Religious Liberty in Prisons under the Religious Land Use and Institutionalized Persons Act following Holt v Hobbs : An Empirical Analysis

Religion in the United States remains a consistent source of conflict not only because of the breadth and depth of personal religious commitment, but also because of guarantees from the United States Constitution. The First Amendment protects religious Free Exercise but also constrains federal, state, and local governments from establishing official government religions, endorsing religions or religion itself. Despite the risk of potential conflicts with the constitution’s text, Congress has supported laws that expand religious liberty. One such example is the Religious Land Use and Institutionalized Persons Act (2000), which significantly enhanced prisoners’ right to religious exercise above the minimum provided by the First Amendment. In the 2015 case of Holt v. Hobbs, the Supreme Court ruled in favor of a Muslim prisoner who had been denied his request for religious accommodations under RLUIPA because the prison failed to satisfy the act’s strict scrutiny standard before it denied accommodations to a prisoner to practice his faith. Via an analysis of case law since Holt v. Hobbs was decided in January 2015 until March 2018, we investigate the extent to which Holt has affected judicial voting in RLUIPA cases and how such voting may have been influenced by judges’ ideological dispositions.


Introduction
It is no secret that the exercise of religious liberty can be divisive. 1Religion remains a perpetual source of conflict not only because of the breadth and depth of people's religious commitments, but also because of the text of the United States Constitution.The First Amendment protects religious Free Exercise 2 but also constrains federal, state and local governments from establishing official government religions, endorsing religions or indeed religion itself. 3Despite the risk of potential conflicts with the constitution's text, Congress has supported laws that expand religious liberty. 4he Religious Land Use and Institutionalized Persons Act ("RLUIPA" or the "Act"), passed by Congress in 2000, 5 is a paradigmatic example of such legislative activity.The law significantly enhanced prisoners' right to religious exercise above the minimum provided by the First Amendment. 6espite a growing body of case law interpreting the Act and commentary on its meaning, 7 empirical research into judicial decision making and RLUIPA have been sparse and rudimentary. 8n 2015 in Holt v. Hobbs 9 the Supreme Court ruled in favor of a Muslim prisoner who had been denied his request for religious accommodations under RLUIPA.This was because the prison failed to satisfy RLUIPA's strict scrutiny standard before it denied accommodations to a prisoner to practice his faith. 10Holt's support for the prisoner's RLUIPA claim was striking, especially when Holt is compared to the only other Supreme Court case interpreting RLUIPA in prison settings, Cutter v. Wilkinson. 11 There, the Cutter Court spoke equivocally about RLUIPA rights.On the one hand Cutter reiterated RLUIPA's strict scrutiny standards but on the other hand simultaneously suggested that a substantial degree of deference is owed to prison officials in applying the strict scrutiny standard to justify burdens on religious exercise.
Oddly, the Holt decision does not even mention Cutter.This might suggest the Holt Court recognized the effect its deference language in Cutter might have in defeating RLUIPA's strict scrutiny standard.Could it be that Holt sent a message of retreat on the deference front?If so, have lower courts received the message and acted upon it as they are obligated to do?We raise this question because sometimes federal judges deviate from what precedent seems to dictate.Indeed, judges frequently vote ideologically when afforded the opportunity to do so. 12hese considerations raise questions about Holt in the context of judges' tendency to vote ideologically. 13Accordingly, in this study we investigate whether Holt has affected judicial voting in RLUIPA cases involving prisoners' requests for religious accommodations and how such voting may have been influenced by judges' ideological dispositions.We deploy a nonparametric matching technique using a dichotomous dependent measure, pro-prisoner plaintiff or pro-defendant, as our dependent measure to assist us in answering these and related questions.Finally, we attempt to draw 3 See, e.g., Engel v. Vitale, 370 U.S. 421 (1962) (holding that a state statute authorizing a short voluntary prayer for recitation at the start of the school day violates the establishment of religious clause of the First Amendment); Lee v. Weisman, 505 U.S. 577 (1992) (including clergy who offer prayers as part of an official public-school graduation ceremony is forbidden by the Establishment Clause); Larkin v. Grendel's Den, Inc., 459 U.S. 116 (1982) (vesting in the governing bodies of churches and schools the power to prevent issuance of liquor licenses for premises located near the church or school violates the Establishment Clause of the First Amendment); but see, Cutter v. Wilkinson, 544 U.S. 709 (2005) (Section 3 of RLUIPA, on its face, qualifies as a permissible accommodation to religious exercise that is not barred by the Establishment Clause); Corporation of Presiding Bishop of Church of Jesus Christ of Latter Day Saints v. Amos, 483 U. S. 327 (1987) (upholding against an Establishment Clause challenge a provision exempting religious organizations from the prohibition against religion-based employment discrimination in Title VII of the Civil Rights Act of 1964).  1(Sisk and Heise 2012) (observing strong ideological voting in Establishment Clause cases). 13Id.
conclusions from the results to learn whether Congress has achieved its goal in enacting RLUIPA of affording institutionalized persons' broad protection for their genuinely felt religious commitments.

