynes et al., 2019), the returns to holding office (Querubin and Jr, 2009), and wealth (Carnes, 2018), among others. Institutional factors, such as the role of party preferences, are also linked to candidate emergence (Fox and Lawless, 2011; Dittmar, 2015; Butler and Preece, 2016; Hassell, 2017).

In tiered political systems, progressive ambition—existing politicians’ aspiration for holding higher office—is intricately linked to candidate emergence for higher levels of elected office. In the United States, for example, more than half of the members of the 116th Congress had previously served in state legislatures. But does lower-level political experience enhance an individual politician's capacity to run for, or win, higher office—and if so, under which conditions?

Drawing on both the individual and institutional determinants literatures, this paper investigates the relationship between state legislative experience, and the characteristics of the institutions in which politicians gain that experience, on candidate emergence for the US Congress. We first quantify the positive causal effect of state legislative experience on candidacy for and election to Congress using data that cover the universe of elections for single-member state legislative districts from 1967 to the present linked to federal elections records. An additional term in the state legislature increases the probability of a politician ever contesting a federal election by nearly four percentage points—approximately equal to the mean probability that state legislative candidates are ever observed contesting a Congressional primary or general election. The majority of this effect is realized in the decision to run for House primary and then general elections. An additional term in a state legislature also doubles the likelihood of winning a Congressional seat and serving as a federal legislator.

We then focus on the ways in which institutional structures affect the experience–progression relationship. Legislative professionalism—a concept that characterizes the nature of the experience as a state legislator via salaries, expected time commitment, and the resources they have available for staff (Squire, 1988, 2007)—proves to be a substantive institutional determinant of state legislators’ career progression.¹ Winning an additional state legislature term in a one standard deviation more professional legislature leads to that person being six percentage points (or 25 percent) more likely to run for Congress than a similar candidate who won in state whose legislature has average professionalization. More professional legislatures see nearly a 50 percent increase in the unconditional rate at which their members win House elections.

Contributing to a literature on how the financial returns to holding office result in political selection (see Dal Bó and Finan, 2018), we also find that states with more stringent regulations around “cooling off” periods—those which restrict private sector lobbying activity after political service—produce more candidates for higher office. This likely contrasts with expectations of increased retention as a result of revolving door policies, which are a consequence of artificially reduced opportunity costs of remaining a state legislator. Our results thus provide new empirical evidence for a tradeoff between reducing regulatory capture and retaining experienced legislators.

In the following section, we discuss existing research on candidate emergence and political selection. We then highlight the competing theoretical predictions in this work and how it relates to institutional features of state legislatures. We then discuss the data and research design before presenting the effect of winning a seat in a state legislature on future office holding. Next, we interrogate the mechanisms behind the differences in state legislature professionalism, including the constituent parts of professionalism and the role played by variation in post-political careers.

1. Institutions and political selection

There is a sizeable literature investigating the determinants of candidate selection at varying levels of government and across institutional contexts. Underscoring the importance of understanding political selection are the established links between politicians’ backgrounds and policy outcomes, with broader implications for representation of diverse societies (Lawless, 2004; Lee et al., 2004; Washington, 2008; Carnes, 2013; Miler, 2018; Thompson et al., 2019). Our paper is also related to the literature examining self-selection into politics (e.g., Mattozzi and Merlo, 2008; Fox and Lawless, 2011; Dal Bó and Finan, 2018). This paper focuses on how institutions affect the decision to run for higher office conditional on already serving in lower level legislatures.

We first examine US state legislators and whether state legislative service causes individuals to eventually run for Congress. As the most common path to becoming a member of Congress is through state legislative service (Carson, 2005), this setting is of per se importance for understanding political selection and candidate emergence at the national level. In line with previous work, we expect there to be a positive effect of winning a state legislative seat on the probability of running for Congress (Wasserman, 2018; Brown et al., 2019; Bernhard and Benedictis-Kessner, 2021). We also examine potential variation in this positive effect through analyzing “windows of opportunities”—that is, whether there are diminishing marginal returns to being as-if randomly assigned additional years of legislative experience. We hypothesize that the effect of state legislative experience is not strictly linearly related to higher probability of seeking higher office: the benefit will primarily come through the first victory.

1.1 State legislatures and career incentives

We expect that institutions that place restrictions on legislators’ careers have the potential to produce heterogeneity in the effect of state legislative office holding on higher office seeking. The first and most obvious of these career limitations comes through term limits. Term limits are a recent feature in the structure of state legislature service that research has shown to have broad effects on legislature performance and legislator behavior (Kousser, 2005; Carey, 2006; Fowler and Hall, 2014; Fahey, 2018). Theoretical predictions on the relationship between experience and higher office seeking are ambiguous. Term limits affect the potential amount of total state legislative experience a politician can accrue, which could encourage state legislators to consider higher office sooner than they otherwise would have. Term limits also work the other direction—they might preclude the accrual of sufficient experience for a viable Congressional candidacy. Term limits might also change political selection into state legislature candidacy due to an institutionally limited career, which could lead to the selection of more progressively ambitious candidates who expect to seek higher office after reaching a term limit, or might increase the selection of candidates who view political service as temporary. Empirically, our comparison will compare the magnitude of the effect among states with and without term limits.²

