Expertise acquisition in Congress

According to many, the US Congress desperately needs reform because its capacity to govern has declined. Congressional capacity cannot be understood without examining how the expertise available to members is fostered or discouraged. We present a theory of expertise acquisition and apply it to the problem of overseeing the Executive. We use this theory to organize a dataset of congressional staff employment merged with new records of invi-tations, applications, and attendance at training sessions produced by three nonproﬁt organizations in Washington, DC. We ﬁnd that staffers are more likely to acquire expertise when their jobs are more secure and there are more opportunities to use their expertise in careers outside of Congress— most notably, when their party takes control of the presidency. Our analysis suggests that oversight expertise is generally not sufﬁciently valuable outside of Congress to entice many staffers to acquire it without subsidies. Dissatisfaction with Congress is a national pastime.

nation as a whole (e.g., Crosson et al., 2021;Furnas et al., 2020;McCrain, 2018;Ritchie & You, 2021).Calls for internal reform have followed, the most visible of which come from the Select Committee on the Modernization of Congress.
But spending taxpayer dollars on Congress' internal functions is unpopular, and may even conflict with the political interests of party leaders.This means most efforts to increase congressional capacity are designed and executed by nonprofit organizations.The same organizations who advocate rule changes and legislation to build Congress' capacity also attempt to fill perceived gaps by providing seminars, training, and other services.Almost nothing is known about how these efforts work, or what they can teach us about expertise acquisition in Congress.
We provide a theoretical framework for thinking about these questions and leverage new data on staff training to evaluate it.Expertise development among unelected personnel has been central to understanding presidential and executive politics (e.g., Callander, 2008;Gailmard & Patty, 2013) but is largely absent from research on the analogous legislative context.Following models in labor economics, we view expertise as a form of human capital that makes staffers more productive and therefore more valuable to members of Congress. 1 Under what conditions will staffers be willing to put in the hard work necessary to become experts?That depends on the kind of human capital they are acquiring.Is the expertise firm specific human capital that makes congressional staffers better at their current job, but does less to make them better at jobs outside of Congress, or is it general human capital that makes congressional staffers more productive in many jobs?The answer determines whether staffers will happily acquire expertise on their own or require some subsidy or prodding from their employer. 2 Thus, to understand congressional capacity and design effective interventions to improve it, we must identify which kinds of expertise are firm specific and which are general.If the expertise is firm specific, then the shortage stems from members' unwillingness to pay their staffers to acquire it, and reformers should focus on making that expertise cheaper and more rewarding for members to train their staff.If it is general, then the shortage stems either from staffers' inability to pay for training or Congress' inability to compensate experts enough to convince them to stay, and reformers should focus on making it cheaper for staffers to get training and helping legislators retain their experts.
We take up this question by studying staffers' decisions to acquire one particular kind of expertise: knowledge about how to conduct oversight of the Executive Branch.Since staff perform numerous functions, this is not the only form of expertise.But it is a crucially important one that attracts substantial attention, and our analysis of it provides a blueprint that can be used to study whether other kinds of expertise.Most notably, as legislative productivity has declined, congressional oversight becomes a more substantial proportion of Congress' workload.In addition, as we describe later, examining this kind of expertise allows us to evaluate additional implications of our theory.
We leverage new data, combining records of staff employment with applications and attendance at training bootcamps and seminars from 2011 and 2021.These bipartisan, voluntary events are meant to build capacity by teaching staff practical skills in overseeing the executive branch.They are funded and organized by the Project on Government Oversight (POGO), the Levin Center, and the Lugar Center and taught by current and former congressional staffers with decades of experience.There is no formal training required of congressional staff.Research on congressional capac- 1 We are not the first to adopt a human-capital framework to study this topic.See Parker (2008) and LaPira and Thomas (2017). 2 Throughout, by "firm specific," we mean specific to Congress as a whole.We expect that most forms of human capital are transferable across congressional offices.Whether the internal market for human capital is sufficiently robust to induce staffers to acquire expertise on their own is an empirical question.
ity typically studies staff tenure, turnover, and pay.To our knowledge, this is the first attempt to observe and analyze skill acquisition within career.
As a preliminary, we establish that the departure of a member from Congress induces significant career uncertainty for their staff.In offices where the member loses or retires, only 30% still work in Congress by the end of the next year-compared to 68% for staffers of winning members.We exploit this variation in career prospects to test whether oversight is firm specific or general.If it is firm specific, the potential departure of legislators ought to make their staffs less likely to attend training because it has a good chance of becoming irrelevant to their work.If it is general, the prospect of members' departure ought to make their staffs more likely to attend training because it will help them get better jobs after their bosses leave.
Our data provides consistent evidence that suggests oversight expertise is firm specific human capital.Staffers are far less likely to attend training or seminars during their member's final term in office.We present a more sophisticated design later, but a rough calculation highlights the magnitude of the difference: offices where the member will stay in office into the next Congress send 36% more staffers to oversight training than offices where the member is in his or her final term.Training sessions scheduled for times when it is inexpensive for members to release staffers for training-when the chamber is on break-attract more attendees, at least from the perpetually busy House.This suggests offices bear at least some of the cost of training, which only happens for firm specific forms of human capital.Staffers who receive training stay in the institution substantially longer than those who do not, which is consistent with the notion that the training increases their value to Congress more than it does to prospective employers outside of Congress.Finally, we find that in circumstances where oversight expertise is more general, staffers are less sensitive to career uncertainty: that is, when a staffer's party controls the presidency-and thus, when attractive executive branch positions that value oversight expertise are potentially available-the effect of career uncertainty on expertise acquisition is diminished.
Our findings have important implications for the study of congressional capacity.If oversight expertise were general, nonprofit organizations could improve congressional capacity by appealing directly to staffers.Our analysis, however, suggests that this is not the case.Many staffers will hesitate to put in the hard work to become experts in oversight because they will not be the ones who benefit.The members of Congress for whom they work will, so legislators must absorb the costs their staffers incur to become experts.This implies that those who want to enhance congressional capacity for conducting oversight must either focus on increasing the value of oversight expertise to legislators or find ways to make it more general so that staffers become more likely to pursue it on their own.

