The National Labor Relations Board (NLRB or Board) has long been criticized for failing to consider empirical evidence when making decisions with broad policy implications.^1×1. See, e.g., Sue J. Henry, Introduction: A Journey into the Future — The Role of Empirical Evidence in Developing Labor Law, 1981 U. Ill. L. Rev. 1, 1 (“For decades, students of labor law have recognized that the law is premised on a series of behavioral assumptions . . . that in the vast majority of instances . . . are wholly unproven and applied without regard to whether they comport with reality . . . .”); Phoebe Taurick, Untested Assumptions in NLRB Proceedings, 27 ABA J. Lab. & Emp. L. 307, 307 (2012) (observing that “the Board’s proceedings usually lack the presentation and consideration of empirical evidence,” relying instead “on [the Board’s] assumption about how people will act or feel in certain circumstances”). One such decision was the NLRB’s 2004 ruling in Brown University^2×2. 342 N.L.R.B. 483 (2004). that graduate student assistants were not employees under section 2(3) of the National Labor Relations Act^3×3. 29 U.S.C. § 152(3) (2012). (NLRA or Act), thus denying student assistants the protections of the NLRA and the permission to bargain collectively.^4×4. Brown Univ., 342 N.L.R.B. at 490. Recently, in Trustees of Columbia University,^5×5. 364 N.L.R.B. No. 90 (2016). the Board overturned this precedent, holding that student assistants working at private colleges and universities are statutory employees covered by the Act.^6×6. Id. at 13. Notably, Columbia University explicitly disavowed Brown University’s failure to weigh empirical evidence^7×7. Id. at 7. and underscored its own efforts to do so.^8×8. Id. at 8–9. Although the NLRB should be applauded for its discussion of empirical evidence in Columbia University, its use of empirical evidence in this decision and others raises significant risks that, until acknowledged, undermine the Board’s credibility.

On December 17, 2014, Graduate Workers of Columbia-GWC, UAW filed a petition seeking to form a union representing all student employees serving as instructors and research assistants at Columbia University (Columbia or University).^9×9. Columbia Univ., Case No. 02-RC-143012, at 1 (Nat’l Labor Relations Bd. Feb. 6, 2015) (order dismissing petition). On January 12, 2015, the NLRB regional office issued an Order to Show Cause, asking why the petition at issue should not be administratively dismissed without a hearing based on the precedential authority of Brown University.^10×10. Id. Petitioners argued that Brown University should be overruled and that, regardless, several facts distinguished the instant case.^11×11. Petitioner’s Response to Order to Show Cause, Columbia Univ., Case No. 02-RC-143012, at 4–6 (Nat’l Labor Relations Bd. Jan. 20, 2015). In particular, petitioners noted the following differences between Columbia University and Brown University: Columbia student employees generate greater sums of income for the University, id. at 4–5; Columbia student employees teach independently and sometimes in situations that bear little on their own courses of study or financial aid packages, id. at 5; and Columbia itself regards the work of its student employees as an apprenticeship, id. at 5–6. The NLRB regional office, relying on Brown University, administratively dismissed the petition on February 6, 2015.^12×12. Columbia Univ., Case No. 02-RC-143012, at 2–3 (Nat’l Labor Relations Bd. Feb. 6, 2015). The Graduate Workers of Columbia petitioned for review of the dismissal.

On March 13, 2015, the NLRB granted the petition for review and remanded the case for a hearing.^13×13. The Trs. of Columbia Univ., Case No. 02-RC-143012, at 1 (Nat’l Labor Relations Bd. Oct. 30, 2015). Because Brown University held that graduate assistants — a term that encompasses individuals that the petitioner sought to represent — are not “employees” within the meaning of section 2(3) of the NLRA, the NLRB regional office concluded that Brown University’s broad holding constrained its decision.^14×14. Id. at 4. Therefore, the regional office dismissed the petition.^15×15. Id. Petitioners then appealed to the NLRB.

