The revolving door between government service and the private sector has garnered attention for decades, with the bulk of that discussion centering on congressional lobbyists and, to a lesser degree, regulatory capture in government agencies (e.g., McCrain Reference McCrain2018; i Vidal, Draco, and Fons-Rosen Reference i Vidal, Draca and Fons-Rosen2012; Cohen Reference Cohen1986).Footnote 2 However, the revolving door of federal criminal prosecution – where officials, most commonly Assistant U.S. Attorneys (AUSAs), leave government service to defend the same types of cases and clients they once prosecuted – has not received systematic scholarly examination. Consider the example of former AUSA Michael Loucks, profiled in the New York Times as “arguably the nation’s most influential prosecutor of health care fraud” and “a cross between a firebrand preacher and a charismatic litigator” (Wilson Reference Wilson2011). A year after leaving the U.S. Attorney’s Office (USAO) in Boston, where he created a health care fraud unit and secured a record $2.3 billion settlement from Pfizer, Loucks joined Skadden, Arps. There he became an advocate for the pharmaceutical companies he had spent years prosecuting (Wilson Reference Wilson2011).
Loucks’s transition attracted media attention, but similar career moves by government attorneys to private practice occur regularly with far less fanfare. Such transitions raise basic questions about the relationship between career incentives and the enforcement of corporate criminal law. For instance, if an AUSA aspires to private practice, to what degree might that career objective impact her conduct while still in government? More to the point, do those who transition from AUSAs to private practice – which we term “revolvers” – behave differently than their counterparts who do not? If such systematic differences do indeed exist, they raise important questions about how career incentives shape prosecutorial conduct. Moreover, if revolvers are positioning themselves for future employment, we would anticipate their lenience to be most pronounced when large law firms – revolvers’ likely future employers – are directly involved in cases.
The answers to these questions have implications for corporate accountability, prosecutorial independence, and discretion. Nor do they operate in a vacuum, especially given widespread criticism of the government’s efforts to ensure corporate accountability in the years following the 2008 financial crisis. Senior U.S. District Judge Jed Rakoff, who sits in the Southern District of New York, notably called this “one of the most egregious failures of the criminal justice system in many years” (Rakoff Reference Rakoff2014). Though he ultimately minimizes the revolving door’s role in that particular failure,Footnote 3 his acknowledgment of its potential influence highlights an important gap: there is a lack of systematic, as opposed to impressionistic, evidence about how the career incentives of prosecutors may shape their behavior in government service. It is this gap we seek to fill.
With that in mind, our paper proceeds as follows. First, we examine existing literature on revolving door lobbying, focusing on the congressional context where most recent contributions have been made. We then extend relevant insights from that literature to the domain of federal prosecution. Here we explore the value proposition of hiring former prosecutors, discussing the informational advantages and network connections they bring to the table. After considering institutional context and its intersection with individual motivations, we introduce our hypotheses and research design. Then we describe the data we have gathered and explain how we harness it to address our research questions. Next, we present our findings, discuss them, and conclude with their implications for future research.
The revolving door: Lessons from the lobbying context
Some of the most extensive research on the revolving door phenomenon has come from the legislative context. That work offers important insights that help frame the ways in which career incentives may relate to behavior in the context of federal prosecution. Although some may conceptualize the value of former governmental officials to private employers in terms of theoretically distinct mechanisms, in practice, they likely operate as complementary avenues by which those in government service are in private sector demand. First, former government officials possess specialized, “insider” knowledge – about policy processes, institutional procedures, and relevant personalities – that they can leverage in support of private clients (e.g., McCrain Reference McCrain2018; LaPira and Thomas Reference LaPira and Thomas2017; Makse Reference Makse2017). In the realm of congressional lobbying, this could include a deep understanding of committee dynamics or knowing which members are most integral or susceptible to influence (e.g., McCrain Reference McCrain2018). Those without such experience cannot as credibly lay claim to that specialized knowledge. There are clear applications of this concept to the prosecutorial context, where specialized knowledge might include intimate familiarity with charging practices, the informal office cultures of particular U.S. Attorneys’ Offices (USAOs) or divisions within them, or the nuances of plea-bargaining norms. Former prosecutors would also seem well-positioned to know which arguments might best resonate with certain judges or how different AUSAs approach cases.
This is no doubt the sort of insider knowledge potential employers expect former prosecutors to deploy – in support of smoother case negotiations, deferred prosecutions, or more favorable plea deals. The profile of former federal prosecutor Zachary Terwilliger, now a partner in Government Investigations & White Collar Criminal Defense at Vinson & Elkins, illustrates this perspectiveFootnote 4:
Zach’s knowledge of how the Justice Department and federal enforcement agencies function in both civil and criminal matters, particularly in regards to discretionary decision making, provides him with unique insights necessary to effectively assist companies and individuals facing federal enforcement matters, including allegations of domestic and international criminal and civil violations.