The final career restriction we consider is policies that place limits on the outside options of state legislators. Opportunity costs for many public servants are high, given the typically lower salaries in the public sector. Some evidence, largely outside of contemporary American politics, suggests there is a substantial financial return to holding office (Eggers and Hainmueller, 2009; Querubin and Jr, 2009; Dal Bó and Finan, 2018). These expected future returns factor into a politician's decision making, and given the substantial differences across states in legislator resources and salary the effects should be particularly pronounced. If legislators’ post-political career options and the commensurate financial returns are limited, pursuing higher office will be a more attractive path than other options. In related research, Strickland (2019) finds that the demand for legislators-turned-lobbyists decreases once the market becomes saturated, leading to fewer state legislators entering lobbying as their expected value decreases. Egerod (2017) finds US Senators strategically retire when, ex ante, they expect higher returns in the lobbying sector. Consistent with this work, we hypothesize that stricter restrictions on outside options will lead to more candidates running for higher office.³

1.2 Legislative professionalism

Former state legislators from professionalized legislatures are, on average, more effective members of Congress (Volden and Wiseman, 2014), holding implications for the quality of representation and legislative outcomes. We then investigate institutional variation in generating candidates for federal office, with a particular focus on legislative professionalism. Professional legislatures are those in which legislators are, most broadly, provided resources and see their legislative role as a full-time job (the United States Congress is typically considered the pinnacle of professionalized legislatures).⁴

Existing research on how legislative professionalism directly relates to political selection includes Maestas et al. (2006), building on Rohde (1979), who argue that legislative professionalism enters into candidate emergence through interacting with ambition (see also Fowler and McClure, 1990). They find that professionalism affects ambition through a survey of sitting, elected state legislators that asks the legislator to assess their own ambition (Maestas et al., 2006). More professional legislatures tend to see their more ambitious legislators run for higher office. However, this leaves open the question of which features of professionalism lead to variation in higher office seeking, or whether the effect is purely through selection (i.e., ambitious types are more likely to select into professional legislatures in the first place). There is also strong evidence that professional legislatures attract “better”—that is, more knowledgeable—legislators (Butler and Nemerever, 2019).

There are various competing predictions about the relationship between professionalism and selection into higher office. Among state legislators—specifically, individuals who have already run for and won an elected office—characteristics of the state legislature may generate variation in higher office seeking by affecting incentives for elected members to remain a legislator or to pursue higher office or other opportunities (e.g., lobbying). The mechanisms behind this relationship fall into two broad categories: (1) the legislature's ability to retain incumbent legislators; (2) the legislature's ability for incumbent legislators to increase their probability of winning a future congressional election. Traits of professionalism affect both of these mechanisms.

Professionalized legislatures feature more resources for their members such as salaries and funds for full-time administrative and/or legislative staff.⁵ Professional legislatures should then be able to attract and retain better legislators, given the more attractive features of the job (Squire, 1988; Ferraz and Finan, 2008; Fisman et al., 2015).⁶ On the other hand, as suggested by Maestas et al. (2006), ambition may be correlated with selection into professional legislatures, which then may result in higher turnover among the pool of state legislators as they are more likely on average to leave the state legislature to run for higher office.

The other broad mechanism through which professionalism affects candidate emergence for Congress is through providing state legislators advantages if they do wish to run for higher office. A feature of professional legislatures is the ability for its members to treat it like a full-time job (competitive salaries, long sessions, and no need to maintain a second job or business) and access to resources for members to build up a personal vote (e.g., Fenno, 1978), including professional legislative staff in the state capitol and in district offices. In other words, professional legislatures permit their members to develop experience in politics similar to that of a member of Congress (Berkman, 1994; Fiorina, 1994). This proposition has not been subject to much empirical scrutiny, however.⁷ Ex ante, the expected direction of the relationship between service in professionalized and higher office seeking is mixed.

Similarly, the expectations surrounding the heterogeneous relationship between professionalism, its individual component parts, and higher office seeking is mixed. For instance, a longer legislative session allows legislators to better establish electoral constituencies and credit claim on policy, but it also limits available time for campaigning and establishing links to the relevant party actors. Additionally, this component of professionalism will be strongly related to other components, such as availability of staffing. Longer sessions in legislatures with more staffing resources enables more policymaking and credit claiming opportunities, but also presents a better opportunity for motivated legislators to develop a career in state public service (and as a result may lead to less upward candidacy). We investigate these heterogeneities below using data that account for the constituent elements of professionalism (Bowen and Greene, 2014).⁸

2. Context and data sources

We use the State Legislative Election Returns (SLER) data set by Klarner et al. (2013). The SLER contains candidate-level election returns for US state legislature elections from 1967 to 2010; we use information on candidate's name, state, district, and the chamber they are running in, as well as total vote counts and the candidate's party. We then merge the SLER to records from Congressional primary and general election returns from 1968 to 2018. This process begins with a fuzzy matching algorithm based on first and last name fields, and requires matches to occur within the same state. The vast majority of those found are perfect matches; the remainder are reviewed manually to ensure accuracy, resulting in relatively little ambiguity in the final matches found.⁹ For later analyses, we combine these data on state legislatures’ professionalism and term limits, as well as individual-level measures of ideological extremity. For the analysis, we restrict our sample to the first appearance of each individual in the data. This is done in order to not overweight individuals who appear in multiple state legislative elections but (by construction) have the same career outcome; results are not sensitive to this choice in any way.¹⁰ Limiting to these races also allows a comparison among candidates without prior experience (and knowledge gained) in congressional campaigning.

