JUDICIAL BEHAVIOR ON IMMIGRATION COURTS: INDEPENDENT OR POLITICALLY RESPONSIVE?

Within modern U.S. immigration courts, immigration judges are tasked with adjudicating hundreds of thousands of noncitizens’ removal cases each year. In these cases, immigration judges make decisions on charged noncitizens’ removability, their applications for relief from removal like asylum claims, their requests for adjustment of immigration status, and similar immigration matters (Chand and Schreckhise Reference Chand and Schreckhise2020; Law Reference Law2010). The stakes of immigration judges’ work are incredibly high: “off the charts—the highest possible” (Martin Reference Martin1983, 190). Every individual case decided by an immigration judge holds the potential to irrevocably alter the “lives of a noncitizen and their loved ones,” while the combined effect of immigration judges’ decisions “reshapes the composition of U.S. society” (Jain Reference Jain2019, 264). The U.S. immigration judge position is arguably more important and salient today than ever before. There has been a stunning rise in the number of immigration judge-ordered deportations in the last two decades—a number equal to those ordered during the full century before (Asad Reference Asad2019). With unprecedented national political attention on immigration law and policy and a lack of movement on comprehensive immigration reform, presidents and attorneys general have looked for openings to exert pressure on immigration judges to issue more decisions that fall in line with their immigration agenda—in other words, to generate politically responsive immigration judge behavior.

Immigration judges’ unique roles as “bureaucrats in robes” (Jain Reference Jain2019) and “judges-as-bureaucrats” (Miller, Keith, and Holmes Reference Miller, Keith and Holmes2014, 54) add a layer of complexity to understanding their decision-making behavior and whether political efforts to influence their decisions might succeed. As “judges,” immigration judges are expected to make decisions based on the law and facts in the cases before them (Chand Reference Chand2019; Chand and Schreckhise Reference Chand and Schreckhise2020). Federal law indicates that immigration judges’ decision-making should be guided by “independent judgment and discretion” (8 CFR §1003.10(b)). Similarly, immigration judges are directed to “play the traditional role of passive arbiter” or neutral decision maker in the cases before them (Goldschmidt et al. Reference Goldschmidt, Mahoney, Solomon and Green1998, 3). In the course of their discretionary decisions, immigration judges’ backgrounds and values are likely to guide how they interpret the case facts and law and affect what decisions they reach in a case.

However, immigration judges are also “bureaucrats,” where “efficient and uniform implementation of public policy” are prioritized over due process and individualized justice (Lens Reference Lens2012, 289). Immigration judges are employed by the DOJ, and that places them within a broader administrative agency framework below the U.S. Attorney General and the president. In their positions, immigration judges work alongside the government’s immigration prosecuting attorneys, with the presidential administration serving as the boss of both (Alexander Reference Alexander2006). This creates a tension with the ideal of judicial independence: “Article III [federal] judges are free from most, although admittedly not all, pressure that can be exerted by the political branches. Yet, administrative judges—serving within the executive branch—enjoy no such autonomy” (Chand Reference Chand2019, 397). This potentially opens up immigration judges to politically motivated attempts to manage and control their behavior in ways that do not happen with many judges who are not simultaneously judges and bureaucrats (Seabrook, Wilk, and Lamb Reference Seabrook, Wilk and Lamb2013; Wolfe Reference Wolfe2002).