CONGRESSIONAL CAPACITY AND EXPERTISE
LaPira et al. define "congressional capacity" as the "organizational resources, knowledge, expertise, time, space, and technology that are necessary for Congress to perform its Constitutional role" (2020, 1).The definition makes clear that congressional capacity is complicated, multidimensional, and difficult to measure.Scholars have worked to identify capacity and the factors that affect it.To that end, a long line of research has noted that Congressional offices are complicated enterprises and have the difficult task of making policy on a huge range of subjects, building on foundational work by Salisbury and Shepsle (1981).Additionally, the capacity of a congressional office is subject to many different competing interests, influences, and constraints (Hall & Deardorff, 2006;LaPira et al., 2020;Lee, 2016).For instance, offices must balance policy and reelection concerns and supplement their constrained capacity with support from interest groups or lobbyists, which has potentially troubling implications.
In practice, much of the work on congressional capacity has focused on the role and institutional support of congressional staffers.Staffers are a critical aspect of congressional capacity, having important roles creating and passing policy, providing information, and influencing the behavior of legislators (DeGregorio, 1994;Montgomery & Nyhan, 2017).Additionally, the network of congressional staffers, their experience, and the resources they are allocated affect the productivity of the institution (Burgat, 2020;McCrain, 2018).
Recent work focuses on the decline of congressional capacity by investigating trends related to funding and personnel.Since the 1990s, funding levels for personal offices have stagnated and member salaries have decreased, despite increasing workloads and responsibilities (Reynolds, 2020).Salaries and the amount of allocated resources for substantive legislative and committee staff have declined, with legislators instead prioritizing communications and constituency service staff (Crosson et al., 2021;Jensen, 2011).Nonpartisan agencies which offer expert information to legislators like the Congressional Budget Office, the Government Accountability Office, and the Congressional Research Service have all been subjected to budget reductions and declines in numbers of personnel to varying degrees (Fagan & McGee, 2022;Reynolds, 2020).
Despite the large empirical focus on experience and pay, most research sees congressional capacity as a function of the expertise of members, staff, and sup-port organizations.Lewallen et al. (2016), for example, argue that staff expertise (especially as manifested in committees) is critical to explaining the decline in problem solving.LaPira et al. see "the level of specialized knowledge and the ability of Congress to tackle complex problems" (2020,19) as the critical resource that contributes to Congress' ability to be representative, responsive, deliberative, and to serve as a watchdog on the executive branch.Not surprisingly, there is also evidence that this kind of expertise matters for policymaking outputs.Crosson et al. (2020) find staff tenure may be one underlying explanation.Most pertinent for our purposes, their findings highlight the importance of the quality, not the quantity, of staff-again, suggesting expertise as an underlying mechanism for capacity.
While scholars agree on the importance of expertise in Congress, studies about the development of expertise have mostly concentrated on executives and the bureaucracy (e.g., Gailmard & Patty, 2013).It adopts a principal-agent framework applicable to congressional staff.But congressional staff serve at the pleasure of their member, and even staff in good standing might lose their job after the next election cycle.The disparity in both compensation and lifestyle between working in Congress and outside options is stark-likely even greater than the disparity between executive offices and private employers.In short, the appropriate question might be why congressional staffers develop expertise at all.This is a starting point, then, both for those interested in reform and further investigation.

WHAT MOTIVATES EXPERTISE ACQUISITION
We follow labor economists in defining expertise as a form of human capital-an attribute that makes a worker more valuable to their employer.Expertise includes the ability to craft more appealing policy proposals, which is how formal theorists have traditionally modeled expertise (e.g., Gilligan & Krehbiel, 1987;Hitt et al., 2017), but it also encompasses a broader set of attributes that could help the staffer's member of Congress, such as the ability to extract valuable information from the executive branch, skill in attracting favorable attention from the media, and an extensive professional network that can be leveraged to advance the member's interests.
However, staffers must incur costs to acquire expertise.The more time they spend researching policy, acquiring skills, and making professional connections, the less time they have left to complete their shortterm work, spend time with family and friends, read Proust, perform household labor, and earn outside income.
Given the costs, there are two reasons why a staffer might nevertheless acquire expertise.First, the expertise might allow them to secure better compensation.Money is one kind of compensation that staffers might value, but they could also be compensated with greater influence over policy, greater access to their member, more flexibility in their work hours, or more interesting responsibilities.Since we are modeling Congress as a single employer, if expertise helps staffers get jobs in more attractive offices, such as jobs on the staffs of more senior legislators, committees, or party leadership that would also qualify as better compensation.If the long-run value of that extra compensation exceeds the short-term cost of the training, the staffer will attend the training and bear the cost on their own.Second, their employer might cover the cost of the training directly, perhaps by scheduling it during work hours in lieu of the staffer's regular responsibilities or perhaps by giving the staffer a one-time, up-front bonus for the training.This shifts the cost of the training from the staffer to their employer.