The second mechanism of influence discussed in lobbying scholarship might be termed “network-based access.” This speaks to the personal relationships former government employees possess by virtue of that service and, in practice, often represents the vehicle through which informational advantages are leveraged (see i Vidal, Draca, and Fons-Rosen Reference i Vidal, Draca and Fons-Rosen2012; McCrain Reference McCrain2018). These connections lower transaction costs for both parties, because current government officials are more likely to trust their former colleagues due to shared experiences and institutional culture. In the prosecutorial context, these might include relationships with sitting AUSAs, judges, or law enforcement agents. It is also likely that staff connections – which are critical in congressional lobbying (McCrain Reference McCrain2018, 1370) – are especially valuable in USAOs due to frequent turnover at the top.Footnote 5
When Boies Schiller Flexner added former AUSA Jacqueline Kelly (SDNY) to its white collar practice in 2025, the firm’s chairman alluded to these connections: “The nature and volume of work that a prosecutor at Jackie’s level acquires at SDNY, and the relationships across all levels of government that she has cultivated, make her an ideal addition to the firm.”Footnote 6 All this is consistent with the conclusion that “What Big Law seeks now is an ability to negotiate the mega-settlements and the inside knowledge of the institution and the government hive mind, to glean what constitutes cooperation for the Justice Department and what settlement it would deem a win.” For their part, many prosecutors “want their adversaries to imagine them as future partners” (Eisinger Reference Eisinger2017, 199). This insight parallels recent theoretical work by Awad, Judd, and Riquelme (Reference Awad, Judd and Riquelme2025), who develop a formal model in which politicians temper their positions to preserve relationships with interest groups. Prosecutors bound for private practice confront a similar trade-off: a reputation for being reasonable with large law firms may serve them better than a reputation for aggressive enforcement.
Recent empirical work has broadened understanding of the revolving door in ways that buttress these conclusions. For one thing, scholars have extended insights beyond the legislative realm to encompass executive branch contexts. Bolton and McCrain (Reference Bolton and McCrain2025) examine federal bureaucrats who become lobbyists; they find that expertise and connections to political leadership predict revolving. Further, employing former executive branch officials has been shown to increase access to policymakers in agencies from the SEC (Ban and You Reference Ban and Young You2019) to the EU Commission (Egerod, Rasmussen, and van der Ploeg Reference Egerod, Rasmussen and van der Ploeg2024). There is also evidence that presidential appointees to the Farm Service Agency who move into lobbying are more proactive in agency communications than those who do not (Liu Reference Liu2024).
Other research reinforces the importance of connections and expertise. One team found that staffers experience a 24% drop in generated revenue and are less likely to remain lobbyists after those for whom they worked leave office (i Vidal, Draca, and Fons-Rosen Reference i Vidal, Draca and Fons-Rosen2012, 3732). Strickland’s (Reference Strickland2020) state-level study confirms the declining effectiveness of lobbyists as their connections with former colleagues attenuate. Black and Owens (Reference Black and Owens2021) uncover a similar relationship in their study of former Supreme Court clerks. They find that attorneys who once clerked for a justice are significantly more likely to capture that justice’s vote than is an otherwise identical attorney who did not clerk for that justice. The authors interpret this to suggest that former clerks, like others who traverse the revolving door, “may succeed because they capitalize on personalized information that few others enjoy” (Black and Owens Reference Black and Owens2021, 796). This idea of officials seeking to position themselves strategically for future opportunities extends beyond post-exit employment advantages. In earlier work, Black and Owens (Reference Black and Owens2016) show that judges on the U.S. Courts of Appeals alter their behavior when angling for Supreme Court appointment – suggesting that career advancement incentives systematically shape judicial behavior even before any transition occurs. This parallel supports our argument that prosecutors will similarly adjust their behavior in anticipation of private sector opportunities.
Yet, while all this underscores the value of knowledge and connections for post-service careers, it raises an important antecedent question: if future prospects depend on demonstrating competence and maintaining relationships, just how might AUSAs seek to do that? As illustrated above, scholarship on revolving-door dynamics has tended to focus on what happens after someone leaves government – how former staffers or regulators utilize their connections and expertise in the private sphere. But recent research suggests career concerns also help explain conduct in government service. Shepherd and You’s study (Reference Shepherd and Young You2020) is instructive here; they find offices employing future lobbyists show increased productivity, particularly when staffers near the end of their time in government. Further, they find that “career-minded staffers showcase their skills to prospective future employers” strategically – they grant more access to lobbying firms than non-revolvers (Reference Shepherd and Young You2020, 28) and focus disproportionately on issues like health care and commerce that are a key focus of the lobbying industry (Shepherd and You Reference Shepherd and Young You2020, 271; Zheng Reference Zheng2015).
The prosecutorial setting offers advantages for testing these pre-exit dynamics. Compared to areas such as regulatory enforcement or legislative attention, prosecutions tend to yield more measurable outcomes – declination decisions, plea bargains, fine amounts, sentence recommendations – most of which are the products of some discretion; this creates room for prosecutors to calibrate their behavior based on various considerations. For reasons we expand upon shortly, that is particularly true of white-collar prosecutions of corporations – the cases we examine here. If, as Lord Scarman (Reference Scarman1981, 633) put it, “discretion is the art of suiting action to particular circumstances,” then individual career motivations could well constitute one such circumstance.