Like other judges faced with judging in a complex environment, immigration judges are not unconstrained; rather, they must be strategic when assessing when and how to be mindful of their competing interests—independence versus potential political constraints—when making decisions. We have long known, for example, that judges and justices in the United States “prefer Court opinions and legal rules that reflect their policy preferences” (Maltzman, Spriggs, and Wahlbeck Reference Maltzman, Spriggs and Wahlbeck2000, 17) and that their career backgrounds and identity affect their legal worldview and decision-making (e.g., Bowie and Songer Reference Bowie and Songer2009; Bowie, Songer, and Szmer Reference Bowie, Songer and Szmer2014; Boyd, Epstein, and Martin Reference Boyd, Epstein and Martin2010; Epstein, Landes, and Posner Reference Epstein, Landes and Posner2013; Glynn and Sen Reference Glynn and Sen2015; Harris and Sen Reference Harris and Sen2019; Hinkle Reference Hinkle2015; Law Reference Law2005; Mak and Sidman Reference Mak and Sidman2020; Nelson, Hazelton, and Hinkle Reference Nelson, Hazelton and Hinkle2022; Williams and Law Reference Williams and Law2012; Zorn and Bowie Reference Zorn and Bowie2010). Immigration judges, like other U.S. judges, will be motivated to accomplish their goals—ideological or other—while on the bench. However, immigration judges are also likely to be mindful of the context in which they operate, realizing that “their ability to achieve their goals depends on a consideration of the preferences of the other actors, the choices they expect others to make, and the institutional context in which they act” (Epstein and Knight Reference Epstein and Knight1998, 10). Strategic immigration judges will “seek to maximize the impact of their decisions by working within political constraints, strategically anticipating reactions to their decisions by others in the policy process” (Schreckhise, Chand, and Lovrich Reference Schreckhise, Chand and Lovrich2018, 126–7)—even if this means compromising on their preferences to “consider a wider array of concerns than merely those presented in the cases in front of them” (Chand and Schreckhise Reference Chand and Schreckhise2020, 179).

While the potential for political factors to affect immigration judging is real—especially, because immigration judges’ independence is more precarious than federal judges or even other agency adjudicators like administrative law judges (Chand Reference Chand2019; Taratoot and Howard Reference Taratoot and Howard2011)—there has historically been little operational reason for immigration judges to feel politically constrained. The “perception of the political environment” determines a great deal about the ways in which bureaucrats are checked (Jowell Reference Jowell1975, 197), and for immigration judges, the unique bureaucratic environment that they serve in helps to explain why political control over immigration judges has been traditionally weak. In effective principal–agent relationships, principals (like presidents) use tools such as monitoring, careful selection, and the threat of sanctions to help ensure agents (like immigration judges) produce desirable outcomes (Brehm and Gates Reference Brehm and Gates1994; Moe Reference Moe1984; Randazzo and Waterman Reference Randazzo and Waterman2014; Randazzo, Waterman, and Fine Reference Randazzo, Waterman and Fine2006). Absent these control mechanisms, agents are free to behave in idiosyncratic, self-serving ways (Brehm and Gates Reference Brehm and Gates1997; Lewis Reference Lewis2008; Moe Reference Moe1984).

Non-appointed, lower-level (a.k.a. “street-level”) bureaucrats of this nature are frequently held up as examples of difficult to monitor and control agents who are often unresponsive to and even, at times, hostile toward their principals (Brehm and Gates Reference Brehm and Gates1997; Krause and O’Connell Reference Krause and O’Connell2019; Lipsky Reference Lipsky1980; Miller and Whitford Reference Miller and Whitford2016; Mummolo Reference Mummolo2018). The reasons for this are plentiful. Lower-level, career bureaucrats “do not have the same perspective as their political superiors” (Lewis Reference Lewis2008, 31) and lack the presidential loyalty and willingness to “comply with administrative policy objectives” that high-level bureaucratic appointees hold (Krause and O’Connell Reference Krause and O’Connell2019, 530). Principals of lower-level bureaucrats also often do not hold credible threats of punishment for ill-behaving agents (Brehm and Gates Reference Brehm and Gates1997; Mummolo Reference Mummolo2018).