General versus firm specific human capital
Under what conditions does the staffer bear the cost of acquiring expertise, and under what conditions does their employer cover the cost?Becker (1962) shows that it depends on how the training affects her ability to find a better job at a different firm.If a staffer could easily find an attractive job that uses the training at another firm, then her current employer must increase her compensation to prevent her from leaving.Since the training will increase her compensation over the long run (either in the form of a raise from her current employer or a better job with a different employer), the staffer will incur the cost of the training on her own without any further inducements from her employer.Becker calls this general human capital, because it generalizes to many prospective employers.
If there are so few firms that value workers with the training that is time-consuming, difficult, or simply unlikely for the staffer to find an attractive job that would use it, then her employer does not need to compensate her much to prevent her from leaving.In fact, it can expropriate the productivity gains associated with her expertise.This deters the staffer from acquiring expertise in the first place, so the employer must help cover the costs of the training.Becker calls this firm specific human capital, because, although it makes the worker more productive at her current firm, it is hard to find another job that makes use of it.
Which kind of expertise are general, and which kinds are specific to Congress?The answer to this question has important practical implications for reformers.
For general expertise, interventions that appeal to the members will not work well.Even if the members appreciate the benefits of having experts on staff, they will correctly anticipate that they would need to compensate those experts more generously to keep them from leaving.The cost of that extra compensation would counterbalance the gains from expertise.On the other hand, the more firm specific expertise is, the less effective interventions that target staffers will be.They will correctly anticipate that their employers would expropriate much of the productivity gains from additional expertise, which makes them less inclined to incur costs for the training.
Of course, general and firm specific are in reality two poles on a continuum, and most kinds of human capital fall somewhere between the two.Acemoglu and Pischke (1999), for example, model human capital on a continuum from equally useful to all employers to useful for only the worker's current employer, and they find that the key predictions from the dichotomous theory carry over to the continuous case.As the human capital becomes more useful for finding a good job with another employer, it gradually behaves less like perfectly firm specific human capital and more like perfectly general human capital.To clarify this point, Section A in the online supporting information (p.A2) provides a simple formal theory based on Acemoglu and Pischke (1999) that allows human capital to vary continuously between perfectly firm specific and perfectly general.It shows that the empirical predictions from the dichotomous case are robust to this extension.
Accordingly, for our main analysis, we use the dichotomous distinction between general and firm specific as a shorthand.When we ask whether expertise is general or firm specific, we are really asking whether it is far enough on the general side of the spectrum for workers and employers to treat it as if it were perfectly general human capital, or whether it is firm specific enough for them to treat it as if it were perfectly firm specific.
Sometimes, the answer is obvious ex ante.Typing, management, interpersonal skills, business writing, and media production are all valuable to many employers for many jobs and therefore fall on the general side of the spectrum.Knowledge of congressional ethics rules and a healthy relationship with the Office of Legislative Counsel have narrower value outside of Congress and are likely on the firm specific end.Many of the kinds of human capital that are relevant to congressional capacity, such as expertise in policy, oversight, and legislative procedure, fall somewhere in between, and it is not obvious ex ante whether they are closer to general or firm specific.They are not useful for many jobs, but there are some attractive jobs in lobbying firms, think tanks, and elsewhere for which they are relevant. 3If these jobs are so scarce relative to the number of staffers who want them that Congress can hold on to most of its experts without offering them much additional compensation, then these forms of expertise are effectively firm specific human capital.If these jobs are sufficiently plentiful (or if they are so extremely attractive that staffers are willing to invest in training for the slim chance of securing one), then these forms of expertise are effectively general human capital.Whether these forms of expertise are more general or firm specific is something to discover from data.

Empirical implications
Theoretical research from labor economics offers a series of tests that researchers can use to characterize whether a particular kind of human capital is closer to general or firm specific.These tests do not rely on measuring which firms want to hire Congress' experts. 4nstead, the theory of human capital encourages us to infer whether a particular kind of expertise is more firm specific or general based on the behavior of the staffers: which staffers attend training, how the contextual factors influences their decisions to attend, and how long staffers stay employed in Congress.
First, we exploit the fact that the effect of career uncertainty on expertise acquisition depends on whether expertise is general or firm specific human capital.The more likely a staffer is to lose their job in the near future, the more likely they are to acquire general human capital, because general human capital makes them more likely to get an attractive job if they must enter the labor market.The more likely a staffer is to lose their job in the near future, the less likely they are to acquire firm specific human capital.From the staffer's perspective, there is a good chance they will have to leave congressional employment whether they want to or not, and their firm specific human capital will not help them much if they have to find a job outside of Congress.This makes them less inclined to incur costs to become experts than they would otherwise be.From their employer's perspective, the firm will have little time to reap productivity gains from the firm specific human capital, which makes it less willing to pay the worker to acquire that human capital in the first place.Therefore, if expertise is general, then career uncertainty ought to be positively associ-ated with the acquisition of expertise, and if it is firm specific, then career uncertainty ought to be negatively associated with the acquisition of expertise.
However, the labor market changes over time, and expertise may move from firm specific to general as job opportunities come and go.As more job opportunities that leverage expertise arise outside of Congress, it becomes more general, and staffers' decisions about whether to acquire expertise will become less sensitive to career uncertainty.
Furthermore, if expertise is firm specific, then staffers will not acquire it unless their employer defrays some of the cost.The employer will be more inclined to bear that cost when they reap greater rewards from having experts on their staffs and the cost the employer incurs to release the worker for training is low.However, if expertise is general, then staffers will bear the entire cost of the training and pursue it on their own time.They may be more willing to bear that cost when the value of the expertise to their current employer is high, but the cost the employer would incur to release the worker for training will be irrelevant.
Finally, staffers who acquire firm specific human capital will tend to stay in Congress longer than staffers who do not.Since they are more productive, their members will be less likely to lay them off.If, for some reason exogenous to their training, they do get an attractive outside offer, their employers will also be willing to pay more to retain them.