Our sample consists of over 35,000 observations representing candidates across more than 20,000 unique contested elections. Table 1 contains summary statistics of the data set, showing that 50 percent of the candidates are Democrats, the average candidate has little previous state legislative experience, and four percent of these candidates are ever observed running for a Congressional primary or general election, and only two percent ever run for any federal general election.

Table 1. Summary statistics

3. Empirical strategy

We first estimate the relationship between state legislative experience and career progression to Congressional candidacy and officeholding. To do this, we conceptualize closely won elections as career shocks that assign an additional term of state legislative experience quasi-experimentally across otherwise-comparable individuals.¹¹

The regression discontinuity specification is as follows:

(1)$$\eqalign{Y_{ist} & = \alpha Won\; state \; seat_{ist} + \beta margin_{ist} + \gamma [ Won\; state \; seat_{ist}\times margin_{ist}] + X_{ist} \delta + \tau_{s} + \phi_{t} + \epsilon_{ist}}$$

where Y _ist stands for a set of binary outcomes which takes the value of one (and is zero otherwise) if state legislature candidate i in election at time t in state s ever: (1) stands as a candidate for a House primary, (2) wins a House primary, (3) stands as a candidate for a House general election, or (4) wins a House general election. Won state seat _ist, takes the value of one if candidate i won their state s legislature election in year t, and is zero otherwise. margin _ist is the margin by which the candidate won (taking negative values if the candidate lost), and the linear effect of the victory margin can vary flexibly on either side of the threshold via γ. X _ist includes the candidate's cumulative (observed) legislative experience, party, term length, a constructed measure of the total number of candidates in the election, and whether individual is contesting a special election.¹² Time-invariant state effects are captured by τ_s, and common shocks by election year in which candidates are contesting the state legislature election are captured by ϕ_t. ε_ist is the error term, and we cluster standard errors by state.¹³ α is the parameter of interest, capturing the regression discontinuity estimate of the relationship between a term of experience in the state legislature and entry and success in Congressional politics. For all analyses, we restrict the sample to single-member districts,¹⁴ the first time that candidates appear in our sample, and elections that fall within Calonico et al. (2019)'s mean-squared-error optimal bandwidth estimated separately for each outcome, specification, and sample used.

There has been substantial critique of the use of close elections for empirical identification since the estimation of incumbency effects in the US Congress by (Lee, 2008). Notably, Caughey and Sekhon (2011) and Grimmer et al. (2011) show that the local continuity assumption does not hold at election thresholds for the sample of candidates for US House elections from 1942 to 2008. The conclusion from this work is that selection increases as races get closer, particularly along a set of measures broadly related to candidates’ “structural advantage.” More recent work, however, shows that some of these concerns may be tempered when correcting for multiple hypotheses and in consideration of the inferential assumptions used (de la Cuesta and Imai, 2016). We are not aware of established evidence of sorting among state legislature elections; following the recommendations from the studies above, implications regarding multi-order polynomials and bandwidths from Caughey and Sekhon (2011), and best practices from Cattaneo et al. (2020), we undertake a battery of tests to show that there is no evidence of sorting akin to that for US House elections.

We first show the McCrary (2008) density test in Figure 1, with no evidence of manipulation around the threshold (p-value: 0.984). Table 2 tests for sorting at the identifying threshold and shows continuity in contemporaneous individual and election observables that should not be systematically related to the winner of the election. Specifically, we test for a relationship between a close win and individual and contextual measures of “structural advantage”—including the candidate's incumbency, prior experience, prior losses, the number of candidates in the election, as well as their party and their party's alignment with the sitting governor or the legislative majority party. As mentioned above, our sample is optimal-bandwidth restricted and uses a local linear specification to detect discontinuities as the threshold; we find no evidence across any of the individual or party-based measures of structural advantage—notably not among party alignment with the sitting governor nor with the contemporaneous state legislature's majority. We also show, in Appendix Figure 1, that our sample of close elections is proportionate to the share of total elections by state over this time period.

Figure 1. McCrary (2008) density test.

Note: Figure reports a depiction of the McCrary (2008) density test. The sample includes all contested elections. The p-value for the test of a discontinuity in the density is 0.984.

Table 2. Test of sorting at the identifying threshold

$^{\ast }{\rm p} < 0.1$; $^{\ast \ast }{\rm p} < 0.05$; $^{\ast \ast \ast }{\rm p} < 0.01$.

Note: This table reports estimates of the effect of an additional state legislative term on individuals’ career progression to Congressional candidacy. The dependent variable is equal to one if the candidate ever runs in the election listed in the column header and is zero otherwise. The sample contains all first-time state legislative elections within the optimal bandwidth given by the Calonico et al. (2019) algorithm. All regressions include state and election year fixed effects, a linear term for the vote margin and its interaction for having won the election, and the full set of candidate and election controls. Standard errors clustered by state legislative constituency are reported in parentheses.

3.1 Estimating the effect of state legislative experience on Congressional candidacy and representation

As primaries are the typical, but not only, path to Congressional candidacy, we first estimate the effect of a narrow victory in state legislative elections on subsequently running in and winning a House primary. Candidate emergence and competition in primary races is of broad interest in political science (e.g., Hall, 2015; Hassell, 2017) since it ultimately affects the pool of candidates for service in Congress—with implications for accountability of elected officials (Rohde, 1979). Figures 2(a) and 2(b) present the focal analysis graphically by plotting the rate at which state legislative candidates ever contest and win House primaries by the binned margin of victory they experienced in their first close state legislature election. A negative value on the x-axis indicates the candidate lost; a positive value indicates winning the focal state legislative election.