Having discussed the theoretical foundations of dynamics related to the revolving door and how they might inform prosecutorial behavior, in the next section we consider the institutional context that shapes AUSAs and their career incentives. But, before doing so, we wish to highlight an aspect of our theoretical perspective. While existing work has emphasized structural or institutional factors that affect federal prosecutorial personnel more-or-less uniformly, we take a more nuanced view that adds space for the operation of individual motivations alongside those systematic pressures. After all, if structural forces were the only explanation of revolving door effects, there would be no reason to anticipate – as we do – that revolving and non-revolving AUSAs might exhibit behavioral differences, much less do so in ways that would appear tactically advantageous for those planning transitions to private practice.
Institutional forces and individual motivations
Federal prosecutors operate within an institutional structure, and that structure shapes their work and career prospects. For example, resource limitations, the need to meet performance metrics, and attention to hierarchical oversight inside the USAO and from the Department of Justice (DOJ) are part-and-parcel to any AUSA’s environment (Richman Reference Richman1999). These can serve as constraints, as Brandon Garrett’s work illustrates in the white-collar context. As one example, he notes that institutional priorities and limited resources necessitate tradeoffs: “bringing large quantities of relatively small individual immigration, firearms, and drug cases” makes it more difficult to channel attention to more resource-intensive corporate crimes (Garrett Reference Garrett2020, 113–114).
In addition to these sorts of resource constraints, there is an additional strategic dimension relevant to white-collar prosecutions – the trade-off between pursuing individual executives versus corporate entities themselves. The evolution of policy directives from Main Justice – another type of institutional factor – helps illustrate this point. In the late 1990s the DOJ’s so-called “Holder Memo” instructed prosecutors to weigh the economic harm to third parties that might result from pursuing criminal cases against large corporations (Eisinger Reference Eisinger2017).Footnote 7 As Professor Jennifer Arlen notes, “What the Holder memo did was announce publicly and officially that the focus of corporate enforcement should be on individual convictions, on getting the individuals responsible” (Berman Reference Berman2013). By 2015, on the heels of such criticism, the “Yates Memo” attempted to reemphasize DOJ’s expectations for accountability at the corporate level. As then-Deputy Attorney General Sally Yates put it, “We’re not going to be accepting a company’s cooperation when they just offer up the vice president in charge of going to jail” (Apuzzo and Protess Reference Apuzzo and Protess2015; Garrett Reference Garrett2020, 129).
Directives like these can create additional layers of oversight and, in theory, bring greater standardization to prosecution by requiring supervisory approval or input from Main Justice on significant decisions (e.g., Miller and Curry Reference Miller and Curry2019). But they also highlight the complexity prosecutors face in white-collar prosecutions. On the one hand, doggedly pursuing high-profile individual convictions can be “career making” (Rakoff Reference Rakoff2014, 5), both because they indicate competence and generate visibility. On the other hand, corporate defendants offer greater opportunities for AUSAs to pursue the sorts of negotiated settlements that may be attractive to future private employers.
This discretion in the face of complexity tends to be especially pronounced in corporate prosecutions. The DOJ’s United States Attorneys’ Manual (USAM) says as much in its “Principles of Federal Prosecution of Business Organizations.” “Due to the nature of the corporate ‘person,’” the USAM directs prosecutors to consider ten distinct factors beyond those applied in individual cases, including “the pervasiveness of wrongdoing,” “collateral consequences,” choosing between criminal prosecution and “civil and regulatory alternatives,” and whether some individual, non-corporate prosecution might suffice (U.S. Attorneys’ Manual § 9-28.000, § 9-28.300). Prosecutors are instructed to “exercise their thoughtful and pragmatic judgment in applying and balancing these factors” (USAM 2006, § 9-28.300).Footnote 8 Our focus in this paper is solely on corporate prosecutions, where these considerations suggest discretion and complexity should be most pronounced. Unlike individual cases where conviction rates may be the primary metric for success, corporate cases give prosecutors significant space to demonstrate the strategic judgment that private employers value.Footnote 9
The distinction between individual prosecutions and prosecutions of corporations is important, but we believe that the signal from individual prosecutions is likely quite noisy for future employers for a few reasons. First, charging an individual is potentially much more about moral attribution (as it might revolve around decisions about what works best for deterrence, for instance) rather than marketable competence (e.g., how to resolve corporate exposure). Second, the signal from charging an individual is likely to be noisy because institutional oversight on this issue is high and frequently oscillates, as illustrated by the Holder and Yates memos, as well as subsequent retrenchment in charging individuals post-Yates. Third, while it is true that individual prosecutions and convictions may be “career-making” in the words of Rakoff (Reference Rakoff2014), it is likely that such convictions are more important for advancement within government service than for outside exit options, given the noisiness of the signal.Footnote 10
Changes in oversight requirements have also been paralleled by broader organizational changes within USAOs. As Barkow (Reference Barkow2009) chronicles, office hierarchies have grown more stratified, and there is typically a clear delineation between so-called “front office” personnel and line prosecutors. In nearly every USAO, this hierarchical expansion has created two high-level supervisors between line prosecutors and the U.S. Attorney (e.g., Miller and Curry Reference Miller and Curry2019, 10). Beyond its implications for internal career advancement – the United States Attorney controls elevation to those supervisory positions – this largely uniform structure means that the line AUSAs who handle most corporate prosecutions face similar constraints regardless of district (Barkow Reference Barkow2009).