Oversight expertise
Members of Congress and their staff perform numerous tasks that require different kinds of expertise.Our theory and research design generalize to any form of expertise, but, for our empirical tests, we focus on one particular kind of expertise: knowledge of how to conduct oversight of the executive branch.
First and foremost, whether this kind of expertise is a form of general or firm specific human capital is up for debate.Oversight expertise is potentially useful to Congress itself, interest groups that play an auxiliary role in oversight as government watchdogs, investigative journalism, and the executive branch.Perhaps the noncongressional demanders of oversight expertise are numerous enough to form a strong outside option for well-trained congressional staffers, or perhaps they are so few as to be negligible.Even if they are too few to matter, expertise in oversight might also provide a foundation that readily generalizes to private-sector jobs in law and auditing.
Second, there is an extensive literature on congressional oversight that demonstrates partisan patterns in oversight, which provide measurement leverage we later use to examine our theory.With few exceptions, studies repeatedly demonstrate that divided government leads to more frequent and more vigorous public oversight (e.g., Kriner and Schickler 2014;Lowande and Peck 2017).We discuss the potential implications for whether this expertise is general and firm specific in a subsequent section.

DATA AND APPROACH
We acquired data from nonprofit organizations that provide remarkably fine-grained measurement of offices' investments in oversight expertise.Our outcome measures are application and attendance records of two types of training events which took place in 2011-21.The first events are monthly seminars that typically last 1-1.5 hours.Each features a different lecturer who typically presents and then answers questions.They are similar in format and time commitment to research seminar presentations in higher education, but the material is more practically oriented.Example seminar topics include "Working with Whistleblowers" (June 2017), "How to Hold an Oversight Hearing" (March 2018), and "How to Write a Request Letter" (April 2021).Overall attendance at each seminar is fairly wide ranging, from a few dozen to over 100.Attendance is not capped.
The second events are biannual bootcamps that typically amount to 12 hours over a two-day period.These bootcamps are accurately described by organizers as "an intensive two-day, bipartisan training."The curriculum for these events consists of hundreds of pages of information on conducting investigations, planning hearings, constructing witness lists, interviewing witnesses, along with writing questions, press releases, and committee reports.Their pedagogical approach involves a mixture of lecturing, open discussion, group projects, and role playing.Attendance at bootcamps is typically capped to maintain the benefits of a smaller cohort.This means attendance ranged between 18 and 32, while applications might be over 100.
For this reason, we expect bootcamp applications to be the best indicator of interest in training, with attendance at seminars second.In addition, attendance at the events themselves is likely to be complicated by considerations less relevant to expertise acquisition.In the records we obtained, there were at least two documented cases of staffers applying but being prevented from attending by their Chief of Staff.The organizers work hard to maintain a reputation for bipartisanship.Of POGO, the Levin Center, and the Lugar Center, the latter two are legacy projects from members of both parties.However, it is possible that bipartisanship may be less appealing for some staff leadership.Note also that all events are free and voluntary.By chamber rules, staff cannot be given any form of compensation for the training.Even the food and drink provided at these events must be sufficiently sparse to avoid being deemed a gift "meal."For this reason, we do not expect participation to be driven by considerations that are unrelated to the desire to learn about oversight.
Another important question is whether the events that take place during these sessions suggest whether oversight expertise is general or firm specific.There are indications of both.For instance, each might involve networking benefits.Attendees at the bootcamps, for example, are placed in small, bipartisan groups of other staff they have never met and who may even work in a different chamber.At a bootcamp one author attended, these small groups exchanged business cards.In addition, the instructors occasionally include people who have been the target of oversight, and many participants have law degrees.This might make the training valuable for employers who specialize in defending targets of congressional inquiry.This suggests the benefits of the training are somewhat general, in that they may increase connections to others and make staffers more valuable to outside employers.
On the other hand, the course materials and content also point to institutional (i.e., firm specific) knowledge.Course evaluations almost exclusively mention the practical advice regarding oversight and almost 15405907, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ajps.12848 by University Of Michigan Library, Wiley Online Library on [26/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License never mention networking opportunities.For one bootcamp, we surveyed the participants before and after, along with another sample of untrained staff who had expressed interest in the training.We found some evidence that the training increased their knowledge of basic oversight procedures.We describe this evidence in more detail in Section D of the online supporting information (p.A11).The effectiveness of the training is a separate question outside the scope of the present study; however, we take this as evidence that the training events are meant to develop expertise valuable within Congress.
We supplement this data with employment records for the population of Congressional staffers over the same period.Both the Senate and House of Representatives are required to regularly publish reports with detailed, time-stamped, itemized lists of expenditures by 2 U.S.C. 104(a).Critically, these reports include all payments to employees made by Congressional offices.
We built a dataset containing the names of staffers, the offices for which they worked, and the dates for which they worked there, and their pay, formatted cleanly and complete with universal office, legislator, and staffer identifiers from 2011 to 2022.Constructing this data set involved considerable challenges.Names of staffers and offices are not standardized or consistent, so there are no ways to universally identify employees or employers over time.Additionally, the frequency and format of the data varied across chambers. 5We developed a series of algorithms and procedures to solve these problems.Ultimately, we were able to develop a mostly automated pipeline for parsing, standardizing, and aggregating the data.We detail our approach in Section B of the online supporting information (p.A5).Similar data sets are available through services like LegiStorm.Our approach offers a few distinct advantages.First, our data is largely constructed using automated scripts and procedures.This means that our cleaning and aggregation procedures are transparent and reproducible.Our method gives researchers the ability to modify various aspects of the cleaning and aggregation procedures to suit their needs.Second and most importantly, our data is free and publicly available.