Reviewing the decision of the regional office, a panel of the Board^16×16. The panel majority consisted of Chairman Mark Gaston Pearce and Members Kent Hirozawa and Lauren McFerran. reversed and remanded the dismissal, holding that student assistants with a common law employment relationship to a university are covered by the NLRA as employees.^17×17. Columbia Univ., 364 N.L.R.B. No. 90, at 2. The panel considered decades of Board precedent^18×18. When it first addressed the issue in 1972, the NLRB had declined to extend the title of statutory employee to graduate student assistants, finding that they were “primarily students” and that the university did not exercise the requisite control over their work. The Leland Stanford Junior Univ., 214 N.L.R.B. 621, 623 (1974). In 2000, the Board held that graduate student assistants were statutory employees. N.Y. Univ., 332 N.L.R.B. 1205, 1205 (2000). on the issue of collective bargaining at private universities and overruled Brown University.^19×19. The Board agreed with the decision in New York University, the 2000 NLRB decision that Brown University overruled. Columbia Univ., 364 N.L.R.B. No. 90, at 4. First, the panel held that the Brown University Board erred in its statutory interpretation of section 2(3) of the Act because student assistants are common law employees and because the Act does not require that an employment relationship must be primary over any other relationship that exists between the employer and the employee.^20×20. Id. at 5. Therefore, any exception to coverage under the NLRA would require “compelling statutory and policy considerations.”^21×21. Id. at 6. Second, the panel concluded that statutory coverage of student assistants would promote, rather than upset, the goals of federal labor policy.^22×22. Id. at 6–7. Rejecting the Brown University Board’s idea that the panel has discretion to decide whether to extend collective bargaining rights or not, the Board found that the separate-but-overlapping concerns of education and employment do not preclude application of the NLRA.^23×23. Id. at 7. Moreover, the panel held that student assistants’ collective bargaining would not infringe upon First Amendment academic freedom.^24×24. Id. at 7–8.

Next, the panel turned to empirical evidence and its own experience to support its jurisdiction over student assistants. With little by way of direct historical grounds on which to decide whether the Act could be applied appropriately to student assistants in a private university setting, the panel used the comparable experiences of students collectively bargaining at public universities and of faculty collectively bargaining at private universities as empirical and experiential evidence in favor of asserting jurisdiction in this case.^25×25. Id. at 8. The panel cited amicus briefs, academic studies, and current collective bargaining agreements at both public universities and at New York University, a private university that continued to voluntarily recognize a graduate assistants union even after the Brown University decision.^26×26. Id. at 8–12; see also, e.g., id. at 9 (describing one amicus brief detailing several universities’ experiences with graduate student unions). The panel also cited survey-based research on faculty-student relationships at public universities and on academic freedom.^27×27. Id. at 9. Further, it addressed the evidence that Columbia and supporting amici raised to support Brown University’s policy, namely, individual examples showing that collective bargaining detracted from universities’ educational goals.^28×28. Id. at 10. Ultimately, the panel concluded that the evidence supporting a continuation of Brown University amounted to little more than speculation.^29×29. Id. at 12–13. The panel “put suppositions aside . . . and . . . instead carefully considered the text of the Act as interpreted by the Supreme Court, the Act’s clearly stated policies, the experience of the Board, and the relevant empirical evidence drawn from collective bargaining in the university setting” to reach its holding.^30×30. Id. at 12.