While these institutional factors essentially apply to all AUSAs, they can create different individual incentives depending, at least in part, on one’s career aspirations. It is precisely within this space between institutional constraint and individual discretion that such motivations may influence prosecutorial behavior. For instance, even when approvals are required by superiors, line prosecutors retain a level of informational asymmetry and, hence, significant control – they can frame recommendations by emphasizing particular facts or in how they present cooperation efforts or describe compliance programs.
Historically, AUSAs tended to serve for a short time before moving on to other opportunities (Miller and Curry Reference Miller and Curry2019, 10). And, despite some geographic variations, the average longevity of AUSAs has increased in recent decades. Better civil service protections, higher salaries, and lifestyle considerations such as the absence of billable hour requirements are frequently cited as factors associated with that change (Suthers Reference Suthers2008; Lochner Reference Lochner2002).Footnote 11 Further, AUSAs bear responsibility for some of the most complex and high-stakes legal matters in the federal system, which can undoubtedly bring them significant personal and professional fulfillment. As we have discussed, the litigation experience, courtroom advocacy skills, and procedural expertise they acquire can also make those who choose to transition to white-collar defense or corporate law highly attractive to prestigious firms.
This growth of careerism and more robust job protection for AUSAs has a potentially important consequence for their behavior: on balance, it has (probably) increased their ability to operate independently of supervisory direction. It has become harder to remove AUSAs, even when their approaches conflict with office priorities or they stand out as underperformers.Footnote 12 In contrast to the days when AUSA spent shorter stints in government and could be more easily replaced, these developments may make it easier to pursue individual approaches in the face of stated institutional priorities.
Say an AUSA feels a corporate defendant warrants lenient treatment, perhaps via a deferred prosecution agreement or a relatively modest fine. A stated institutional preference toward aggressive prosecution notwithstanding, there are practical limits to the monitoring supervisors can engage in (see Miller Reference Miller2005). Lead prosecutors inevitably retain discretion when deciding what constitutes sufficient cooperation or an adequate fine (and when it comes to reporting their views to superiors); in effect, informational asymmetry allows them to apply their own standards within broad institutional guidelines before requesting approval for an action up the line. Even in the face of more uniform hierarchical and policy directives, individual AUSAs possess meaningful autonomy in how they approach cases – autonomy that may be influenced by personal career considerations rather than purely institutional factors. This is especially relevant for prosecutors considering a move to the private sector, because it allows them to calibrate their behavior in ways they believe can enhance their marketability. The significant discretion imbued in prosecutorial decision-making is precisely what allows these individual motivations to operate alongside more uniform institutional constraints.Footnote 13 Finally, as referenced in the previous section, former prosecutors offer corporate clients distinct advantages in white-collar defense. They have a refined understanding of prosecutorial decision-making and the plea negotiation dynamics that may well distinguish favorable outcomes from unfavorable ones. Existing relationships with key actors on the government side can supply valuable intelligence about priorities. Perhaps most importantly to corporate practice, former AUSAs bring instant credibility to the negotiating table – their experience serves as a signal that they can size up the situation accurately and appreciate the stakes.
The financial incentives for developing these skills are substantial: the highest salary earned by an AUSA with ten years’ experience is approximately $190,000, equity partners in top law firms earn about $1.9 million, with non-equity partners earning roughly $550,000 (Weiss Reference Weiss2024). These financial incentives go some distance in explaining why AUSAs might cultivate the expertise and relationships that would make them attractive to corporate clients. That credibility is an important commodity for the private firm, both in terms of its dealings with government prosecutors and with prospective clients who seek the assurance of experienced counsel.
Hypotheses, data, and research design
These theoretical insights from the lobbying literature discussed above suggest several testable expectations about how, in the context of federal prosecution, career incentives could influence AUSA behavior in corporate white-collar cases. If revolving door dynamics are such that future career prospects shape current conduct, we should observe systematic differences between prosecutors who transition to private practice (“revolvers”) and those who do not (“non-revolvers”). Our first two hypotheses embody this expectation using different indicators of prosecutorial leniencyFootnote 14:
Hypothesis 1 (Case Outcomes): Revolvers will be significantly more likely than non-revolvers to reach lenient case resolutions with corporate defendants in white-collar cases.
Hypothesis 2 (Fine Amounts): Compared to their non-revolving colleagues, revolvers will be associated with the imposition of significantly smaller fines on corporate defendants.
Our theory also suggests an additional, more contextual implication: if revolvers do, indeed, position themselves for future employment, their leniency should be most pronounced in corporate cases where large law firms are present as defense counsel. This flows from the reality that these prestigious and lucrative firms represent the most coveted private practice destination for departing AUSAs. This leads to two conditional expectations:
Hypothesis 3: The tendency of revolvers to produce lenient case outcomes will be exacerbated in cases where a large law firm is present.
Hypothesis 4: The tendency of revolvers to produce smaller fine amounts will be exacerbated in cases where a large law firm is present.