As is the case with supervisors of other lower-level bureaucrats, the president and his administration at a baseline lack powerful tools of control over immigration judge outputs. Under ordinary circumstances, presidents are unlikely to closely monitor immigration judges since doing so in a high caseload setting like immigration courts is expensive and inefficient (Kiewiet and McCubbins Reference Kiewiet and McCubbins1991). Principal selection and removal powers are also depoliticized for immigration judges. The U.S. Attorney General delegates vast selection responsibility to the DOJ’s Executive Office for Immigration Review (EOIR). The EOIR follows a standardized bureaucratic advertisement, application, interview, and background check process that reduces politicization and increases competence (Hausman et al. Reference Hausman, Ho, Krass and McDonough2022; Krause and O’Connell Reference Krause and O’Connell2019; Lewis Reference Lewis2008; U.S. Department of Justice 2022b). Also limiting selection power is that immigration judgeships are classified as “Schedule A career positions, not political appointments” in the federal bureaucracy (U.S. Department of Justice 2008, 115). Unlike appointed bureaucratic positions, immigration judges serve across political administrations and vacancies only arise sporadically. While the threat of firing will increase the likelihood of agents behaving in their principals’ interests (Kiewiet and McCubbins Reference Kiewiet and McCubbins1991), politically motivated firings of immigration judges—as civil servants—have been difficult to achieve (Barnett et al. Reference Barnett, Reddick, Cornett and Wheeler2018; U.S. Department of Justice 2008). This combination of impractical monitoring, weak selection and removal powers, and encouraged judicial independence means that presidents are not likely able to count on immigration judges as reliable, responsive political agents in their day-to-day judicial behavior. Most prior empirical studies have confirmed this, finding that immigration judges do not consistently adhere to their appointing presidential administration’s policy goals in their decisions (Chand, Schreckhise, and Bowers Reference Chand, Schreckhise and Bowers2017; Hausman et al. Reference Hausman, Ho, Krass and McDonough2022; Kim and Semet Reference Kim and Semet2020; Miller, Keith, and Holmes Reference Miller, Keith and Holmes2014; Ramji-Nogales, Schoenholtz, and Schrag Reference Ramji-Nogales, Schoenholtz and Schrag2007; Ryo Reference Ryo2016; United States Government Accountability Office 2008; 2016).

With little incentive to be politically responsive in their behavior, strategic-minded immigration judges are free to use their judicial discretion to make decisions based on their own preferences, traits, and experiences—and that’s exactly what the prior immigration judging literature has found. Liberal leaning immigration judges are more likely to rule in favor of the noncitizen than conservative judges (Keith, Holmes, and Miller Reference Keith, Holmes and Miller2013; Miller, Keith, and Holmes Reference Miller, Keith and Holmes2014; Stobb, Miller, and Kennedy Reference Stobb, Miller and Kennedy2023). The same is true for female immigration judges (Beougher Reference Beougher2016; Chand, Schreckhise, and Bowers Reference Chand, Schreckhise and Bowers2017; Keith, Holmes, and Miller Reference Keith, Holmes and Miller2013; Ramji-Nogales, Schoenholtz, and Schrag Reference Ramji-Nogales, Schoenholtz and Schrag2007; United States Government Accountability Office 2008; 2016) and immigration judges with prior experience working for nongovernmental organizations focused on indigent legal aid (Kim and Semet Reference Kim and Semet2020; Ramji-Nogales, Schoenholtz, and Schrag Reference Ramji-Nogales, Schoenholtz and Schrag2007). And while not studied in prior empirical immigration court research, we would expect a similar effect to be present for many Latinx judges (Achury et al. Reference Achury, Casellas, Hofer and Ward2023; Hofer and Casellas Reference Hofer and Casellas2020). By contrast, other immigration judges are more likely to issue removals or deny asylum compared to their colleagues, including immigration judges with prior career experience working in immigration enforcement (e.g., for the Department of Homeland Security [DHS] or Immigration and Naturalization Service [INS]), employed as a prosecutor, or serving in the military (Miller, Keith, and Holmes Reference Miller, Keith and Holmes2014; Ramji-Nogales, Schoenholtz, and Schrag Reference Ramji-Nogales, Schoenholtz and Schrag2007). In short, the identity of the immigration judge hearing the case, including his or her characteristics and background, can often be the most important factor affecting whether the noncitizen receives a positive outcome (Alexander Reference Alexander2006).