The panel concluded that, in Columbia’s case, both instructors and student research assistants were statutory employees because the University controlled their work, which was performed in exchange for compensation.^31×31. Id. at 14–18. With regard to student research assistants, the Board first overruled Leland Stanford, which, like Brown University, relied in part on the idea that an employment relationship must be primary over a student relationship; the Board then found that research assistants were common law employees because their funding, regardless of source, was a form of compensation for work that was university-controlled (compensation and control are common law indicators of employment relevant to employment status under the NLRA). Id. at 16–18. Having established jurisdiction over all of the petitioned-for student assistant classifications, the panel concluded that the petitioned-for bargaining unit was appropriate because, although its members were in various educational programs, they shared a “community of interest.”^32×32. Id. at 18; see also id. at 18–20. In finding that students shared a community of interest, and therefore were an appropriate bargaining unit, id. at 20, the Board considered, for instance, the fact that the students were all part of the same system designed to meet the University’s teaching needs, and that all took directions from the University, id. at 19. Although master’s and undergraduate students spend a relatively short time as research assistants as compared to doctoral students, the panel found that they still were part of the community of interest and therefore did not need to be excluded from the bargaining unit as temporary employees.^33×33. Id. at 20–21. Finally, the panel remanded the issue of the voter eligibility formula to the NLRB’s regional office because it did not have sufficient information about what an appropriate formula would be in this case.^34×34. Id. at 21–22. The Board remanded for consideration of whether employees on intermittent semester appointments should still be eligible to vote based on their “continuing interest in the unit.” Id. at 21.

Member Philip Miscimarra dissented.^35×35. Id. at 22 (Member Miscimarra, dissenting). After first agreeing with the Board’s reasoning in Brown University, Miscimarra then explained his concerns with his colleagues’ reasoning in the instant case.^36×36. Id. at 25–33. Miscimarra emphasized the large financial investment associated with a university education as the primary reason that a university cannot be characterized as a traditional workplace.^37×37. Id. at 23. In his view, the “economic weapons”^38×38. Id. at 29. like strikes and lockouts available to students and universities as a result of the majority’s decision would fundamentally change the “college experience.”^39×39. Id. at 29–30. Moreover, Miscimarra argued that the NLRB’s processes are not suited to application in the university setting.^40×40. Id. at 31–32. For example, Board proceedings involve significant time.^41×41. Id. at 31 (noting that unfair labor practices cases may take from three to ten years, or longer). By the time a decision is made, the student assistant involved may no longer be a student, or the academic department involved may cease to exist.^42×42. Id. at 31–32. Finally, he took issue with the specifics of the petitioned-for bargaining unit, noting that the dissimilarities between various types of student assistants would likely cause the unit to fail any community-of-interest test.^43×43. Id. at 32–33.

Columbia University is a significant decision in terms of the potential substantive effects it will have on student employment at private universities across the country. At the same time, it elevates the use of empirical evidence in the NLRB’s decisionmaking process: First, a primary reason that the Columbia University Board overruled Brown University, which spoke almost directly to the question at issue, was that the Brown University Board did not support its decision with empirical evidence.^44×44. Id. at 3 (majority opinion) (“Disavowing the need for empirical analysis, the Brown University majority instead relied on what it perceived to be a fundamental tenet of the Act . . . .”). The majority also cited scholarship questioning Brown University’s refusal to consider empirical evidence. See id. at 4 & n.28 (first citing Michael C. Harper, Judicial Control of the National Labor Relations Board’s Lawmaking in the Age of Chevron and Brand X, 89 B.U. L. Rev. 189, 222 (2009); then citing Catherine L. Fisk & Deborah C. Malamud, The NLRB in Administrative Law Exile: Problems with Its Structure and Functions and Suggestions for Reform, 58 Duke L.J. 2013, 2076–77 (2009)). Second, the Board here explicitly stated its disagreement with the Brown University Board’s “view that ‘empirical evidence’ is irrelevant to the inquiry.”^45×45. Id. at 7 (citing Brown Univ., 342 N.L.R.B. 483, 492–93 (2004)). Third, the Board devoted almost a quarter of its decision — and half of its discussion section — to an analysis of empirical and experiential evidence that supported the Board’s jurisdiction over student assistants.^46×46. See id. at 8–13. Yet upon closer examination, the Board’s discussion failed to address a common pitfall of empirical evidence: cherry-picking. The Board relied on a variety of external sources without providing information as to its own methodology — a common failing of NLRB decisions, regardless of administration or the Board’s political bent. Therefore, while Columbia University’s recognition of empirics is a step in the right direction, the NLRB should directly acknowledge the limitations in its methodology, lest it appear as though it is attempting to promote biased claims under a neutral guise.