The data we use to test these expectations comes from the Corporate Prosecution Registry (CPR) (Garrett and Ashley Reference Garrett and Ashley2025), which provides comprehensive information on every federal organizational prosecution since 2001. We limited our search to the 16 federal judicial districts in or adjacent to the largest U.S. legal markets and gathered a list of all cases from those districts in the database from 2001 through July 1, 2025.Footnote 15 We utilized several variables previously coded in the database for our analyses and discuss those below. We also examined relevant docket sheets or additional information (e.g., copies of plea agreements) to identify the lead prosecutor(s) on each case, as well as up to three firms representing defendants in the action.Footnote 16 In the event more than three defense firms appeared, we prioritized capturing any large law firms whenever they were involved in the case. Based on that information, we undertook supplementary research to determine each lead AUSA’s next position – and specifically captured whether they joined an NLJ 500 law firm. We also recorded their law school and its ranking, the year they were admitted to the bar, their gender, and their federal district of service.
While complete descriptive statistics appear in the appendix, we offer a brief overview here. Our dataset includes 1,053 AUSAs from 16 districts who handled 1,588 corporate white-collar cases from 2001 to mid-2025. Of these AUSAs, 460 transitioned to private practice (revolvers) while 593 remained in government service or pursued some other career path after serving as an AUSA (non-revolvers). There is also variation in our outcome variables, with approximately one-third of all outcomes qualifying as “lenient” (i.e., something other than a trial or plea).
We utilize two dependent variables in our analyses, both of which are drawn from data compiled by the CPR. First, we relied on the CPR’s coding of case disposition – the categories are dismissal, declination, non-prosecution agreement, deferred prosecution, plea, and trial – to create an indicator of whether a corporate defendant was treated leniently. Any case CPR coded as a dismissal, declination, non-prosecution agreement, or deferred prosecution agreement was coded 1; the remaining outcomes (plea or trial) were given a value of zero. We draw a line between deferred prosecution agreements and pleas because this marks the boundary between a formal prosecution and a case where charges are held in abeyance. Additional information on the distribution of outcomes is available in the appendix. Our second dependent variable captures the total fines and fees paid by a defendant in each case.
With these dependent variables in hand, we coded a series of independent variables which can generally be clustered as capturing attorney- or case-based characteristics. Revolver is our central variable of interest, and those who enter private practice after leaving government service are coded 1; all else is set to zero. We incorporate the individual’s year of bar admission as a measure of experience; we capture whether or not the individual is a woman (1 if yes; zero otherwise); and law school rank as measured by U.S News for the year 2024.Footnote 17 Of note, we will use these pre-treatment (pre-revolving) variables (along with district identifiers and time) to help predict who will be a revolver to mitigate concerns about selection on type.
We incorporated variables tapping three case-related characteristics. Most centrally, given our hypotheses, we incorporated a dichotomous indicator capturing whether one or more firms associated with the defense in a case was a Large Law Firm. Cases involving a firm or firms appearing on the NLJ 500 in 2024 were coded 1, and cases where the NLJ 500 was not represented were coded zero.Footnote 18 The CPR codes variables to capture whether or not the case involved a financial institution and whether or not the case involved a U.S. Company (1 if yes; zero otherwise), and we include these in our models as well. Lastly, we include a control for partisan change to account for the possibility that changes in presidential administration may play some explanatory role in our modeling. This variable is coded 1 for years in which the administration changed partisanship, so in our data these are the years 2001, 2009, 2017, and 2021. In addition, we include fixed effects in most models for the district in which a case is heard, fixed effects for the area of law in which charges are brought, and a cubic spline to control for potential change over time based on the year in which a case is decided. Furthermore, most models also include two-clustered standard errors based on clustering at the level of both the attorney and the case. More information on these variables is included in the appendix.
Results
We begin with our model for whether a defendant received a lenient outcome. Table 1 below presents a series of logit models, adding additional complexity with each model including various fixed-effects and clustering of standard errors. In Models 2 through 4 the standard errors are clustered on both the case and the attorney. The most important takeaway in Table 1 is that in each model attorneys who revolve are significantly more likely to grant leniency to corporate defendants than are those who do not. The size of the effect for revolving attorneys varies somewhat across the models, ranging from a 14-percentage point increase in the likelihood of leniency in Model 1 to a 5-percentage point increase in Models 3 and 4. We take Model 5’s estimate as likely to be closest to the truth (given the presence of various controls), with an estimate of a 7-percentage point increase in the likelihood of a lenient outcome. Put slightly differently, the baseline likelihood of a lenient outcome is 30%, but in cases involving attorneys who eventually leave government directly for private practice it jumps to 37%. This evidence supports hypothesis 1. In the appendix, we present a multinomial logit model to highlight exactly what outcomes are increased by the presence of revolvers in a case. There we show that the major difference between revolvers and non-revolvers is that revolvers are significantly more likely to grant defendants non-prosecution deals than are non-revolvers (and are substantially less likely to grant plea deals).
Logit Model – Lenient Outcome
Table 1. Long description
A statistical table titled Logit Model – Lenient Outcome with columns for Lenient Outcome and Models 1 through 6.
Attorney Characteristics:
- Revolver: Coefficients range from 0.28 to 0.65, all significant at p < 0.05 or p < 0.01.
- Bar Admission: Values are near zero and non-significant.