The Religious Land Use and Institutionalized Persons Act of 2000 (RLUIPA)
Section 3 of RLUIPA states: No government shall impose a substantial burden on the religious exercise of a person residing in or confined to an institution, ... even if the burden results from a rule of general applicability, unless the government demonstrates that imposition of the burden on that person-(1) is in furtherance of a compelling governmental interest; and (2) is the least restrictive means of furthering that compelling governmental interest. 14LUIPA's protections are enforceable against states and their subdivisions and persons acting under color of state law. 15RLUIPA does not protect the interests of federal prisoners. 16LUIPA provides for a private cause of action for violations of its provisions. 17However, RLUIPA does not create a cause of action for monetary damages against state entities, 18 and states have not consented to waive their sovereign immunity with respect to official capacity monetary damage claims under RLUIPA. 19Moreover, RLUIPA does not impose on state employees individual liability for violations of the Act. 20This is because RLUIPA was adopted pursuant to Congress's Spending Clause powers.As such personal liability may only be imposed on those parties receiving the funds. 21LUIPA provides an alternative ground for invoking judicial intervention through the Commerce Clause.That provision kicks in where the Act's substantial burden requirement affects "commerce with foreign nations, among the several states, or with Indian tribes." 22Plaintiffs may not obtain declaratory or injunctive relief under RLUIPA where an order can have no effect on the defendant's behavior toward him. 23Would-be plaintiffs must exhaust available administrative remedies before seeking relief under RLUIPA. 24Since failure to exhaust is an affirmative defense that must be pleaded, it can be waived. 25ith RLUIPA came a burden-shifting framework for claim resolution. 26The RLUIPA plaintiff's initial burden is two-fold: he or she must show that (1) the relevant religious exercise is "grounded in a sincerely held religious belief" and (2) the government's action or policy "substantially burden[s] that exercise" by, for example, forcing the plaintiff "to 'engage in conduct that seriously violates [his or her] religious beliefs.'" 27Under RLUIPA there is no requirement that plaintiffs' sincerely held religious belief be fundamental to, or a central tenet of, their religion. 28However, an adherent's mere preference for a practice does not establish a substantial burden for the religious exercise. 29Moreover, challenges to the sincerity of religious convictions by motions to dismiss the complaint may not be made until later phases of the litigation; 30 this is because plaintiff's beliefs are largely a credibility determination and sincerity can rarely be determined, even on a motion for summary judgement. 31inally, RLUIPA establishes a subjective test of whether religious beliefs are sincerely felt; thus, they need not be consistent with majority views of the tenets of a particular faith or even if some may find the beliefs illogical. 32If the plaintiff carries this burden, the government bears the burden of proof to 23 2006).First, it "protects administrative agency authority" by allowing the agency to "correct its own mistakes" rather than being immediately "haled into federal court."Id. (internal citations omitted).Second, it promotes efficiency insofar as administrative review processes are generally faster and more economical than is litigation.Id.Even where the parties subsequently seek judicial remedies, the administrative-review process often produces a useful record to ease and expedite further proceedings.(2014)).See, e.g., Washington v. Klemm, 497 F.3d 272, 28 (3d Cir.2007) (for purposes of RLUIPA, substantial burden exists where: a follower is forced to choose between following the precepts of his religion and forfeiting benefits otherwise generally available to other inmates versus abandoning one of the precepts of his religion in order to receive a benefit; or , the government puts substantial pressure on an adherent to substantially modify his behavior and to violate his beliefs); Smith v. Governor for the State of Alabama, 562 Fed.Appx.806 (11th Cir.2014) (granting summary judgment to defendant on ground that plaintiff failed RLUIPA's substantial burden test which "requires at a minimum that a RLUIPA plaintiff demonstrate that government's denial of a particular religious item or observance was more than an inconvenience to one's religious practice."). 28 show that its action or policy (1) is in furtherance of a compelling governmental interest and (2) is the least restrictive means of furthering that interest. 33he Supreme Court has emphasized that "[s]everal provisions of RLUIPA underscore its expansive protection for religious liberty." 