Figure 2. Regression discontinuity plots: running for and winning house primaries: (a) run in House primary and (b) win House primary.

Note: These plots show empirical regression discontinuity plots estimating the effect of winning a state legislature election on running for or winning House primary elections. The fitted lines are second-order polynomials. The x-axis is the fractional vote share margin ranging from − 1 to 1. The plots include 95 percent confidence intervals calculated from a linear regression of the raw data within the optimal bandwidths on each side of the cutoff.

We see a clear effect of winning a state legislature seat on running in a House primary, winning a House primary, and then running in a House general election.¹⁵ Relative to marginal losers, marginal winners exhibit a more than doubling of the likelihood of ever contesting a Congressional election. These differences, which mirror the results from regressions, consistently approximate a doubling of the probability of ever contesting up of the course of a politician's career.

Figure 3 depicts the effect of winning a state legislature election on whether the individual ever runs in a House general election. Not only are candidates much more likely to run in House primaries upon narrowly winning a state legislative election, they are also more likely to win those primaries and become a candidate in the general election. The effect sizes are similar to primaries as well, almost doubling the probability of running if they win close state legislative elections. But do these candidates ultimately win House elections? Put differently, does state legislative experience influence the pool of Congressional candidates as well as those ultimately elected to Congress?

Figure 3. Regression discontinuity plots: running for House seats.

Note: This plot shows empirical regression discontinuity plots estimating the effect of winning a state legislature election on ever running in a House general election. The fitted lines are second-order polynomials. The x-axis is the fractional vote share margin. The plots include 95 percent confidence intervals calculated from a linear regression of the raw data within the optimal bandwidths on each side of the cutoff.

In Figure 4 we show that there is about a 1 percentage point increase in the unconditional probability of winning a general election based on a near-victory in a state legislative election. While this magnitude is small, it is large relative to the rate at which state legislative candidates are ever observed winning a House election (under 1 percent, see Table 1). That is, a marginally won term in the state legislature effectively doubles the likelihood that a politician will ever win a Congressional seat.

Figure 4. Regression discontinuity plots: winning general elections, win House general election, win any general.

Note: These plots show empirical regression discontinuity plots estimating the effect of winning a state legislature election on winning a House election and winning any general (house or Senate) election. The fitted lines are second-order polynomials. The x-axis is the fractional vote share margin ranging from − 1 to 1. The plots include 95 percent confidence intervals calculated from a linear regression of the raw data within the optimal bandwidths on each side of the cutoff.

3.2 Point estimates from local linear RDD

The graphical analysis in the above figures are akin to the local linear estimation of Equation 1, the results of which we show in Table 3. We estimate three indicator variables that take a value of one if the state legislator ever runs in a House primary, wins a House primary, or runs in a House general election, respectively, and is zero otherwise.¹⁶

Table 3. Effect of state legislative service on career progression to Congressional candidacy and representation

$^{\ast }{\rm p} < 0.1$; $^{\ast \ast }{\rm p} < 0.05$; $^{\ast \ast \ast }{\rm p} < 0.01$.

Note: This table reports estimates of the effect of an additional state legislative term on individuals’ career progression to Congress. The dependent variable is equal to one if the candidate ever runs in the election listed in the column header and is zero otherwise. The sample contains all first-time state legislative elections within the optimal bandwidth based on the Calonico et al. (2019) algorithm. All regressions include state and election year fixed effects, a linear term in the election margin as well as its interaction with the indicator for having won, and the full set of candidate and election controls. Estimations are triangular kernel-weighted. Standard errors clustered by state are reported in parentheses.

Column 1 of Table 3 establishes that an individual who closely wins a state legislature seat is approximately 3.6 percentage points more likely to be seen contesting for a House primary election, is 2.6 percentage points more likely to win the House primary. Additional state legislature experience increases the probability of running in a House general election by 2.3 percentage points, and increases the probability of ever winning by approximately 1 percentage point.

The mean rate of contesting a House election in the sample is only 2.2 percent, and that of winning a House election is less than 1 percent. The effects of a marginal term in the state legislature on these outcomes are thus substantial: the magnitude of the effect on any of the four outcomes is approximately the mean rate at which state legislature candidates are ever observed contesting or winning primaries or general elections.¹⁷ That is, serving an additional state legislature term results in a doubling of the rate of contesting up, as well as a doubling in the likelihood of ever being elected to Congressional office.

There is minimal effect on Senate candidacy and representation (see Appendix Table 5); results are also robust to a range of alternative bandwidths (see Appendix Table 1) as well as estimation via logistic regression (see Appendix Table 6). Thus, not only does state legislative experience create Congressional candidates that otherwise wouldn't contest a Congressional election, but generates candidates who win their elections and become Congress members. Across measures, the effect on winning an election is around half the effect on candidacy—meaning that around half of the politicians induced to run for Congressional office due to their state legislative service eventually win a Congressional seat.