- Woman: Coefficients range from -0.09 to -0.11, non-significant.
- Law School Rank: Values are -0.00 across all models.
- Predicted Rev.: 0.09 in Model 6.
Case Characteristics:
- Financial Inst.: Significant in Model 2 (0.77) and Model 3 (0.53).
- U S Company: Highly significant across Models 2-6, ranging from 1.31 to 1.38.
- Large Law Firm: Significant in Models 2-4 (0.40 to 0.46).
Controls:
- Partisan Change: Highly significant negative coefficients from -1.22 to -1.41.
- District F E: No in Model 1 and 3; Yes in others.
- Law F E: No in Model 1 and 2; Yes in others.
- Year Cubic Spline: Yes in Models 5 and 6.
Model Statistics:
- N: Ranges from 2339 to 2502.
- Likelihood Ratio: Increases from 54.41 in Model 1 to 552.55 in Model 5.
- P R E: Ranges from 0.00 to 0.18.
* is significant at p <0.05; ** is significant at p <0.01 (both two-tailed).
Finally, Model 6 helps to account for the potential that our results are caused by differences in type – that attorneys who revolve are fundamentally different than those who do not and that this underlying difference, and not a desire to appeal to future employers, is the source of the differences we observe. We accomplish this by estimating a model, provided in the appendix, to predict whether an attorney will revolve based on the following pre-treatment characteristics: whether the attorney is a woman, their date of first admission to the bar, the ranking of the law school attended, and fixed effects for judicial district and year. This model predicts who will be a revolver well, and we include those predictions in model 6 (predicted revolver). The key takeaway is that our results remain unchanged – revolvers continue to exhibit greater leniency than those who do not revolve. These results help to mitigate selection before treatment as a threat to inference, conditional on the observable characteristics of attorneys.
Several other variables are important predictors of outcomes in these cases. In most models, a defendant that is a financial institution is no more likely to receive leniency, while defendants that are publicly traded U.S. companies are significantly more likely to receive leniency – about 22 percentage points more likely in most models. Also, in some models (but not our preferred Model 5), defendants represented by large law firms receive more leniency, about 7 percentage points more likely in the models where the variable is significant. Notably, none of the other attorney characteristics, such as law school rank, are significant predictors of leniency in any model. Lastly, years involving partisan change see decreased leniency for corporate defendants. On average, about 22-percentage points less likely.
Table 2 repeats the exercise illustrated in Table 1 but substitutes the natural log of the total fine assessed against the defendant in a case as the dependent variable. As before, we progressively add variables and fixed effects to the regressions across the models. Models 8 through 11 utilize standard errors clustered on both case and attorney. Model 12 includes our prediction of who is likely to revolve, as in Model 6 above.
Regression Model – Log of Total Payment
Table 2. Long description
A table with seven columns. The first column lists variables, and the subsequent columns are labeled Model 7 through Model 12.
Attorney Characteristics:
- Revolver: Values range from 0.07 in Model 7 to -0.90 in Model 8 (significant at p < 0.01) and -0.69 in Model 12 (significant at p < 0.05).
- Bar Admission: Values are approximately 0.04 to 0.05, significant in Models 9 and 10.
- Woman: Values range from 0.02 to -0.15, none significant.
- Law School Rank: Consistent at -0.01 across models.
- Predicted Rev.: Only present in Model 12 with a value of 1.39.
Case Characteristics:
- Financial Inst.: Significant positive coefficients around 2.13 to 2.66 across Models 8 through 12.
- U S Company: Significant positive coefficients around 3.59 to 3.85 across Models 8 through 12.
- Large Law Firm: Significant positive coefficients around 1.72 to 2.28 across Models 8 through 12.
Controls and Fixed Effects:
- Partisan Change: Values range from 0.15 to 0.48, none significant.
- District F E: No in Models 7 and 9; Yes in others.
- Law F E: No in Models 7 and 8; Yes in others.
- Year Cubic Spline: No in Models 7 through 10; Yes in Models 11 and 12.
Model Statistics:
- Constant: Ranges from 10.64 in Model 7 to 425.19 in Model 12.
- N: Decreases from 2484 in Model 7 to 2339 in Models 9 through 12.
- F Test: Significant at 0.00 for Models 8 through 12.
- Adj-R super 2: Increases from 0.00 in Model 7 to 0.16 in Models 11 and 12.
* is significant at p <0.05; ** is significant at p <0.01 (both two-tailed).
Results on fines are slightly less clear than they are for leniency of outcome, but a clear enough pattern emerges: revolvers typically levy smaller fines than non-revolvers. Using the estimates from Model 10 (again, the model with the most controls), revolvers levy fines that are $28,500 ($20,150, $40,260) smaller than in cases with non-revolving attorneys. This is evidence in favor of hypothesis 2. Further, as with leniency, we estimate a model including a predictor of who is most likely to revolve based on pre-treatment characteristics. Our results are robust to the inclusion of this predictor variable, which again strengthens our confidence that (conditional on observed characteristics of attorneys) it is not the case that some type of selection before treatment is responsible for our observed results.