FINDINGS
To investigate alternative implications of expertise type, we examine the effect of member departure on staff turnover, the effect of career uncertainty on expertise acquisition conditional on party control of the presidency, and finally, the effect of expertise seeking on career longevity.

Member turnover and staffer career uncertainty
We first develop and justify a way of measuring variation in job security among congressional staffers.We focus on one important and readily observable determinant of the staffer's job security: whether the member they work for stays in office.When a staffer's member leaves office, the staffer must find a new job.They might be able to find a job elsewhere in Congress, but there is no guarantee.
In fact, Figure 1 shows that staffers working for legislators who lost reelection or retire leave Congress at higher rates than staffers working for legislators who stay in office.Legislators who won, lost, or retired experienced similar levels of staff retention during the election year: the median office in each group retained 84.2%, 81.8%, and 80.0% of their staff, respectively.But staffers were often unable to find new employment in Congress after their members left office.For the median office of a successfully reelected legislator, 68.0% of election-year staffers were still working in Congress by the end of the year after the election.In contrast, just 30.0% of election-year staffers who worked for defeated legislators and 39.1% of election-year staffers for legislators who retired were still working in Congress by the end of the year after the election.
To supplement this descriptive analysis, we adopt a regression discontinuity design (RDD) based on the results of close elections to show that member turnover actually causes staff turnover (e.g., Gerber & Hopkins, 2011). 6These RDD specifications represent a conservative estimate of such effects.Staffers working for members who retire or are unlikely to win reelection will generally know this information well in advance of the general election.This enables staffers to secure jobs outside of Congress and thus raise turnover rates via mechanisms in addition to the actual election loss itself.
We specify two regression discontinuity models utilizing narrowly decided election results from the 2014, 2016, and 2018 elections.We focus on the offices of incumbents who received between 45% and 55% of votes cast for the top-two vote-receiving candidates in their respective elections.Both analyses are at the staffer level.One model is designed to test for anticipation effects in close elections and uses an indicator for whether staffers left Congress in the year of the election as the outcome variable.The other model uses an indicator for whether staffers left Congress the year after the election.We use a binary "treatment" variable indicating whether the

Leaves year after election (1)
(2) staffer worked for a legislator who lost their reelection campaign.
To adjust for the possibility that some offices always have higher turnover than others, we control the relevant office's turnover rate two years before the election.Some literature suggests that a staffer's gender influences their opportunities and role in the congressional workforce (Ritchie & You, 2021).Accordingly, we include the gender of staffers in our models.We also account for the party affiliation of staffers.
Key results are shown in Table 1.Our results suggest that, in close elections, there are minimal anticipation effects by staffers.Staffers working for a member who ultimately lost reelection do not leave Congress during election years at statistically significantly higher rates than those that do not.Critically, however, we find that member turnover causes staffer turnover.Specifically, staffers working for members involved in narrow election losses have about 56% higher odds of leaving Congress the following year, relative to staffers who work for successfully reelected members.This stylized fact alone, in our view, justifies using elections as a proxy for career uncertainty, as we do in the next section to assess our theory.

Career uncertainty and training
Since member turnover makes that member's staffers less likely to stay in the congressional workforce, we can use it to test whether expertise is a general or firm specific.If expertise is firm specific, then member turnover decreases the staffer's incentive to attend training and the member's incentive to bear the cost of the training.Oversight expertise loses much of its value if the staffer leaves Congress, so the shorter their expected tenure, the less likely they are to acquire the expertise.
To test for this possibility, we analyze attendance at bootcamps and seminars.The unit of analysis is a legislator training-session dyad.The outcome is how many staffers that legislator's office sent to that particular seminar or bootcamp.We regress this outcome on whether that particular training session took place during the legislator's final term in office, as this represents a period of increased career uncertainty for staffers.Furthermore, we include legislator and training-session fixed effects.The legislator fixed effects make the analysis a within-legislator comparison and thereby accounts for the possibility that higher quality legislators are both more likely to send staffers to training and less likely to lose office.Holding the number of staffers a member sends to trainings over the span of the data constant, do fewer staffers attend trainings held