THE TRUMP QUOTAS AS A POLITICAL INTERVENTION

Although immigration judges, in practice, have not been effectively constrained by presidential politics, such a political constraint may be possible under the right conditions. In other contexts, evidence confirms that principals who intervene with high-profile reforms have been able to overcome their otherwise inadequate mechanisms for monitoring and constraining agents. For example, while police officers are a classic example of difficult-to-control street-level bureaucrats (Brehm and Gates Reference Brehm and Gates1997), Mummolo (Reference Mummolo2018) finds that a high-profile NYPD reform was able to generate responsiveness among officers. When the NYPD mandated in 2013 that officers engaging in controversial stop and frisk actions must “provide thorough, narrative descriptions to superiors justifying the reasons for stops of criminal suspects” (2), officers began to worry about supervisor scrutiny of their work. As a result, officers altered their behavior, significantly reducing questionable stops. Might a high-stakes presidential intervention—where political oversight is specific and discipline for failure to comply is daunting—result in more political responsiveness among strategic immigration judges along the lines observed by Mummolo (Reference Mummolo2018) in the policing context? President Trump would put this to the test in 2018.

In June 2017, the U.S. Government Accountability Office (GAO), an independent legislative-branch agency tasked with monitoring the federal government’s performance, issued a report on the case backlog in immigration courts (United States Government Accountability Office 2017). Immigration cases pending per year had skyrocketed in the prior decade, rising from under two hundred thousand per year in 2008 to well over four hundred thousand per year in 2015 (and continuing to rise thereafter) (U.S. Department of Justice 2021). The GAO’s report concluded with recommendations for executive branch action to help address the backlog including instituting immigration judge performance measures with case completion goals (United States Government Accountability Office 2017).

The Trump administration was quick to seize on the GAO’s invitation to intervene in the immigration courts’ backlog crisis. In an aggressive policy that went into effect in October 2018, Trump’s Attorney General Jeff Sessions introduced the “EOIR Performance Plan” which focused on performance goals for immigration judges. The plan required that individual immigration judges complete a minimum of seven hundred cases per fiscal year and that no more than 15% of their cases could be overturned on appeal (Sessions Reference Sessions2018). According to the performance plan, an immigration judge who does not meet these performance quotas will be rated as someone who “needs improvement” or is “unsatisfactory” in their civil service evaluations.Footnote ¹

In public comments and memos, Trump administration officials defended the introduction of the quotas as needed to address the backlog crisis and provide more expeditious operation of our immigration courts (McHenry Reference McHenry2018; Sessions Reference Sessions2017). However, by most metrics, the Trump administration’s “EOIR Performance Plan” was more than just an effort to carry out the nonpartisan GAO’s call to reduce case backlog. It was a politically motivated intervention aimed at achieving immigration outcomes favorable to the Trump administration. Among the quota policy’s critics’ arguments were:

• • “[B]acklog is being used as a political tool to advance the current law enforcement policies” (Long Reference Long2018).

• • The policy would serve as a “death knell for judicial independence” for immigration judges (Sacchetti Reference Sacchetti2017).

• • “Sessions is treating them [immigration judges] like immigration officers, not judges” (Benner Reference Benner2018).

The political intentions of the administration’s immigration judge-directed policy were made all the more credible by Trump’s success in destabilizing the position of careerists throughout the federal bureaucracy (e.g., Doherty, Lewis, and Limbocker Reference Doherty, Lewis and Limbocker2019; Lewis Reference Lewis2019; Moynihan and Roberts Reference Moynihan and Roberts2021) and his administration’s open flaunting of “its desire to impose political loyalty over administrative expertise” (Chand Reference Chand2019, 395).

The executive branch’s direct efforts to generate political responsiveness among immigration judges through the quota policy stand in stark contrast to past efforts. As we describe above, the president typically holds only weak influence options over immigration judges, resulting in little incentive for compliance among these agents. Prior DOJ efforts to push for increases in immigration judge productivity lacked the “teeth” necessary to elicit mass immigration judge responsiveness. For example, in prior years, the DOJ had initiatives focused on nudging immigration judge case processing speed via sending emails to all immigration judges that praised judges for their efficient case processing and listed individual immigration judges by name and case completion rate (Jain Reference Jain2019). The DOJ had also previously issued case completion goals for immigration judges but, contrary to the Trump policy of 2018, these were framed as aspirational rather than mandatory, and there were no disciplinary actions attached for immigration judge failure to meet them (Slavin and Marks Reference Slavin, Marks, Kanstroom and Brinton Lykes2015).