34Courts must construe RLUIPA "in favor of a broad protection of religious exercise, to the maximum extent permitted by . . .[RLUIPA's] . . .terms and the Constitution." 35In addition, RLUIPA "may in some circumstances require [a][g] overnment to expend additional funds to accommodate [inmates'] religious beliefs." 36LUIPA defines "'religious exercise' capaciously to include 'any exercise of religion, whether or not compelled by, or central to, a system of religious belief.'" 37Although RLUIPA subjects governmental action to exacting scrutiny, "it also affords prison officials ample ability to maintain security." 38When applying RLUIPA, "courts should not blind themselves to the fact that the analysis is conducted in the prison setting." 39In particular, they must recognize that "[p]rison officials are experts in running prisons and evaluating the likely effects of altering prison rules." 40However, judicial deference is not unyielding; courts are not "bound to defer" to a prison system's assertions. 41"[I]t is the obligation of the courts to consider whether exceptions are required under the test set forth by Congress." 42Thus, while courts "should respect" the prison officials' expertise, they cannot abandon "the responsibility, conferred by Congress, to apply RLUIPA's rigorous standard." 43LUIPA creates certain procedural obstacles to which prisoners must adhere.As mentioned above the Act imposes on prisoners an administrative exhaustion requirement: "No action shall be brought with respect to prison conditions under section 1983 of this title, or any other Federal Law, by a prisoner confined in any jail, prison, or other correctional facility until such administrative remedies as are available are exhausted." 44RLUIPA sets out a four-year statute of limitations for bringing judicial actions for violations under the Act. 45LUIPA and its predecessor statute, the Religious Freedom Restoration Act of 1993 (RFRA), 46 were Congressional reactions to the Supreme Court's decision in Employment Division., Department. of Human Resources of Oregon.v. Smith. 47In this case the Court held that neutral, generally applicable laws that incidentally burden the exercise of religion usually do not violate the Free Exercise Clause of the First Amendment. 48Smith largely overruled the method of analysis used in prior free exercise cases that used a balancing test that considered whether a challenged government action that substantially burdened the exercise of religion was necessary to further a compelling state interest. 49n its place Smith abandoned the strict scrutiny standard and substituted a rational basis test for asserting a Free Exercise claims against the government, thereby narrowing the constitutional ground upon which to assert free exercise infringements against the government.50Thus, when inmates claim that prison policy substantially burdens their Free Exercise rights the state may defeat the claim by showing that its rules are "reasonably related to legitimate penological interests." 51ongress enacted RFRA to overcome the effects of Smith by returning protection for religious exercise to pre-Smith standards. 52In the prison context Congress was concerned that government officials retained too much power in denying capriciously one more liberty to prisoners where they had already been stripped bare. 53In making RFRA applicable to the States and their subdivisions, Congress relied on Section 5 of the Fourteenth Amendment. 54In City of Boerne v. Flores 55 the Supreme Court held that RFRA exceeded Congress' powers under that provision. 56City of Boerne did not, however, address enforcement of RFRA-created rights against the federal government.
In 2006, in Gonzales v. O Centro Esperita Beneficente Uniao do Vegetal, the Supreme Court held unanimously that members of a Christian spiritualist sect could enforce RFRA-based rights against the federal government. 57Since the question of whether RFRA exceeds Establishment Clause limits was not decided in O'Centro or in subsequent cases decided by the Supreme Court, it remains an open question whether RFRA oversteps constitutional limits.Those circuits that have addressed this issue have rejected claims that RFRA violates the Establishment Clause, 58 and considering the Court's unanimous ruling it seems likely that it would not find that RLUIPA exceeded Establishment Clause limits.The importance of RFRA in the present context is that its provisions, not those of RLUIPA apply to prisoners' right to religious accommodations in federal prisons.
In the next section we next discuss our research methodology.This includes an explanation of how we selected cases for our data base, the classification of judicial decisions, coding protocols, and statistical tools applied to answer our research questions.