Case characteristics are key predictors of the total fines paid by corporations in these cases, but attorney characteristics (other than being a revolver) are not. Cases involving financial institutions also see larger fines, as do cases involving U.S. public companies. Lastly, cases defended by large law firms also tend to result in higher fines. Each of these results suggests, intuitively, that fines are larger in cases that involve wealthier defendants.
Taken together, the results clearly suggest that attorneys who are revolvers grant more lenient outcomes for defendants and ask them to pay smaller fines than do non-revolvers. Although we are not entitled to make causal claims given the nature of our data, it is worth noting that we control for factors that are likely to make a comparison between revolvers and non-revolvers unfair. That is, fixed effects for district and area of law, along with the cubic spline and other case-level controls, make it unlikely that case assignments (which are not random in federal districts) are driving our results. Further, we have attempted to account for selection by type as a cofounder as well, without any change in our substantive conclusions. However, our case would be strengthened if we could show that revolver behavior is especially sensitive to the potential to audition for future positions, a task we set ourselves to in the next section.
Heterogeneous effects
To test hypotheses 3 and 4 about the circumstances where revolving door effects should be most visible, we turn to a series of models in which we interact revolving status with characteristics of the case. Figure 1 below displays the results of this exercise for the key variables of interest (full model results are available in Table 7A of the appendix). For the sake of comparison, we have included the effect from Model 5 as well in the first column. None of the interaction effects reach standard levels of statistical significance, although there is some evidence in favor of hypothesis 3 – revolvers are about 5 percentage points more lenient than non-revolvers in cases involving large law firms (NLJ) but this effect fails to reach traditional levels of statistical significance (p = 0.12). A similar effect, but larger (about 10 percentage points), exists for the interaction between revolvers and cases involving financial institutions (also with a p-value of 0.12). There is, then, only suggestive evidence that revolvers go easier on defendants represented by large law firms (hypothesis 3).
Interaction Results for Leniency.
Figure 1. Long description
A coefficient plot with a Y-axis labeled Change in Pred. Prob. of Lenient Outcome, ranging from negative 0.2 to positive 0.2. A horizontal dashed line marks the zero threshold. The X-axis contains four categorical labels.
Moving from left to right:
* Revolver: The point estimate is approximately positive 0.07 with error bars extending from 0.02 to 0.12, remaining entirely above the zero line.
* Revolver x N L J: The point estimate is approximately positive 0.05 with error bars extending from negative 0.01 to positive 0.11, crossing the zero line.
* Rev. x Fin. Inst.: The point estimate is approximately positive 0.11 with error bars extending from negative 0.01 to positive 0.23, crossing the zero line.
* Rev. x U. S. Comp.: The point estimate is approximately negative 0.02 with error bars extending from negative 0.15 to positive 0.11, crossing the zero line.
Figure 2 below displays a similar set of results looking at total fines rather than case outcomes; full model results appear in Table 8A of the appendix. Once again, we plot results from Model 10 in the first column for the sake of comparison. Only the interaction between revolvers and the presence of a large law firm is significant. Figure 2 displays results in terms of the change in the size of the log of total fines and so to get a better sense of the size of the effects we can translate into real dollar terms. When a revolver is involved in a case with a potential future employer, they levy considerably smaller fines than non-revolvers do. The average fine levied by non-revolvers in cases involving these large firms is $147,000, but the average fine levied by revolvers is $49,000, for a difference of $98,000. Interestingly, we do not find evidence that revolvers are more likely to go easy on financial institutions or U.S. public companies. These non-results are important because they suggest that the effect size we see for revolvers and large law firms is not likely a result of these firms simply defending large or more wealthy defendants. Rather, leniency in terms of smaller fines seems to be confined to the pairing of an AUSA who eventually revolves and a potential future employer.
Interaction Results for Total Fines.
Figure 2. Long description
The Y-axis is labeled Change in Pred. Size of Fine (Log Scale) with a range from -4 to 4 in increments of 2. A horizontal dashed reference line is set at 0. The X-axis contains four categorical labels.
From left to right, the data points are as follows.
* Revolver: A point estimate slightly below -0.5 with error bars extending from approximately 0 to -1.5.
* Revolver x N L J: A point estimate at approximately -1 with error bars extending from just below 0 to nearly -2.
* Rev. x Fin. Inst.: A point estimate at 1 with error bars extending from 3 to -1.
* Rev. x U.S. Comp.: A point estimate slightly below -0.5 with error bars extending from 1 to -2.5.