With its 2018 policy, it seemed like the Trump administration had taken a page straight out of Lipsky’s guidebook for exerting control over street-level bureaucrats. Lipsky (Reference Lipsky1980) had long before argued that clear goals and well-developed performance measures for agents can increase “managers’ ability to exercise control over policy” (40). Unlike any immigration judge-focused policy of the past, with the quota policy’s detailed performance metrics, noncompliant judges face the new and potent threat of poor marks on their civil service evaluations, which could be used to trigger at-will firings and/or block career advancement goals. The threat of political constraint on immigration judge behavior that had sat largely fallow for so long was now active. As the New York Times’ Editorial Board observed: “…Mr. Trump came along and reminded everyone just how much power the head of the executive branch has when it comes to immigration [courts and judges]” (Editorial Board Reference Board2021). Strategy-minded immigration judges would have to adjust to this new world of active political constraint. In particular, for those immigration judges not already behaving in ways that were pleasing to the Trump administration or who felt targeted by the new policy,Footnote ² continuing to follow their sincere preferences in their judging behavior could risk their jobs.

LOOKING FOR IMMIGRATION JUDGE RESPONSIVENESS IN DECISIONS

It seemed clear, particularly in light of the Trump administration’s many other aggressive tactics and rhetoric in the immigration law and policy arena, that the administration’s quotas were aimed at producing a high quantity of efficient removals (Trovall, Ortiz, and Prendergast Reference Trovall, Ortiz and Prendergast2018). From his campaign through his presidency, Trump had “placed the deportation of ‘illegals’ at the center of his policy agenda, staking much of his political future on the ability to remove these individuals from the country” (Kim Reference Kim2018, 3). The quotas were likely to aid the issuance of these desired removal orders. A focus on speedy case resolution “obstructs the noncitizen’s ability to present his or her case or obtain counsel” and “compromises the [immigration judge’s] ability to engage in an accurate assessment of the facts at issue,” meaning that immigration judges may deny noncitizens relief “notwithstanding their legal eligibility” (Kim Reference Kim2018, 48). Surveyed federal agency judges have similarly argued that political pressure to speed case decisions can “impede decision making” and “thwart due process” (Chand Reference Chand2019, 405) in a way that prioritizes bureaucratic consistency while sacrificing individualized justice. Given the Trump quota policy’s focus, we expect responsive, strategy-minded immigration judges to prioritize greater efficiency and higher removal rates in their decision-making. We look to two areas of immigration court outcomes for producing this potential responsiveness in the post-quota era: in absentia removal orders and pro-government removal orders on the merits.

For in absentia removal orders, current immigration law states that an immigrant who does not attend a scheduled hearing in the immigration court “shall be ordered deported in absentia” (Immigration Act of 1990, §242B(c)(1)). This legal provision means, in effect, that if at any time, an immigrant fails to appear at a hearing before the immigration court as scheduled, she may be ordered, at the discretion of the immigration judge presiding over the case, removed from the United States because of her absence without the opportunity or extended time in future hearings to defend her case on the merits (Eagly and Shafer Reference Eagly and Shafer2020; Gomez Reference Gomez1993). Due to the innate efficiency of in absentia removal orders relative to allowing cases to proceed on the merits, the quota policy and the administration’s messaging surrounding it implicitly encourage immigration judges to increase their issuance of in absentia removal orders. For example, Attorney General Sessions urged immigration judges that “We have to be very productive. Volume is critical.” and “The timely and efficient conclusion of cases serves the national interest. Unwarranted delays and delayed decision making do not” (Benner Reference Benner2018; Sessions Reference Sessions2017). In addition to their efficiency, in absentia removal orders are more likely than other outcomes to be protected from being overturned on appeal since direct appeals of the orders are not permitted and a very narrow set of criteria must be met for immigration judges to rescind the orders (Boyd et al. Reference Boyd, Carlos, Taylor, Baker and Blasingame2023; Eagly and Shafer Reference Eagly and Shafer2020; Koh Reference Koh2017). Given the combination of their efficiency benefits and their greater protection from reversal on appeal, we expect that immigration judges will increase their rate of in absentia removal orders (especially those whose previous behavior was at odds with the Trump administration’s preferences).