Case Selection
To isolate effects attributable to Holt v. Hobbs, we employed a nonparametric matching technique to estimate if there was a change in judicial voting after Holt.We selected cases decided at the U.S. Courts of Appeals before and after the Supreme Court's ruling in Holt on 20 January 2015.
This approach enabled us to more plausibly attribute effects to the Holt decision in the specific legal environment surrounding the timing of Holt.We included decisions in the data base that met the following criteria: 1.
The prisoner's claim was brought and decided under RLUIPA.Prisoners often brought claims under the First Amendment's Free Exercise Clause or the Fourteenth Amendment's Due Process Clause.A decision was included only if the text of the decision revealed the case was disposed of under the Act rather than an alternative legal ground.2.
The data was derived from published and indexed/" "unpublished" courts of appeals decisions.Most of Courts of Appeals decisions are decided as "unpublished."This does not mean they are unavailable on various data bases but only that the circuit judges deciding the case concluded the holding will not be considered to have formal precedential value.However, in every other respect these decisions purport to following legal precedent applicable to the case before the court.

3.
The decision was rendered by one of the United States Circuit Courts of Appeals.Thus, United States District Court and U.S. Supreme Court decisions were not included in the data base.4.
Prisoners bringing claims are or were incarcerated in state facilities.This is because RLUIPA only provides for religious accommodations to prisoners housed in state facilities. 5.
The cases were decided during the period following the Holt decision in January 2015.Those cases ranged from June 2012 to March 2018.
Applying these criteria our dataset was constructed from 135 decisions at the Courts of Appeals before and after Holt, for a total of 405 votes. 9Because courts analyze RLUIPA cases under a variety of procedural and substantive lenses we coded the decisions in a way that tracked the process courts followed in deciding them.

Procedural Dismissals
With some frequency RLUIPA claims are dismissed on procedural grounds for such reasons as the statute of limitations having expired or the prisoners' failure to exhaust administrative remedies prior to suing.Other non-merits grounds for dismissal include mootness and lack of standing.We categorized these decisions as non-merits dismissals ("NMD").The idea here was to separate those cases from the ones which ruled on the merits of the RLUIPA claims and addressed core purposes of the Act.Of the 135 decisions 33 were categorized as NMD cases; these comprised 99 of the 405 votes.

Merits Decisions
On the merits side, RLUIPA plaintiffs must initially show that they hold a sincere religious belief and that the government substantially burdened its exercise.Here we distinguished those decisions that held the prisoner's claim failed to allege a substantial burden on religious exercise from those in which the courts found the prisoner did so.Among the 102 decisions that remained after the NMDs were subtracted from 135 decisions . . .47 decisions (comprised of 141 votes) were dismissed for failure to allege or prove adequately a substantial religious burden.This left 55 decisions comprised of 165 votes to apply RLUIPA's compelling interest-least restrictive means test.
This approach tracked the Act's analytic framework since under RLUIPA analysis stops when no burden is found.In other words, there is no reason for the courts to continue with the compelling interest and least restrictive means tests when the prisoner fails to show the government imposed a substantial religious burden on the prisoner.On the other hand, where the court found the plaintiff alleged a substantial burden the court would have to decide whether the state showed it had a compelling interest justifying its actions and had applied the least restrict means in doing so. 59mong the cases where the prisoner made an adequate showing on the religious burden prong 40 decisions went in favor of the prisoner and 15 for the prison.
These numbers reveal that 24.4% of the RLUIPA claims were never reviewed on the merits because of NMDs.Most of these decisions were mootness-dismissals based on the prisoner's transfer to another facility, an outright release, or the provision of an adequate accommodation to the prisoner since the claim was filed.Far fewer of these dismissals were based on failure to exhaust administrative remedies or limitations failures.About 34.8% of all 135 claims filed were dismissed because the prisoner failed to make a prima-facie case showing that exercise of sincerely held religious beliefs was substantially burdened.Of the 102 claims that survived NMD 47 of these, or about 46.1%, were dismissed on "no substantial burden" ground.Thus, out of the original 135 claims filed only 55 (165 votes) proceeded to a compelling interest-least restrictive means analysis.This represents only 41% of the initial 135.Of this group 40 plaintiff prisoners obtained relief from the courts in [state] cases.

Courts of Appeals Voting Patterns
A potential issue in our data analysis was the fact that prior to Holt v. Hobbs Courts of Appeals applied different degrees of deference to prison-defendants' claims that religious accommodations under RLUIPA should not be granted because they would compromise the states' interest in safety or the orderly administration of their oversight function, for example.Some circuits exercised substantial deference to such government claims while others applied the strict scrutiny required by RLUIPA in a more demanding fashion.The latter group took a "hard look" at compelling interest and least restrictive means claims made by prison officials; they examined much more closely the factual assertions made by the government to justify their refusal to accommodate the prisoners' requests for accommodations.Such circuits may be labeled "hard look" jurisdictions. 60n this regard Bollman concluded that the Fifth, Sixth, Eighth, Tenth and Eleventh Circuits were "deferential" circuits and the First, Third, Fourth, Seventh, and Ninth Circuits were hard look circuits. 61e concur with Bollman's contention but conclude the Second Circuit was a hard look jurisdiction pre-Holt and have adjusted our data analysis accordingly.
The voting dispositions and circuit court legal standards, deferential or hard look, applied pre-Holt had the potential to affect our ability to determine whether Holt affected judges' voting.Since Holt was definitive in its insistence that courts take a "hard look" at the reasons given and the availability of less restrictive means to accomplish prisons' policy objectives, Holt should result in less substantial movement in voting between the pre-Holt period and the post-Holt one since those circuits were already voting in a manner seemingly more consistent with what Holt required.On the other-hand the deferential circuits it would seem might move more substantially in a pro-prisoner direction since they had, so to speak, a greater distance to travel to meet the Holt standard.We accounted for the circuit effects in this regard in our statistical analysis as explained more fully below.