3.3 Upward career progression from state legislatures: discussion

One of the contributions of the paper is to establish empirically that gaining state legislative experience leads to a much higher probability of holding higher elected office, influencing who is represented by congressional policymaking (Fox and Lawless, 2011; Carnes, 2018). This contributes to a vast literature on political selection given that, to date, an empirical estimation of the return to state legislative service on career progression has not been established—for the USA, or any other country. These results also provide context to the broader literature on candidate selection by clarifying that state legislative service increases the marginal candidate's likelihood of running for Congress—even when holding constant unobservables such as ambition or other valence traits (Maestas et al., 2006; Fox and Lawless, 2014). Below, we unpack heterogeneity in this effect across dimensions of institutional settings that may be hypothesized to produce variation in the capacity or propensity to seek higher office.

A typical concern about estimates recovered by a regression discontinuity design is their generalizability away from the threshold-based locality around which estimates are internally valid. In the current context, the concern is that competitive elections (or districts) differ systematically from non-competitive districts such that effects on career progression from close-won elections are either larger or smaller than those when the candidate wins by a large margin. On the one hand, a competitive election could drain a candidate's resources, precluding a Congressional run in the near future and thus muting the effect relative to a candidate who won by a large margin. Alternatively, a state legislator who won a competitive election may have received greater exposure to voters during their campaign, thereby increasing their valence and the viability of a future Congressional campaign. Existing research sheds some light onto this concern,¹⁸ though there is not a consensus on how competitive and non-competitive state legislature districts vary in their ability to generate candidates for federal office.¹⁹ While we cannot say that our estimates represent upper or lower bounds of the average effect across all candidates and elections, the value in the estimates we recover is in their locally valid interpretation as the effect of state legislature service for marginally elected state legislators. This subset comprises approximately 46 percent of all contested elections, and the approach yields estimates that represent an effect among candidates whose win was far from guaranteed—rather than strong incumbents or political elites who win by large margins, or generally unelectable candidates. This means that the positive effects of state legislature service to career progression are relevant to otherwise marginal politicians, whose political career ostensibly could have stalled after losing their state legislature race.

3.4 Heterogeneity and robustness of main results

In this section, we examine the heterogeneity surrounding the primary quantity of interest: near electoral victories to state legislatures and upward candidacy. We analyze possible sources of variation by legislators’ career differences, and other sources of district and legislature traits that might influence the probability of running for higher office conditional on a near win.

Career lengths. For how long is state legislative experience valuable in generating a Congressional bid? We might expect any of three profiles of the effect of experience over time: (1) immediately valuable, but with high depreciation, (2) constant with no depreciation, or (3) growing in value over time. Figure 5 plots the probability of running for higher office by or at different points of time, as measured by the number of Congressional elections that occur from the time of the close state legislature election. The top two plots of the panel display the cumulative probability of running and winning a House election over an increasingly expansive window relative to the time of the close-won state legislature election. That is, we estimate the probability that somebody will run or win by the first n opportunities to do so; our main results above are for when n = ∞. As these show, the probabilities increase until around the fifth opportunity (given that most state legislature terms of office are two years, this is within ten years after the focal state legislature election)—at which point they begin to plateau. The bottom two plots estimate the contemporaneous (rather than cumulative) quantity: the probability that an individual will run for (panel C) or win (panel D) a federal election relative to when their state legislature election occurred. These marginal probabilities at a given point in time further highlight the fact that the value of state legislative experience in generating Congressional candidates and representatives is around ten years.

Figure 5. Congressional candidacy and representation across expanding time horizons: (a) run in House election, (b) win House election, (c) run in House election and (d) win House election.

Note: These plots display the coefficients from regression discontinuity estimates of the same form as above, along with 95 percent confidence intervals. In the top two figures, each outcome is an indicator of whether an individual runs for higher office given a certain opportunity to do so. For example, the third opportunity would represent their third congressional election while in state legislative office, and the probability is cumulative across those three opportunities. In the bottom two panels, the probability of running is plotted within a given window of opportunity. Instead of a cumulative probability, it is the probability that a candidate will run in their third term, not by their third term, for instance.

Career stage. We next investigate the point in an individual's career at which additional state legislative experience is most valuable for higher level candidacy by investigating the magnitude of effects in a different sample of state legislators—those who win or lose a close election after already having served one, two or three (or more) terms in a state legislature. Figure 6 estimates the specifications in Table 3 across samples of candidates who contest a state legislature election having already served. (The primary analysis uses candidates who are observed for the first time in our sample, and thus would never have been observed having previously held state legislative office.) We see quite clearly that the effects of additional service are positive and approximately similar for those who have never previously served (similar to the sample in our main analysis) and those who win an additional term after already serving one term, but begins to consistently decline at the point of winning a third term (two previous terms served) to the point of being statistically indistinguishable from zero for those who already served three terms. We are cautious with interpretation here, because it is not possible to distinguish the pattern of effects as being attributable to heterogeneous effects versus selection on types (i.e., those who run again after having served three terms might have lower levels of progressive ambition). However, this does provide additional evidence that the value of state legislative service is highest at the earlier stages of a legislator's career.

Figure 6. Model results by number of previous legislative terms served.

Note: Figure depicts coefficient estimates and 95 percent confidence intervals for the estimation of Equation 1 for separate samples based on the number of previous terms served and the outcomes defined in the legend.