House and Senate
Only Senate

Hours of training
Final term in chamber during that member's last term in office?The training fixed effects account for the fact that bootcamps have fewer attendees than seminars, that the popularity of the program may fluctuate over time, and that some sessions might be held at more convenient times of the year than others.We cluster standard errors at the legislator level.This analysis assumes that staffers anticipate that their member might leave office at end of the term.Because these contingencies affect whether the staffer needs to look for other work, staffers have strong incentives to figure out if their member faces a threatening primary or a difficult general election, suffers from health issues, or otherwise does not wish to remain in office.
Table 2 shows that staffers are less likely to attend trainings during their member's last term in office.The first column looks at the number of staffers from that office attend each training, and the second column looks at the number of hours staffers from that office spend at that training, which reflects that bootcamps take 12 times as long as seminars.The average office sends .025staffers to each event and log an average of .036hours of training per event, so the final term effects of .015fewer staffers and .020fewer hours of training per event is a substantively significant, relative effect.
The third and fourth columns offer a robustness check that does not depend on the assumption that staffers can anticipate if their member is about to leave.It restricts the analysis to the Senate and makes the main independent variable whether senators who are up for reelection that cycle.Members who are up for reelection could lose in either the general or the primary, but that cycle is also a particularly attractive time to retire because it maximizes the amount of time the legislator is in office while avoiding a costly reelec-tion campaign.Of senators up for reelection, 27.7% leave by the beginning of the next congress, compared to only 4.1% who are not up for reelection that cycle.
The third and fourth columns of Table 2 finds similar results to the baseline analysis, albeit with much less precision.The point estimates are about the same, but the standard errors are much wider, which is also understandable because being up for reelection is a noisy proxy for whether the senator will actually leave office and because restricting the analysis to senators discards over 80% of the data.Nevertheless, the fact that an analysis which discard so much data and uses a coarser but conceptually related version of our main independent variable yields such similar results serves to bolster our main results.

Expertise acquisition and the separation of powers
We next examine how the separation of powers moderates the impact of career uncertainty.If staffers hesitate to acquire expertise on their own because it is a form of firm specific human capital that is difficult to transfer to other careers, then staffers ought to be more likely to acquire expertise when there are more job opportunities where that particular form of human capital would be valuable.We test this additional implication of the theory using a major shock to the congressional labor market: changing the party of the President.
The President and his administration control a massive number of appointed positions in the Executive Branch, and they give virtually all of these positions to members of their own party.The Executive Branch is one of the few places outside of Congress where expertise in oversight would be useful, because the

House and Senate
Only Senate

Hours of training
Opposite Executive Branch is the target of congressional oversight.Not surprisingly, numerous "alumni" of the oversight training we examine go on to work in federal agencies.
Accordingly, congressional staffers should be less responsive to career uncertainty if their party controls the presidency.In Congress, their oversight expertise will be valuable for helping provide a counterpoint to the opposition party in hearings.In executive branch positions, that expertise should be valuable for defending the targets of oversight from the inside.Thus, we expect the importance of career uncertainty for the in-party to diminish, relative to the opposition.
To test this hypothesis, we replicate the main analysis in Table 2 with an interaction for whether the staffer's member of Congress is from the opposite party as the President.The theory predicts that the interaction of the Final Term in Chamber and Seat Up for Reelection variables with Opposite Party of the President should be negative.
This is what Table 3 shows.The first-order effect of being from the opposition is positive.This is not surprising.Members from the opposite party of the President have more to gain from oversight, so members might encourage their staff, to attend these trainings.More importantly for our theory, the interaction between being from the opposition and both measures of career uncertainty is negative.Staffers are more sensitive to career uncertainty in Congress when their party does not control the presidency.The coef-ficient is only statistically significant in conventional levels in one of the models, but it nearly attains statistical significance in the other specification that uses both House and Senate data (p = .14).As Table 2 showed, the baseline analysis suggests that the Senateonly specifications are underpowered even without the interaction term, so it is not surprising that the interaction terms are not precisely estimated.
Taken together, this analysis suggests that the technical knowledge about how to conduct oversight is firm specific human capital.If a staffer works for a member who is about to leave office, then they are less likely to attend, unless the party that controls the presidency gives them an attractive outside option.These substantive findings are also robust to other specifications that include different factors affecting the Congressional labor market.For instance, we might be concerned that legislators' final term in office may be positively correlated with changes in majority status.We show in Section G (in the online supporting information, p. A15) that our findings about the relationship between career uncertainty and expertise acquisition are robust when majority status is treated as a moderating variable.

Cost to the employer and training
If expertise in oversight is firm specific human capital, then the employer must bear at least part of the cost of training.This implies that the lower the cost to the legislator, the more likely the legislator's staffers are to acquire training.Even though there is no tuition for the bootcamps or seminars, attendance still requires a sacrifice from the staffers' offices: staffers' time.Seminars were held on Fridays and 84.6% of the bootcamps (all but four) were held entirely on weekdays.Any time staffers spend at these events is time they did not spend on other work.While staffers' time is always valuable to their members, sometimes it is slightly less valuable.The House and Senate occasionally take extended breaks from legislative session, such as for major holidays, district work periods, and the August recess.Staffers in Washington are still expected to work during these breaks, but they have fewer responsibilities.Consequently, offices give up less by sending staffers to training when their chamber is on a break and ought to be more inclined to send staffers during break than while the chamber is in session.
Accordingly, Table 4 conducts a training-level analysis of how many staffers from each chamber attended as a function of whether the relevant chamber was on a break.A chamber is on a break if, on the day the training was held, it has not held a roll-call vote in the previous five days or if it will not hold a roll-call vote in the next five days.
Table 4 shows that, at least in the House, more staffers attend the training if it is held during a break.This stronger effect relative to the Senate, and the pooled sample likely reflects the greater demands on the House's time.House offices have smaller staffs than Senate offices, and the House generally considers more legislation than the Senate does.These compounding pressures make staffers' time more valuable in the House, so House offices are particularly sensitive to opportunities to train their staffers.While we believe we would also observe this pattern in the Senate if we could find a measure of the costliness of the training to the legislator that senators felt more acutely, Table 4 offers only provisional support that increasing the cost of the training to the legislator decreases the amount of expertise staffers acquire.