Case Locating and Coding
The data are drawn from all cases obtained from a search of the Westlaw and LexisNexis databases for all U.S. Court of Appeals cases meeting the criteria set forth above.We assigned a value of "1" to 59 We recognize that some judges might dispose of cases they were inclined to dismiss anyway by simply finding no substantial burden and thereby avoid the demanding work of analyzing the governmental interest asserted by prison officials and whether they applied the least restrictive means to achieve their policy objectives.Although the high proportion of cases dismissed on no substantial burden ground suggests this could be the case this is far from certain. 60(Bollman 2018) (Applying descriptive statistics to study among other things the effects of Holt v. Hobbs, 135 S.Ct.853 (2015)  on case outcomes at the U.S. District Courts and U.S. Courts of Appeals). 61 votes that were in favor of the plaintiff-prisoner and "0" in favor of the prison-defendant.Votes for cases to be remanded for further proceedings were also coded as "1" since in such instances the prisoner achieved substantial success: the court of appeals required the lower court to apply RLUIPA in a way more favorable to the prisoner than it had when it first confronted the case.As such, of all the votes in our dataset, 137 (33.8%) were classified as pro-prisoner-plaintiff while 268 (66.2%) were classified as pro-prison defendant.

Results
Our main interest is in assessing whether the Holt v. Hobbs case decided at the US Supreme Court at the end of 2015 had any impact on judicial behavior.Of all votes cast, 192 (47%) were cast before Holt and 213 (52.6%) were cast after Holt. Figure 1 shows the distribution of pro-prisoner-plaintiff and pro-prison-defendant votes before and after Holt.As shown, the descriptive statistics show an increase in the percentage of votes in favor of the plaintiff-prisoners, going from around 29% of votes before Holt to 38% of votes after Holt.To isolate the part of these outcomes that can be attributed to changes in judicial behavior, we consider the potential confounding impact of both party affiliation and political ideology.In our dataset, 197 (48.6%) of votes were cast by Democratic Party appointees, while 208 (51.4%) were cast by Republican Party appointees.
In all a higher percentage of Republican votes (36%) were cast in favor of the plaintiff-prisoners compared with Democrats (31%).As shown in Figure 2, this party difference is almost entirely attributable to the post-Holt era.Before Holt, just over 70% of both Republican and Democratic affiliated votes (77 and 59 votes respectively) were in favor of the prison-defendant.After Holt, the percentage of votes in favor of the prison-defendant decreased to 56% of Republican affiliated votes (56 votes) while the Democratic affiliated proportion of votes in favor of the prison only dropped to 67% (76 votes).To isolate the part of these outcomes that can be attributed to changes in judicial behavior, we consider the potential confounding impact of both party affiliation and political ideology.In our dataset, 197 (48.6%) of votes were cast by Democratic Party appointees, while 208 (51.4%) were cast by Republican Party appointees.
In all a higher percentage of Republican votes (36%) were cast in favor of the plaintiff-prisoners compared with Democrats (31%).As shown in Figure 2, this party difference is almost entirely attributable to the post-Holt era.Before Holt, just over 70% of both Republican and Democratic affiliated votes (77 and 59 votes respectively) were in favor of the prison-defendant.After Holt, the percentage of votes in favor of the prison-defendant decreased to 56% of Republican affiliated votes (56 votes) while the Democratic affiliated proportion of votes in favor of the prison only dropped to 67% (76 votes).Party affiliation, while often used as a proxy for political ideology, may be an inadequate device by which to match judges, since it potentially masks significant ideological differences among those of the same party affiliation.This notion is captured by common phrases such as "blue dog" Democrat or "moderate" Republican which denote party members who are respectively somewhat to the right or left of members of their party in their voting records.As such, we also consider a more finely grained measure of ideology based on DW-NOMINATE continuous scores.
In total, there were 196 judges in our dataset, 90 (46%) of whom were appointed by a Democratic Party president, and 106 of whom were appointed by a Republican counterpart.The DW-NOMINATE score ranges from −1 (most liberal) to +1 (most conservative).The average score of Democratic Party appointees was −0.32 and that of Republicans +0.44.The distribution of these scores is shown in Figure 3.62 62 Here, we use the same formula employed in (Sisk and Heise 2012), where judges are assigned the DW-NOMINATE score of the appointing President if there is no Senator of the same party in the judge's state, the score of the Senator if there is one of the same party as the appointing President for the judge's state, or the mean of the scores of both Senators if there are two Senators of the same party as the appointing President for the judge's state.However, whereas Sisk and Heise use "common-space" DW-NOMINATE scores, we use the specific scaling for the Senate, since only Senators are involved in the judicial appointment process.Party affiliation, while often used as a proxy for political ideology, may be an inadequate device by which to match judges, since it potentially masks significant ideological differences among those of same party affiliation.This notion is captured by common phrases such as "blue dog" Democrat "moderate" Republican which denote party members who are respectively somewhat to the right or left of members of their party in their voting records.As such, we also consider a more finely grained measure of ideology based on DW-NOMINATE continuous scores.
In total, there were 196 judges in our dataset, 90 (46%) of whom were appointed by a Democratic Party president, and 106 of whom were appointed by a Republican counterpart.The DW-NOMINATE score ranges from −1 (most liberal) to +1 (most conservative).The average score of Democratic Party appointees was −0.32 and that of Republicans +0.44.The distribution of these scores is shown in Figure 3. 62 Beyond judicial ideology, we also consider the nature of the jurisdiction in which a case was heard.As described earlier, there are strong theoretical grounds to distinguish between "deference" circuits and "hard look" circuits.Consequently, in assessing whether judges changed their voting behavior following the Holt decision, we would ideally want to compare the record of judges from a similar jurisdiction.We coded the Fifth, Sixth, Eighth, Tenth and Eleventh circuits as "deference circuits" and the First, Second, Third, Fourth, Seventh and Ninth circuits as "hard look" circuits.In all, 189 (47%) votes were cast in "deference" circuits, and 216 (53%) were cast in "hard look" circuits.As discussed, there are reasons to expect that judges in deference circuits would be more likely to change their voting propensities following Holt, since the U.S. Supreme Court decision could be interpreted as sending a signal to the lower courts to accord more weight to the religious needs of prisoners.
Finally, we also consider the type of case.Our cases involved the following classifications: whether a non-merits dismissal(NMD)occurred on procedural ground such as statute of limitations or mootness, whether there was a determination of "no substantial burden" that would lead to an automatic victory for the defendant/prison, whether the court found the plaintiff-prisoner stated an adequate claim for a burden on religious exercise that would have proceeded to further analysis, and whether the circuit from which the decision was taken was a deferential or hard look circuit.