Deterrence. Next, we test whether the effect might ultimately be derived from the behavioral response among state legislature election losers by restricting the sample to candidates in close elections who previously lost a state legislature election. Among this subsample, we hypothesize, the deterrence/hazard effects of losing an election will be smaller, and thus if our main effect were primarily due to an exit from politics as a result of a previous election loss, we would see a smaller difference among winners and losers in the effect of winning the focal state legislature election on higher-level candidacy. Appendix Table 7 estimates the specification in Table 3 for this sample, and finds effects of equivalent magnitude—providing indirect evidence against this explanation of our main results.

Term limits. We estimate samples separated by state-years with term limits versus no term limits in Appendix Tables 8 and 9.²⁰ Most states that enacted term limits did so in the 1990s, and thus the term-limited legislatures comprise a more recent sample which is also subject to greater truncation, though effects for both samples remain positive. As discussed previously, we expect a positive relationship between state legislative experience and running for Congress among both term limited and non-term limited states: conditional on running for office in either institutional equilibrium, there will still be a positive effect of serving in a state legislature on upwardcandidacy among similar types of candidates (a comparison produced by the RD design). Appendix Figure 3 displays the main results from above, disaggregated by decade. Our substantive interpretations remain the same—a positive relationship in both institutional arrangements—and the results are almost all statistically significant, despite the drop in sample size as a result of dividing the data.

Chamber heterogeneity. We then test heterogeneity by which chamber the legislator is initially elected to, using the same specifications as above. On average, the upper chamber of a legislature will offer more opportunities in terms of resources and policy impact for a state legislator, perhaps increasing the probability that they run for higher office. However, as is the case in the professionalism discussion previously, this might also lead to higher retention as it increases the quality of the job, leading to countervailing hypotheses as to the effect of state legislative service on upward candidacy. In Appendix Tables 10 and 11 we split the sample by the chamber in which the legislator is initially elected and find that the effect of narrowly winning an election remains positive and statistically significant in both samples. However, the effect is larger among those elected initially to the upper chamber.²¹

District competitiveness. Finally, one potential concern with the main results above is that they are driven primarily by state legislative districts which overlap competitive congressional districts, offering more opportunities for legislators to run for higher office. To provide some additional evidence about whether this might or might not be the case, in Appendix Figure 4 we present maps of state legislative districts for the lower house and congressional districts, filled in based on competitiveness. These maps show that there is substantial overlap in which districts are competitive at the state legislative level versus the congressional level, and no systematic difference in where one is more likely to find a certain type of district.²²

4. Institutional determinants of upward political selection

We now discuss and interrogate some mechanisms of upward candidacy related to a near-victory in a state legislative election, and examine heterogeneity in the relationship based on features of the legislature to which the candidate is elected. As discussed theoretically above, we specifically focus on legislative professionalism—the resources available to legislators—which varies significantly across states. We then analyze if ideological extremism predicts running for higher office. Finally, we assess how variation in outside options available to state legislators, specifically within the lobbying industry produced by revolving door restrictiveness, produces differences in which state legislatures experience higher turnover and more produce more candidates for higher office.

4.1 Professionalism

Next, we aim to better understand how the relationship between state legislative experience and Congressional candidacy varies across institutional contexts—namely, across legislatures that vary in their level of professionalism. Our primary measures of state legislative professionalism come from Bowen and Greene (2014).²³ Our primary measure of interest is the first dimension of a state's overall legislative professionalism (see Bowen and Greene, 2014, for more details on construction). In later analyses, we also use the constituent parts of the index, including the length of the legislative session, legislators’ salary, and expenditures per legislator.

There is broad heterogeneity across states in their professionalism providing substantial heterogeneity in this measure. For instance, in California legislators are full-time and are paid over $ 110, 000 a year in salary. They are allocated personal staff and maintain offices in the capitol as well as their districts. However, California is largely an outlier. In 25 state legislatures, legislators are provided one or fewer personal staff. In North Carolina, legislators are paid less than $ 14, 000 a year in salary. In many legislatures, then, legislators are expected to hold full-time jobs outside of their elected service.

Next, we turn to the relationship between these features of a state legislative seat and the propensity for state legislators to compete for higher office. We motivate this undertaking by establishing the basic correlation between states’ mean level of professionalism (described in more detail below), and the share of state legislators serving in our data that we ever observe running in a Congressional primary or general election. Figure 7 depicts this relationship graphically, suggesting a strong positive correlation between professionalism and career progression.

Figure 7. Correlation of state legislative professionalism and upward candidacy.

The x-axis is the Bowen and Greene (2014) professionalism index and the y-axis is the proportion of state legislative candidates that ever run for a congressional office.

An interpretation of this descriptive correlation at face value would suggest that professionalism is strongly correlated with career progression from local to national politics, or, alternatively, that professionalized legislatures attract more upwardly mobile politicians. To first assess the heterogeneity in the probability of running for higher office on legislative professionalism, we interact the focal RDD regressor in Equation (1) (Won state seat _ist) with a unit-standardized professionalism score from Bowen and Greene (2014), and include the main effect of the professionalism score. The specification then becomes:

(2)$$\eqalign{Y_{ist}& = \alpha_0 Won\; state \; seat_{ist} + \alpha_1 [ Won\; state \; seat_{ist}\ast Professionalism_{ist}] \cr & \quad + \alpha_2 Professionalism_{ist} + \beta margin_{ist} \cr & \quad + \gamma [ Won\; state \; seat_{ist}\times margin_{ist}] + X_{ist} \delta + \tau_{s} + \phi_{t} + \epsilon_{ist}}$$

Table 4 shows, consistent with Figure 7, that there is a positive relationship between professionalism and running for higher office—although the quasi-experimentally estimated effects are approximately half the magnitude of the correlational magnitude. Taking the results from columns 3 and 4, a winning candidate from a legislature one standard deviation over the mean value of professionalism is 25 percent more likely to run for a House seat and 50 percent more likely to win a House seat.