Beyond in absentia removal orders, the implementation of the quota policy may also have driven those immigration judges who were more likely to find in favor of noncitizens pre-policy to strategically alter their merits decision-making in a pro-government, anti-immigrant direction. Immigration judges seeking to be responsive to the president saw, just as other onlookers did, that the Trump administration’s quota policy focused on “maximizing the number of deportations” (Kim Reference Kim2018, 49). Sessions had explicitly emphasized to immigration judges, for example, the need for them to help “to enshrine what the law contemplates and what the people desire—an end to unlawfulness in our immigration system” (Sessions Reference Sessions2017). As a former immigration judge put it, “Evaluating somebody’s performance on the number of cases they close is obviously going to have some effect on the substance of the decisions …You know the boss wants removal orders, not grants” (Topan Reference Topan2018). A pragmatic explanation also helps us to understand an increase in compliance in merits-level decision-making: during the Trump administration, removal orders would be perceived as being more likely to survive appellate review before the Attorney General and Board of Immigration Appeals—where Trump administration-selected judges held a supermajority of positions (Misra Reference Misra2020a; Reference Misra2020b)—than would granted applications. Staying under the new policy’s 15% reversal threshold may thus require immigration judges who were previously empathetic in their judging behavior toward noncitizens to shift in anticipation of appellate review.

As we document above, prior research indicates that in the pre-quota era, immigration judges held vast discretion and were free to be guided by sincere personal preferences or other individual idiosyncrasies (like gender, ethnicity, and career experience) in their decision-making. Under these circumstances, some immigration judges’ baseline attitudes toward immigration, along with their prior experiences, made them more or less likely to rule (on average) in pro-removal ways that mirrored the Trump administration’s policy goals. We expect that judges whose pre-quota decision-making already aligned with the Trump administration’s goals (such as Republican judges) will exhibit a smaller “response” to the quota policy. Immigration judges whose sincere preferences diverge from those of the Trump administration (such as Democrats, women, and Latinx judges) will have to strategically alter their behavior following the introduction of the 2018 quota policy, in order to avoid potential career sanctions.

DATA AND METHODS

Our empirical analysis of the effects of Trump’s policy requires data tracking individual immigration judge decision-making in immigration court hearings both before and after the implementation of the October 2018 policy. The EOIR has made these data available to the public through FOIA (U.S. Department of Justice 2022a), and they record fine-grained information on every immigration court case, including the immigration judge’s name, along with details on the noncitizen participant, outcomes, and other case information. Our study focuses on removal proceedings for which an immigration judge held a hearing (for our in absentia analysis) or made a substantive decision (for our merits analysis) between January 1, 2012 to March 1, 2020.Footnote ³ To credibly identify the causal effect of the quota policy and ensure that any observed change in immigration court decision-making trends is not driven by a change in the composition of immigration judges over time, we restrict our sample to the 335 judges who were actively hearing cases both the year before and after the policy change.

We have two outcomes of interest: In Absentia Removal Order, which measures whether the immigrant is ordered removed in absentia, and Merits Removal Order, which captures whether the immigrant lost her case on the merits and the government won. For In Absentia Removal Order, our unit of analysis is the immigration court hearing, which accounts for the potential that an in absentia removal order can be issued at any scheduled case hearing where the noncitizen does not appear. For Merits Removal Order, our unit of analysis is the immigration court case. Each dependent variable is dichotomous, coded as 1 if the noncitizen is ordered removed and 0 otherwise. Our treatment variable, Post Policy, is also dichotomous, measured as 1 if the immigration judge’s decision occurs after the October 1, 2018 implementation of the quota policy, and 0 if it occurs before.