Columbia University was a response to Brown University, which overtly rejected the need to consider empirical evidence, or any evidence at all. The majority in Brown University defended its decision to disregard empirical evidence by stating that its inquiry required only statutory interpretation and nothing else.^47×47. See Brown Univ., 342 N.L.R.B. at 491–93. Moreover, the Brown University Board privileged its own “25 years of untroubled experience under pre-NYU standards” above the scholarly studies that the dissent cited.^48×48. Id. at 493. According to Professors Catherine Fisk and Deborah Malamud, Brown University “offered no empirical support, and instead simply reiterated its arguments from definition.”^49×49. Fisk & Malamud, supra note 44, at 2076. Likewise, Professor Michael Harper suggests that Brown University was on shaky ground because of its “refus[al] to even consider available evidence relevant to how coverage or exclusion [of graduate student assistants] would actually serve or disserve [the NLRA’s] goals.”^50×50. Harper, supra note 44, at 218. Thus, Harper argues that Brown University might not withstand arbitrariness review.^51×51. Id. at 219. These scholarly criticisms do not take place in a vacuum; rather, they speak to a larger concern about the lack of empirical evidence in NLRB decisions generally.

Although the Columbia University Board criticized Brown University’s claims as “almost entirely theoretical,”^52×52. Columbia Univ., 364 N.L.R.B No. 90, at 7. it did not mention that Brown University’s failure to offer empirical evidence — even though the relevant assertions were “empirically testable”^53×53. Id. (quoting Fisk & Malamud, supra note 44, at 2077). — was less the exception and more the rule in NLRB decisions.^54×54. See Fisk & Malamud, supra note 44, at 2050–51 (“The Department of Labor has the expertise and facilities to produce high-quality empirical analyses of the myriad questions that arise in NLRB cases, but it lacks the jurisdiction to intervene (formally or informally) in NLRB debates. The NLRB has the jurisdiction, but, as already noted, Congress denied it the authority and the resources to engage in independent empirical analyses (as a result of which the Board lacks the internal expertise to evaluate empirical analyses offered by interested parties).” Id. at 2051.). Fisk and Malamud outline the NLRB’s limits as a policymaking agency, using a diverse set of cases to illustrate that the NLRB’s rejection of empirical evidence is not limited to the student-employee context or to the current Board. Because the Board operates like a court, it “us[es] rights rhetoric as a way to mask what would otherwise be its obligation to seek out (let alone generate) empirical assessments of the effects of its policies.”^55×55. Id. at 2057. These criticisms may have struck a chord with the Board. Two years after Fisk and Malamud published their recommendations, the Board explicitly invoked empirical evidence in Lamons Gasket.^56×56. 357 N.L.R.B. 739, 742 (2011).

Columbia University appears to be part of a changing tide, in line with Lamons Gasket. Columbia University cited recent data showing that more than 64,000 graduate student employees are organized at twenty-eight state universities.^57×57. Columbia Univ., 364 N.L.R.B. No. 90, at 9 (pointing to universities in California, Florida, Illinois, Iowa, Massachusetts, Michigan, Oregon, Pennsylvania, Washington, and Wisconsin). Looking at both the union contracts at these state institutions and at survey-based research, the Columbia University Board concluded that there was no empirical evidence showing that collective bargaining harmed either mentoring relationships or academic freedom.^58×58. Id. To the Board, it may have been intuitive that systematic data was more reliable than the anecdotes that Columbia referenced.^59×59. See id. at 9–10. However, the Board did not explicitly state why it found the studies to be more convincing or how it chose to weigh various forms of evidence. Because the Board did not explain its methodology, it missed an opportunity to legitimize its choices and instead invited counterevidence, which will arguably exist with regard to any difficult question.