Table 4. Effect of state legislative service interacted with state legislative professionalism

$^{\ast }{\rm p}< 0.1$; $^{\ast \ast }{\rm p}< 0.05$; $^{\ast \ast \ast }{\rm p}< 0.01$.

Note: This table reports the same model specifications as in Table 3, but includes an interaction with a globally unit-standardized professionalism score for the state-chamber-year based on Bowen and Greene (2014). The dependent variable is equal to one if the candidate ever runs in the election listed in the column header and is zero otherwise. The sample contains all first-time state legislative elections within the optimal bandwidth based on the Calonico et al. (2019) algorithm. All regressions include state and election year fixed effects, a linear term in the election margin as well as its interaction with the indicator for having won, and the full set of candidate and election controls. Estimations are triangular kernel-weighted. Standard errors clustered by state are reported in parentheses.

We also examine heterogeneity in effects across the constituent parts of the professionalism score, including session length, legislator salary, and expenditures per legislator. Appendix Table 12 displays results from the regression discontinuity specification interacting each of these components with having narrowly won an election. Each constituent part is associated with a positive increase in Congressional candidacy, but high collinearity reduces precision substantially. We also estimate separate models for each interaction (see Appendix Table 13), which shows each component being statistically significant and with an overall effect similar in magnitude to the interaction with the professionalism index from Table 16.²⁴

Finally, we consider additional heterogeneity driven by the interaction between term limits and professionalism. One concern is that those states that adopted term limits are fundamentally different from those that did not, and that pooling these states together could hide that the effect is only among states who never adopted term limits. Kousser (2005) in particular outlines the “dismantling” of legislative professionalism produced by term limits and how term limits can change the composition of legislatures (as well as inter- and intra-branch interactions), which may have downstream impacts on the propensity to run for Congress and win Congressional seats. Descriptively, there is a slight correlation between term limit adoption and professionalism: 10 percent of “high” professionalism state-years are term-limited compared to 7 percent of “medium” professionalism state-years and 4 percent of “low” professionalism state-years. But is there a unique feature of the states that do adopt term limits that masks the relationship between winning a state legislature seat and seeking higher office? And how does this relate to professionalism in the same way as described above? Appendix Table 8 contains estimates based on elections in the sample of states that ever adopted term limits, interacting the indicator for winning with the professionalism score. Among these states, the main effect of winning a state legislative election is substantively large and statistically significant at conventional levels. Additionally, the interaction with professionalism is substantively large in the outcomes of running for or winning a House election. In other words, there does not appear to be evidence that states that adopt term limits see fundamentally different behavior among (barely winning) state legislators in terms of seeking higher office.

We show that additional experience in a highly professional state legislature results in a higher probability of running for a congressional seat in the future. This is consistent with previous research (Maestas et al., 2006) which shows observationally among a sample of sitting state legislators that more professional legislatures see higher rates of candidate emergence for Congress—with professional legislatures potentially enabling ambition among already ambitious types. Our results confirm this finding quasi-experimentally in a framework that avoids a confounding of electoral success in the state legislature with politicians’ ambition. Our results thus lend further support to the idea that service in the legislature itself, and the institutional features associated with this service, lead to higher rates of running for Congress. In the appendix (Tables 12 and 13), we show that it is difficult to provide even suggestive evidence that certain features of professionalism are more likely to contribute to this overall result.²⁵ Next, we examine another feature affecting legislator career concerns: the returns available vis-a-vis outside options.

4.2 Outside options

To investigate this relationship, we employ data from Strickland (2019) on the severity of lobbying “cooling-off” periods in state legislatures. Cooling-off periods determine how long (if at all) legislators must wait to lobby their former institution. (In the Congressional setting, cooling-off regulations have been shown to change the career trajectories of congressional staff who become lobbyists (Cain and Drutman, 2014).) Over the course of our sample, most states implemented a one year ban for legislators, with a two year ban less common, followed by a session or six month ban, the least common. Currently, nearly 40 states have some version of a cooling off period.²⁶ We take these time-varying measures by state and bin them into “none,” “low,” “medium,” and “high” bins of strictness, following Strickland (2019). We then assign each of these bins a number from 0 to 3, respectively. “Low” states consist of those with cooling off periods of only one term whereas “high” states have a ban of two years or more. The expectation is that in states with less strict cooling off periods, running for higher office becomes less attractive given the availability of good outside options. These states should see fewer of their legislators run for higher office.

We use the same model specifications as above and separate the states by the low and high categories. Table 3 displays the coefficient on the RDD estimate for these models. Consistent with the expectation from previous work, state legislators in states with stricter cooling off periods tend to run for higher office at higher rates. The difference is quite large between states with no or lax restrictions compared to those with high restrictions. Since the restrictions variable is by construction linear and additive, a legislator who narrowly wins an election to a state with high restrictions is four percentage points more likely to run in a House primary than a similar legislator from a state with no restrictions.