Matching Analysis
As explained above, our main purpose in accounting for whether the court found a substantial burden (or in which no such determination was made at all) require more analysis and therefore are more likely to invite differences of opinion.It would not necessarily be meaningful to compare the behavior of two judges, one considering a "slam dunk" case of NMD and the other considering a case that was considerably more complicated.
Our analysis to follow uses a nonparametric matching technique to ascertain if there was a change in judicial behavior pre-and post-Holt.This approach uses the "potential outcome framework" to estimate an Average Treatment Effect (ATE).This approach is perhaps best explained by an illustration.Take, for example, the case of a judge considering an RLUIPA claim.Of interest is Beyond judicial ideology, we also consider the nature of the jurisdiction in which a case was heard.As described earlier, there are strong theoretical grounds to distinguish between "deference" circuits and "hard look" circuits.Consequently, in assessing whether judges changed their voting behavior following the Holt decision, we would ideally want to compare the record of judges from a similar jurisdiction.We coded the Fifth, Sixth, Eighth, Tenth and Eleventh circuits as "deference circuits" and the First, Second, Third, Fourth, Seventh and Ninth circuits as "hard look" circuits.In all, 189 (47%) votes were cast in "deference" circuits, and 216 (53%) were cast in "hard look" circuits.As discussed, there are reasons to expect that judges in deference circuits would be more likely to change their voting propensities following Holt, since the U.S. Supreme Court decision could be interpreted as sending a signal to the lower courts to accord more weight to the religious needs of prisoners.
Finally, we also consider the type of case.Our cases involved the following classifications: whether a non-merits dismissal(NMD)occurred on procedural ground such as statute of limitations or mootness, whether there was a determination of "no substantial burden" that would lead to an automatic victory for the defendant/prison, whether the court found the plaintiff-prisoner stated an adequate claim for a burden on religious exercise that would have proceeded to further analysis, and whether the circuit from which the decision was taken was a deferential or hard look circuit.