It could be the case that the result of lower fines in cases involving revolvers is a by-product of case assignment mechanisms – perhaps eventual revolvers are assigned (or get themselves assigned) to cases where lower fees are likely. Although we cannot fully dismiss this as a possibility, our data suggest otherwise. To help rule out this mechanism as a cause of our results, we re-estimated model 10, but limited ourselves only to cases involving large, publicly traded U.S. companies, thereby making a closer like-to-like comparison of defendants. This approach results in a lowered N (only 218 attorney-defendant dyads), but we still find a large negative difference between revolvers and non-revolvers when appearing before large law firms. Revolvers, compared to non-revolvers in cases involving only publicly traded U.S. companies, levy fines that are $28 million less. There is a tremendous difference, but it is based on a small number of dyads and is not statistically significant (p = 0.31). Nevertheless, the direction and the size of the difference are preliminary evidence that the differences in fines that we observe are probably not due simply to differences in the type of defendant revolvers and non-revolvers face.Footnote 19
In a small number of cases, the results differ between defendants within the same case. As an example, in a case involving fraud among medical providers in the Southern District of California, some defendants entered into plea deals while others got deferred prosecution agreements. Such divergence is useful from a research design perspective because it allows us to hold everything except the attorneys representing the United States and the defendant constant – given that these defendants are all handled under the same docket number the underlying facts giving rise to the case are the same. Only 6% of the cases in our data exhibit this kind of variation, which reduces our N considerably but still allows for some inference. Even in cases where the facts are virtually the same, cases handled by revolvers are treated differently. Revolvers are 6-percentage points more lenient than non-revolvers (p = 0.03) and levy smaller fines than do non-revolvers, by about $10,750 (p = 0.02). Put differently, this last set of results gives us added confidence that the differences we observe between revolvers and non-revolvers are not a by-product of non-random assignment of cases. Modeling results supporting these conclusions are available in the appendix.
Discussion
In this investigation of corporate white-collar prosecution and the revolving door, we found clear evidence that revolvers and non-revolvers behave differently – and that they do so in ways that seem unlikely to stem from case assignment differences. We also demonstrated the robustness of this difference on two fronts: the overall leniency of case resolutions as well as the fine amounts involved in those dispositions. Although there may be systematic institutional forces that suppress the prosecution of corporate crime generally, individual career incentives nevertheless create meaningful behavioral variation. Because those who revolve behave differently from those who do not within the cases that do get prosecuted, the results challenge purely structural explanations of prosecutorial behavior in this area. Ultimately, this suggests that individual career considerations can shape the enforcement of corporate criminal law.
In addition to this difference between the behavior of revolvers and non-revolvers generally, we uncovered evidence that such tendencies are exacerbated in circumstances where a revolver faces a large firm defending a corporate client. Specifically, fine amounts exhibited a significant interaction with a large law firm’s presence, but the dichotomous indicator of leniency did not – although this second set of results does trend in the predicted direction. While we can do little more than speculate about the sources of this distinction, several explanations seem plausible. One possible interpretation could be that revolvers may simply be more strategic about fine negotiations than general case dispositions. Perhaps case dispositions are more visible to supervisors, the media, and the public than are fine amounts. The latter are more technical and may be both less subject to constraints from office policies and more insulated from supervisory review or DOJ oversight. The pattern may also partially reflect the more limited variation in case outcomes themselves; fine amounts offer more continuous variation and may be better able than the dichotomous leniency indicator to capture subtler differences in prosecutor behavior in the presence of large firms. We see no reason why these explanations need to be mutually exclusive – the more technical nature of fine negotiations, their potentially reduced visibility, and the granularity differences in the measures could all contribute to the significant interactive pattern. If these mechanisms indeed have merit, we should expect this to offer prosecutors greater discretion to tailor their behavior toward the predilections of future potential employers and our continuous measure would be better equipped to detect that behavior.
All this is consistent with what we call a “strategic portfolio-building approach” where different aspects of prosecutorial behavior serve distinct but harmonic functions. Instead of relying on just one mechanism, AUSAs aspiring to private sector positions may use certain types of decisions to signal different attributes to potential employers. This allows AUSAs to demonstrate the sorts of skills and attributes private legal employers value – things like reasonableness, deal-making acumen, and pragmatism in negotiations. The patterns we observe suggest this portfolio-building approach may be more nuanced than any single behavioral strategy or decision point.
One limitation of our research design is that we still cannot definitively distinguish between multiple explanations for the differences we observe between revolvers and non-revolvers. That is, we cannot say whether those who revolve are different from the beginning or whether they take advantage of the opportunities offered to them. The best we can do is take the significant fine amount interaction with large law firms and what we gleaned from the within-case variation – where revolvers and non-revolvers handle defendants under the same docket with identical underlying facts – and pair it with the theoretical logic that career-minded prosecutors could seek to tailor their behavior to the desires of potential employers. Taken together, this evidence is suggestive of strategic behavior.
Our findings contribute to the robust literature on revolving door dynamics by extending evidence of pre-exit behavioral effects beyond traditional legislative settings and into the realm of federal law enforcement. Previous research has shown that legislative staffers strategically calibrate their behavior for potential employers while still in government service, and our results suggest similar anticipatory dynamics among AUSAs. The prosecutorial context offers unique advantages: it is focused on decisions that yield measurable outcomes, and the discretionary nature of charging and plea negotiations creates a window in which to observe how AUSA career motivations may help shape their official conduct.
In documenting systematic behavioral differences that correlate with future career transitions, we add to scholars’ understanding of how career incentives can condition the exercise of governmental authority in different institutional settings. Future research might examine whether the sorts of differences we observe in the corporate context persist in individual white-collar cases, where the “career-making” potential of high-profile individual convictions could create conflicting professional incentives (e.g., Rakoff Reference Rakoff2014). Finally, research could also examine civil enforcement actions or other attorneys in federal agencies where both the presence of large law firms and monetary settlements are especially common.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/jlc.2026.10016.