In the appendix, we further investigate the relationship between professionalism and revolving door restrictions, which has not been taken on by previous research. It could be the case that less professional legislatures are also those with more lax revolving door restrictions and willingness to enact revolving door restrictions is correlated with professionalism. However, these states may simultaneously be the most likely to have strict revolving door restrictions, because revolving door-type opportunities are more attractive relative to the low salaries paid to their legislators. Similarly, professional legislatures may be those that also enact more revolving door restrictions—but they also may be less meaningful because legislator salaries are higher. We shed some light onto these questions by first showing that descriptively professionalism and revolving door restriction strictness are correlated.²⁷ Appendix Table 17 shows results from the same interaction as Table 5 with the sample split by high or low professionalism. In both subsamples, the interaction is positive and similar in magnitude to the pooled results, though lacking in statistical precision, in part due to smaller sample sizes.

Table 5. Effect of state legislative service interacted with state revolving door legislation

$^{\ast }{\rm p} < 0.1$; $^{\ast \ast }{\rm p} < 0.05$; $^{\ast \ast \ast }{\rm p} < 0.01$.

Note: This table reports the same model specifications as in Table 3, but includes an interaction with an indicator of revolving door strictness from Strickland (2019). Data are available from 1990 to 2008 and the median year is 2002. The dependent variable is equal to one if the candidate ever runs in the election listed in the column header and is zero otherwise. The sample contains all first-time state legislative elections within the optimal bandwidth based on the Calonico et al. (2019) algorithm. All regressions include state and election year fixed effects, a linear term in the election margin as well as its interaction with the indicator for having won, and the full set of candidate and election controls. Estimations are triangular kernel-weighted. Standard errors clustered by state are reported in parentheses.

Policymakers designing revolving door restrictions may jointly aim to reduce regulatory capture and increase retention among legislators through limiting outside career options and lowering the opportunity cost of remaining a legislator. Our results, however, lend nuance to these considerations: “cooling off” periods are associated with more state legislators seeking higher office. While revolving door regulations may address one aspect of good governance by reducing capture and limiting lobbyists’ influence, our results suggest this comes at the cost of a tradeoff in legislator retention.

5. Discussion and conclusion

A hallmark of democratic governance is the belief in the ability of competent citizens to earn their way to prominent leadership roles. Examples of successful individuals with “humble beginnings” are often used as proof against a preponderance of elite capture or dynasticism in US politics. Service in a state legislature is the single-most dominant career path that precedes Congressional service—but are those who follow this path destined to higher offices, or does serving in the state legislature facilitate career progression for the marginal state legislator? Candidate emergence and political selection, and what drives individuals to run for higher office, holds substantial implications for descriptive representation (Lawless, 2004; Washington, 2008; Carnes, 2013; Thomsen and King, 2020), accountability of elected officials to citizens (Schlesinger, 1966; Rohde, 1979), and the prevention of corruption (Ferraz and Finan, 2008; Dal Bó and Finan, 2018).

Using data that trace the political careers of candidates for US state legislatures since 1967, we quantify the value of serving in a state legislature for career progression to national politics. Winning an additional term in the state legislature increases the probability of ever running in a Congressional primary by four percentage points, in a Congressional general election by two percentage points, and increases the unconditional likelihood of ever winning a Congressional general election and serving in Congress by approximately 1 percentage point. These effect sizes are large, as they constitute a more than doubling of the likelihood of these outcomes relative to the mean rate at which state legislature candidates are seen contesting or serving for Congressional office. These results add important context to competing theoretical debates on political selection broadly, and candidate emergence for Congress in particular. Scholarship on candidate emergence for Congressional seats identifies an ambition gap which drives variation in which candidates run, or are selected to run, for Congress (Maestas et al., 2006; Fox and Lawless, 2011; Dittmar, 2015). If we assume that office-seeking ambition is similar among candidates who barely win or lose an initial state legislative election, we find a meaningful career return to state legislative service.

The average effect, however, masks substantial heterogeneities across states whose institutions differentially favor or hinder political career progression. We find that highly professionalized legislatures are more likely to see their members run for and win higher office, as those legislatures that provide more resources to legislators lead to more Congressional candidates and enable more successful Congressional campaigns (e.g., Berkman, 1994; Berry et al., 2000). Better pay and more resources may thus have countervailing effects on the retention of state legislators, as increased resources may enable some legislators to instead seek federal office. Additionally, we find that each element of professionalism—salaries, staff, and term length—are positively correlated with upward candidacy. Furthermore, the restrictiveness of revolving door regulations for state legislators also produces variability in higher office seeking: states with stricter regulations—in other words, those states that provide more limits on the ex post careers of their legislators—are more likely to see their members run for Congress. For policymakers, this is a complex problem. Does restricting the later careers of state legislators affect selection into office and turnover once there?

Our work contributes to the understanding of how the returns to office affect political selection (Eggers and Hainmueller, 2009; Querubin and Jr, 2009) and specifically the role of professionalism in higher office seeking (Maestas et al., 2006; Butler and Nemerever, 2019). Broader implications derive from a better understanding of the determinants of the pool of congressional candidates. As Congress members who were formerly state legislators in professionalized legislatures are generally more effective relative to their peers without state legislative experience (Volden and Wiseman, 2014), then the fact that some states are more likely to send their legislators to Congress generates substantive consequences for the quality of democratic representation. Future work should investigate the implications these dynamics have for representation via policy outcomes, as well as quantify the effects of local political service on career progression to the various public service opportunities available to state legislators for subsequent career moves.