Expertise and career longevity
Finally, we examine the downstream implications of expertise acquisition on time spent working in Congress.If expertise is firm specific, then staffers interested who acquire expertise will stay in Congress for longer than those who do not.First, staffers who have firm specific expertise are more valuable to their employers, so they are less likely to be laid off.Second, although it is difficult for staffers to convert firm specific expertise into better-compensated outside options, if a staffer does get an attractive outside offer, the legislator they work for will be willing to increase the staffer's compensation in order to retain them.This makes staffers with oversight expertise less likely to quit.Suppose, for example, the staffer has a J.D. with a focus in securities law.If a bull market creates sudden demand for lawyers who know securities law, the staffer might want to quit Congress to return to the legal profession.However, if she also has expertise in oversight, her employer might be willing to increase her compensation by enough to prevent her from leaving.
Unfortunately We argue that interest in the training sessions is an expression of some general, inherent motivation to acquire expertise.Thus, staffers with this predisposition may pursue expertise acquisition in a variety of ways aside from bootcamp or seminar attendance, make decisions to actively improve their performance, and be generally incentivized by the development of specialized skills and knowledge.If expertise in Congress is firm specific, then we predict staffers interested in training will have longer careers on the Hill.Conversely, if expertise is general, then we would expect interested staffers to have shorter Congressional careers.
This analysis also controls for a number of important time-varying covariates that may affect career length: staffer's earnings, dummy variables for whether or not they worked in the Senate, held a policy-focused position, held a senior staff position, or worked for a committee.The online supporting information describes these job categorizations and provides summary statistics about attendees with respect to these categorizations in greater detail in Sections I (p.A17) and F (p. A14) respectively.Table 5 shows that staffers who express interest in training stay in Congress for longer (have a lower hazard rate) than those who do not.We do not mean to suggest that merely expressing interest in attending a bootcamp or seminar causes staffers to have longer careers.
Rather, staffers who express an interest in training are more likely to get oversight expertise somewhere, that this firm specific human capital makes them more valuable to the legislators they serve, and hence these legislators work to retain staffers who would otherwise leave.
Another possibility is that the results follow from selection bias: only staffers who seek out training are those who intend to stay in Congress for a long time.However, this alternative mechanism would lead to the same substantive conclusion as our original interpretation: oversight expertise is firm specific.If it were general, then even staffers who did not intend to stay in Congress would want it.Another objection is that this test follows from our first test.Table 2 has already shown that staffers are less likely to attend training during their member's final term in office, so perhaps the effect in Table 4 is mechanical: some staffers do not express interest because they anticipate their member will soon leave office and that member's staff will then be forced out of the congressional workforce.
To account for this possibility, Table E3 in the online supporting information (p.A14) replicates our analysis using only staffers who either left Congress before their member did, committee staffers, and staffers who were still working in Congress at the end of the study period.This restricted sample excludes all staffers who had to leave Congress because their member was not reelected.The results in Table E3 are consistent with the results in Table 4, which shows that the effect cannot be attributed to forced exit due to member turnover and that the results in Table 5 provide a distinct and informative test about whether oversight expertise is general or firm specific.
The theory explains that the reason that the staffers who acquire oversight expertise stay longer is that Congress compensates these staffers more generously to prevent them from quitting.Compensation can take many forms, but we focus on just one: getting promoted to a better job.To support the theory's predicted mechanism for our findings on staffers' longevity, we show in Section H in the online supporting information (p.A16) that staffers with oversight expertise are more rapidly promoted to policy-oriented positions