Our modeling also includes judge and case-level factors deemed relevant to immigration decision-making by prior research. As our above-discussion highlights, since the degree of responsiveness required of immigration judges may depend on their baseline attitudes toward immigration, it is also important for us to include the political identities of the judges in our data. We measure the political partisanship of the immigration judges in our dataset (Judge Party) based on their party registration and/or primary ballot information found in statewide voter registration databases. This judge partisanship measurement technique, which has been used in prior smaller-scale studies of bureaucratic judges (Seabrook, Wilk, and Lamb Reference Seabrook, Wilk and Lamb2013; Taratoot Reference Taratoot2014; Taratoot and Howard Reference Taratoot and Howard2011), is specific to the judge rather than being dependent on the appointing president or attorney general. Following the lead of scholars using commercial data sources on voters and public officials (e.g., Chyn and Haggag Reference Chyn and Haggag2019; Einstein, Ornstein, and Palmer Reference Einstein, Ornstein and Palmer2022; Enamorado, Fifield, and Imai Reference Enamorado, Fifield and Imai2019; Fraga Reference Fraga2015; Hersh and Ghitza Reference Hersh and Ghitza2018; Velez and Newman Reference Velez and Newman2019; Yoder Reference Yoder2020), we use voter registration information developed and made available commercially through L2, Inc. Within our coded data, 30% of judges are Republicans, 51% are Democrats, 10% are independent or non-partisan, and 9% could not be located. The Supplementary Material provides additional information on our partisanship measurement technique and its desirability over alternative options.

We also include variables related to key immigration judge background and demographic characteristics that, as described above, have been found in prior literature to affect discretionary immigration judicial behavior. Three of our judge background variables involve prior work experience for the U.S. government and are likely predictive of pro-government immigration judging behavior. DHS, INS, or EOIR Experience captures whether the immigration judge had previous experience working for an immigration enforcement agency, such as INS, DHS, or the EOIR. Prosecutor/Government Experience measures instances where the immigration judge previously worked for the government or served as a state or federal prosecutor. Military Service codes immigration judges with some prior service in the U.S. military. Legal Aid Experience, which captures prior experience working for organizations providing legal aid to the indigent population including legal aid societies and public defense work, Latinx Judge, and Judge Gender are likely to be related to a judge’s baseline propensity to be more pro-noncitizen in their behavior.Footnote ⁴

We also measure a large number of case-specific characteristics that are likely to influence judicial decisions, including the nationality of the noncitizen, language spoken, and whether the noncitizen has legal representation, is applying for asylum, or is in custody at the time of the hearing. Further details on variable measurement and coding are provided in the Supplementary Material.

To estimate the effect of the quota policy on immigration judge behavior, we employ an interrupted time series approach (e.g., Mummolo Reference Mummolo2018), comparing the rate at which immigration judges remove noncitizens before and after the policy on October 1, 2018. In order to ensure that any increases in removal decisions are not driven by a change in the composition of immigration judges over time or the type of cases heard, we pre-process the data using exact matching (Ho et al. Reference Ho, Imai, King and Stuart2011). For every judicial decision in the post-policy period, we identify the set of decisions in the pre-policy period heard by the same judge, in the same location, and with the same set of observable case-level characteristics. This creates a matched dataset in which every observed covariate likely to influence judicial decision-making is perfectly balanced between treatment (post-policy) and control (pre-policy) groups. Further details on our matching procedure are provided in the Supplementary Material.

For each of our outcomes of interest, we model the probability that an immigration judge will order an immigrant removed (p) with the following logistic regression:

$$ \begin{array}{rll}\mathrm{logit}(p)=\alpha +{\beta}_1x+{\beta}_2J+{\beta}_3(x\times J)+\gamma Z+\varepsilon, & & \end{array} $$

where x is a vector of treatment indicators, J is a matrix of judge-level characteristics, and Z is a matrix of case-level covariates. Standard errors ε are clustered by match strata (i.e., the group of observations heard by the same judge, in the same city, with the same set of observed case-level characteristics). Interactions between the treatment vector and each judge-level characteristic (x × J) allow us to estimate conditional average treatment effects. A principal advantage of the matched dataset is that our estimate of the average treatment effect on the treated (ATT) is not sensitive to the precise functional form of our regression model (Boyd, Epstein, and Martin Reference Boyd, Epstein and Martin2010; Ho et al. Reference Ho, Imai, King and Stuart2007; Smith Reference Smith1997). Full regression tables with alternative specifications are included in the Supplementary Material.