Matching Analysis
As explained above, our main purpose in accounting for whether the court found a substantial burden (or in which no such determination was made at all) require more analysis and therefore are more likely to invite differences of opinion.It would not necessarily be meaningful to compare the behavior of two judges, one considering a "slam dunk" case of NMD and the other considering a case that was considerably more complicated.
Our analysis to follow uses a nonparametric matching technique to ascertain if there was a change in judicial behavior pre-and post-Holt.This approach uses the "potential outcome framework" to estimate an Average Treatment Effect (ATE).This approach is perhaps best explained by an illustration.Take, for example, the case of a judge considering an RLUIPA claim.Of interest is whether the same judge would vote differently on a similar case following the Supreme Court's Holt decision.As such the Holt decision is here a kind of "treatment effect."In this framework there are two "outcomes," one observed and one notional based on how the judge "would have" voted if the case had been decided in a different era.The average treatment effect is then the average of the differences between each observed and corresponding notional outcome.
Proceeding to our matching analysis, we were able to specify perfect matching on all the categorical covariates: party affiliation (Democratic or Republican), circuit type ("hard look" or "deference"), and case type (no substantial burden, substantial burden, or non-merits dismissal).This means the votes being compared-aside from occurring before or after the Holt case-are those in which we can assume that the judges involved were both of the same party affiliation, both from either a "hard look" or "deference" circuit, and both considering a case of the same type.For the continuous ideology score, we were also able to obtain near perfect matching.The success in matching judges according to ideology scores is also depicted in the balance plot of Figure 4, which contrasts the closeness of ideology scores for the matched data versus those of the original raw data.These statistics imply that in assessing whether Holt had an impact on individual judges we are essentially comparing judges of the same party affiliation and political ideology on cases of the same type and being decided in similar circuit types.
whether the same judge would vote differently on a similar case following the Supreme Court's Holt decision.As such the Holt decision is here a kind of "treatment effect."In this framework there are two "outcomes," one observed and one notional based on how the judge "would have" voted if the case had been decided in a different era.The average treatment effect is then the average of the differences between each observed and corresponding notional outcome.
Proceeding to our matching analysis, we were able to specify perfect matching on all the categorical covariates: party affiliation (Democratic or Republican), circuit type ("hard look" or "deference"), and case type (no substantial burden, substantial burden, or non-merits dismissal).This means the votes being compared-aside from occurring before or after the Holt case-are those in which we can assume that the judges involved were both of the same party affiliation, both from either a "hard look" or "deference" circuit, and both considering a case of the same type.For the continuous ideology score, we were also able to obtain near perfect matching.The success in matching judges according to ideology scores is also depicted in the balance plot of Figure 4, which contrasts the closeness of ideology scores for the matched data versus those of the original raw data.These statistics imply that in assessing whether Holt had an impact on individual judges we are essentially comparing judges of the same party affiliation and political ideology on cases of the same type and being decided in similar circuit types.
Based on this matching our results provide evidence that judicial voting propensities did indeed change following the Holt decision.The result is statistically significant although the effect size is quite modest.Specifically, we find that the marginal probability, here the average treatment effect or ATE, of a judge voting in favor of the defendant-prison dropped by 13% following Holt (Table 1) We should note here that this result is not evident from the descriptive statistics but is only obtained after close matching of judicial and case attributes when comparing cases before and after Holt.

Discussion
Using a non-parametric matching design and aligning judges with applicable case level [circuittype (deferential or hard look)], judge-level [judges' ideology (party affiliation, DW-nominate scores)] and plaintiffs' religious affiliation into our model, we found a significant difference in voting Based on this matching our results provide evidence that judicial voting propensities did indeed change following the Holt decision.The result is statistically significant although the effect size is quite modest.Specifically, we find that the marginal probability, here the average treatment effect or ATE, of a judge voting in favor of the defendant-prison dropped by 13% following Holt (Table 1) We should note here that this result is not evident from the descriptive statistics but is only obtained after close matching of judicial and case attributes when comparing cases before and after Holt.