DISCUSSION
Congress needs expertise to perform its constitutional duties, but it has struggled to attract, train, and retain a large workforce of expert staffers.In response to broad beliefs that congressional performance has declined, nonprofits have stepped in to assist Congress in training its workforce, but very little is known about these efforts.Their effectiveness is, in part, a function of the individual incentives of congressional staffers.This study investigates these incentives using novel records of training conducted in Congress for one particular kind of expertise: congressional oversight.
Taken together, our tests suggest that expertise in oversight is a form of firm specific human capital.One implication relevant for contemporary debates about congressional capacity is that, for this form of expertise, staffers' compensation does not increase in proportion to the value of their expertise.There are practical implications for this insight.Nonprofits may be able to train more staffers by making participation cheaper for employers-scheduling training during recesses and the lame duck session.However, scheduling the trainings on nights or weekends could backfire, since that would shift the cost from employers to staffers.
Our analysis also yields important insights for the study of Congressional capacity.High turnover among legislators impedes the acquisition of firm specific expertise from two directions.First, it makes legislators less inclined to send their staff to training, since legislators who will soon leave office gain less from investments in expertise.Second, it makes staffers less inclined to acquire expertise, since there is a better than even chance that they will leave Congress if their boss does, and expertise in oversight does not do much to help them find a better job after they leave.As a result, anything that induces legislators to stay in office for longer-like more input into the legislative agenda, higher salaries, or fewer fundraising obligations-would also increase the staff's expertise in oversight.Institutions that approximated civil service protections, such as expanded committee staffs with some degree of job security across successive committee leaders, would likewise encourage the development of firm specific expertise.Conversely, reforms that would decrease the typical tenure of a staffer-such as legislative term limits-would reduce their motivation to become experts in the particular functions of Congress.This point is somewhat counterintuitive, as opponents of legislative term limits sometimes argue they empower staff.Short time hori-zons may discourage both legislators and staff from specializing.
The same may not be true of other forms of expertise.Although oversight expertise is firm specific, other kinds of expertise, such as substantive policy expertise, might be general.Getting staff to develop general human capital is much easier.Since getting it increases their compensation, staffers seek out training on their own, without any prodding or subsidies from their employers.Turnover among the legislators themselves would not deter these staffers; in fact, knowing that they might need to find a new job soon could actually spur them to seek out training.But developing general human capital poses its own unique challenges; it replaces the problem of incentivizing staffers to acquire expertise with the problem of retaining them once they have it.The scarce financial resources and unpredictable work schedules on Capitol Hill could make it difficult for Congress to keep its best staffers in the face of outside competition.
This implies discussions about congressional capacity should disaggregate capacity into its constituent parts.Some elements of capacity require firm specific expertise, and others probably require general expertise.Reforms that would help cultivate firm specific kinds of expertise could backfire if applied to cultivate general expertise, and vice versa.
Fortunately, there are several possible paths to further investigating these questions.We examined a single type of training.But there are many more, offered by different institutional partners, and some may be more or less firm specific.In addition, recent rules changes in the House mean that congressional staffers will have greater access to subsidies for outside trainings.These smaller-scale reforms, limited to one chamber, represent another opportunity to analyze this labor pool using the framework we have applied.
Our analysis also suggests new questions for political science.We found that the negative effect of career uncertainty on a staffer's willingness to acquire firm specific expertise diminishes when their party controls the presidency.Presumably, opportunities in the executive branch may also siphon experts off the congressional staff.What is the net effect of these two countervailing forces on the level of different kinds of expertise on the congressional staff?Additionally, our analysis treats Congress as one giant firm because it is a useful starting point for understanding human capital in the legislative process.In fact, Congress is more like a cartel composed of many semiautonomous offices and committees.Our empirical results suggest that this internal market for expertise is not large or robust enough to make Congress-specific expertise effectively general, so our treatment of Congress as one giant firm is accurate to a first approximation.Nevertheless, future research could develop a more 15405907, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ajps.12848 by University Of Michigan Library, Wiley Online Library on [26/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License complete understanding of the obstacles to expertise on the congressional staff by studying competition between congressional offices.

A C K N O W L E D G M E N T S
We thank Jesse Crosson, Sean Gailmard, Tim LaPira, Tessa Provins, and seminar participants at APSA, MPSA, UC San Diego, the Center for Effective Lawmaking, and the Center for Effective Government American Politics Conference for their helpful comments and suggestions.We acknowledge funding from the College of Literature, Sciences, and the Arts at the University of Michigan.This research has been deemed exempt from IRB review (#HUM00175889).We thank the Project on Government Oversight, the Levin Center, and the Lugar Center for sharing their data.Special thanks are due to Elise Bean.The authors bear sole responsibility for any errors.The views expressed in this study are those of the authors and not of the Project on Government Oversight, the Levin Center, or the Lugar Center.

R E F E R E N C E S
Turnover in Congressional Offices.Note: Plots turnover rates by office across different electoral outcomes for every Congressional election from 2010 to 2018.T A B L E 1 Re-election and departure.
Career uncertainty and training attendance.
Career uncertainty, party control, and training.
15405907, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ajps.12848 by University Of Michigan Library, Wiley Online Library on [26/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 15405907, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ajps.12848 by University Of Michigan Library, Wiley Online Library on [26/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Reports regression coefficients and standard errors clustered by legislator.Outcomes are indicators of participation in training sessions.
15405907, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ajps.12848 by University Of Michigan Library, Wiley Online Library on [26/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 15405907, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ajps.12848 by University Of Michigan Library, Wiley Online Library on [26/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Attendance and timing.Reports coefficients and standard errors for models that predict the number of staff attendees at 82 unique training sessions.
, our data does not tell us which staffers have expertise in oversight.It tells us which staffers attended a particular set of trainings.This rules out what would otherwise be an attractive research design: comparing staffers who attended oversight training to those who applied for oversight training but did not attend.Those staffers who did not get to attend might get oversight training elsewhere.There are other opportunities to acquire training while working in Congress, offered through Congressional Staff Academy, the Brookings Institution, and the Congressional Research Service, to name a few.The presence of other trainings would attenuate the measurable effect of attending one of the trainings we observe.Consistent with this possibility, our analysis in the online supporting information (see Table E2 on p. A13) finds no evidence that attending training increases how long staffers stay in Congress, compared to those who just express interest.Instead, we test whether staffers who express interest in the training stay in Congress longer than those who do not.We utilize a number of staffer-level statistical tests conducted with Cox proportional 15405907, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ajps.12848 by University Of Michigan Library, Wiley Online Library on [26/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License hazards models with fixed and time-dependent covariates.The outcome is the number of days a staffer remained in Congress after the first training session that he or she could have attended. 7Thus, the outcome variable in all tests is the number of days between the date of this training session and either the last date the staffer was employed by Congress or the latest date available in our Congressional staff data (March 31, 2022).A staffer is considered to have expressed interest in training if they appear anywhere in the records.
Training and career length.
T A B L E 5Notes: Reports coefficients and standard errors from Cox proportional hazard models predicting departure from Congress.† p < .1;*p < .05;**p < .01.
15405907, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/ajps.12848 by University Of Michigan Library, Wiley Online Library on [26/03/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License and to senior staff than staffers who do not express interest in the training.