The dissents in both Lamons Gasket and Columbia University raised pertinent critiques on this very point, signaling the need for changes in the NLRB’s treatment of empirical evidence. The Lamons Gasket dissent characterized the majority decision as “a purely ideological policy choice, lacking any real empirical support and uninformed by agency expertise.”^60×60. 357 N.L.R.B. at 748 (Member Hayes, dissenting). The dissent went on to reveal the empirical evidence’s inadequacies, along with a different interpretation of the same data.^61×61. See id. at 750–52. Likewise, the Columbia University dissent raised a valid concern with the majority’s “selective attachment of significance to the examples of peaceful negotiations involving student assistants . . . and with their summary discounting of examples that go the other way.”^62×62. Columbia Univ., 364 N.L.R.B. No. 90, at 28 n.35 (Member Miscimarra, dissenting). While the dissents in both cases offered perhaps unnecessarily harsh criticisms, they aptly pointed to cherry-picking as a potential shortcoming of the NLRB’s use of empirical evidence in its decisions. The Board should address this criticism directly in future decisions.

In the present case, for example, the majority referred to an American Federation of Teachers amicus brief noting that collective bargaining agreements at the University of Illinois, Michigan State University, and Wayne State University gave management specific rights regarding educational elements of courses as well as graduate students’ degree-related performance.^63×63. Id. at 9 (majority opinion). The majority used these agreements as evidence that “parties can and successfully have navigated delicate topics near the intersection of the university’s dual role as educator and employer.”^64×64. Id. On the other hand, the majority characterized Columbia and supporting amici’s “few individual examples” of labor disputes as “a fact of economic life.”^65×65. Id. at 10. These examples in juxtaposition rightfully drew the attention of the dissent,^66×66. Id. at 28 n.35 (Member Miscimarra, dissenting). because, without additional information about the way in which the majority weighed evidence, it seems arbitrary to elevate some examples to empirical evidence and dismiss others as background noise. The majority could have legitimized its choice and addressed the seeming arbitrariness of its consideration of evidence by discussing its methodology and limitations directly.

Ultimately, the NLRB is unlikely, at least in the short term, to develop the kind of robust system of rigorous collection and evaluation of data that would alleviate all methodological concerns. Nevertheless, the NLRB’s commitment to using empirical evidence here is laudable. In order to honor that commitment going forward, however, the NLRB should state its methodology and acknowledge its limitations. In advocating for a more robust decisionmaking process, Columbia University takes a clear stance on the importance of evidence that could alter the framework for Board decisions. Irrespective of changes in administration and composition of the Board, which have profound effects on the substance of highly political decisions,^67×67. See Michael H. LeRoy, The Formation and Administration of Labor Policy by the NLRB: Evidence from Economic and ULP Strike Rulings, 22 J. Lab. Res. 723, 735 (2001) (“[O]n the dimension of policy formation, the Board seems sensitive to its external environment.”). the NLRB’s adjudicatory reasoning has been largely consistent.^68×68. See id. (“The emerging empirical picture of Board decision making shows that . . . the Board administers the NLRA in a routine and remarkably consistent manner.”); see also Fisk & Malamud, supra note 44, at 2019 (describing, across the Boards appointed by Presidents Bush and Clinton, “a formalistic style of adjudicatory reasoning that packages questions of policy as questions of law, and, in so doing, deems social science data and analysis (what might be characterized as ‘legislative facts’) irrelevant to Board policymaking”). Therefore, a procedural change has the potential to last for longer than a single political cycle. This advantage makes it all the more critical that the Board begin stating its methodology and limitations. This simple but important step will allow future Boards to continue discussing empirical evidence while mitigating some of the practice’s harshest criticisms.