Discussion
Using a non-parametric matching design and aligning judges with applicable case level [circuit-type (deferential or hard look)], judge-level [judges' ideology (party affiliation, DW-nominate scores)] and plaintiffs' religious affiliation into our model, we found a significant difference in voting outcomes before and after Holt v. Hobbs, which was decided on 20 January 2015.Judges' pro-prison voting decreased after Holt.We observe however the effect size was a modest one.Specifically, we find that the marginal probability, here the average treatment effect or ATE, of a judge voting in favor of the defendant-prison dropped by 13% following Holt.This result was not evident from the descriptive statistics; we ascertained this change only after close matching of judicial and case attributes when comparing cases before and after Holt.It seems then the Holt precedent matters and to some degree made the rights protected under RLUIPA a bit more secure.
In hindsight prisoners' claims tended to fall into three main types: grooming, ceremonies, or services, and requests for religious materials. 63Among these categories, ceremonial or group services for prisoners would seem to pose (from a prison official's perspective) the most threat to good order; judges might take this same view.Since we did not account for accommodation-type in our study it is potentially a confounding factor that contributed to variance in judges' voting behavior.Future researchers may wish to include this as a variable in modeling judicial voting on RLUIPA claims.
Methodologically, the use of non-parametric matching for our subject [judges] served this study well and may be an improvement on traditional logistic regression modeling for dependent measures which are dichotomous Our matching approach enabled us to more accurately compare how judges voted before and after Holt; it enabled us to make more cogent comparisons between these precedential eras by accurately isolating Holt's effects.Use of this approach may make more accurate the conclusions we draw about why judges behave as they do and should perhaps be employed more often in studying judicial behavior.
We now consider relevant issues as to why we obtained only a modest effect when it might have been reasonable to expect a greater impact of the Holt precedent.The 135 RLUIPA decisions we analyzed produced 408 judicial votes.At first blush it is striking that nearly 24.4% (33) of these claims were dismissed on non-merits ground, for example, filing the claim after the statute of limitations had run, mootness, based on prisoner transfer or release, and other reasons not related to RLUIPA's core purpose of protecting prisoners' sincerely felt religious commitments.The most likely reason for this result is that a substantial majority of these claims were initiated by pro se prisoners who are unschooled in the arcane world of civil procedure.It is hard to attribute these outcomes to anything but poor "lawyering." Why 47 of the remaining 102 claims were dismissed for failure to allege or establish a substantial burden on religious exercise is less clear.The results are perhaps a bit surprising since RLUIPA defines religious exercise broadly as "any exercise of religion, whether or not compelled by, or central to, a system of religious belief." 64Certainly Holt reinforced this Congressional directive by requiring an individualized assessment of the prisoner's subjective sincerity and a determination of whether the prisoner must actually choose between following religious belief or following prison policy. 65hile some portion of these claim failures may be attributable to pro se prisoners' poor drafting skills, this seems an incomplete explanation since trial courts are forgiving in processing pro se complaints which dominate the landscape in RLUIPA proceedings.
It is possible that claims dismissed for lack of substantial burden were simply weak ones where the prisoners assumed that rights created by RLUIPA covered mere religious preferences rather than the substantial burden RLUIPA requires.For example, prisoners would with some frequency admit in the papers they submitted that they could practice their faith without the prison accommodating those preferences.In other words, their requests for accommodations appeared to be only a gloss on their subjective beliefs about what their faith required.In the same vein, prisoners' assertions of what their beliefs required and what prison mandates were unconnected logically.Such claims were virtually guaranteed a dismissal on RLUIPA's substantial burden prong.
We can only speculate on the reasons for the high proportion of NSB dismissals since there could be ideological or other factors at play in explaining why 46% of the surviving claims were defeated on RLUIPA's first prong. 66Our data, however, did not reveal substantial differences in voting patterns between Democratic and Republican appointees or conservative and liberal voters when judges' DW-nominate scores were considered.In any case, even if there had been such differences, our methodology that matches vote comparisons according to both judicial and case attributes gives us a strong basis for asserting that we have successfully isolated the impact of the Holt case alone, without other confounding factors.
Although it is possible that the substantial burden dismissals might be a mere cover for deference to prison officials' claims of security needs or orderly administration of prisoner concerns, our data gives us no reason to suspect this is true.Indeed, the limited research in this area suggests this theory may not be a viable one. 67otably, of the 55 claims which proceeded to a compelling interest-least restrictive means analysis 40 plaintiff-prisoners obtained relief in the Courts of Appeals.This represents about 73% of these cases.Given RLUIPA's burden shifting approach, the direction of the results is not surprising.Recall that RLUIPA requires a compelling governmental justification to say "no" to the prisoners' request for an accommodation and proof that the denial is the least restrictive means of accomplishing that purpose.The legal challenge for the prison-defendants may be daunting once the plaintiff establishes that a substantial burden on his or her religious exercise has occurred, especially in the wake of Holt.
It seems then, on a practical level, prisoners and their advocates would be well advised to devote as much time and effort as may be feasible in drafting complaints which: link their religious beliefs and their implementation, clearly and precisely while being sensitive to the safety, budgetary and staffing issues prison officials typically confront.As to the latter point selecting locations and times where security will be adequate may be helpful in obtaining the requested accommodation.So too, selection of objects for use in religious activities which pose a diminished potential for use as weapons may help advance the strength of the RLUIPA claim.Perhaps those advocates, even those working in a ministerial capacity, could serve as advisors in connection with prisoners' requests for religious

Figure 1 .
Figure 1.Distribution of Votes Before and After Holt v. Hobbs (June 2012 to February 2018).

Figure 1 .
Figure 1.Distribution of Votes Before and After Holt v. Hobbs (June 2012 to February 2018).

Figure 2 .
Figure 2. Distribution of pro-prison votes before and after Holt by party affiliation.

Figure 2 .
Figure 2. Distribution of pro-prison votes before and after Holt by party affiliation.

Figure 4 .
Figure 4. Balance Plot of ideology (DW-NOMINATE) scores for raw and matched data.

Figure 4 .
Figure 4. Balance Plot of ideology (DW-NOMINATE) scores for raw and matched data.
See, Banks v. Secretary, Pennsylvania Dep't of Corrections, 601 Fed Appx.101, **5) (3d Cir.2015) (denying prison-specific injunctive and declaratory relief, holding that claims were moot on transfer because injunctive and declaratory relief for injury at institution where he no longer resided could have no effect on defendant's behavior toward him).24ThePrison Litigation Reform Act (PLRA) provides that "[n]o action shall be brought with respect to prison conditions . . .by a prisoner confined in any . . .prison . . .until such administrative remedies as are available are exhausted."42 U.S.C.

Table 1 .
Estimated change in voting propensities following Holt matching by ideology, party affiliation, circuit type, and case type.