There’s No There There: The Trump Administration’s Use of Misleading Empirical Evidence to End Collective Bargaining for Most Federal Employees

Introduction

President Donald Trump has issued several executive orders affecting the federal workplace. One directive with sweeping impact is Executive Order (EO) 14251, issued on March 27, 2025. The EO excludes federal employees in more than a dozen cabinet-level and other agencies from labor law coverage and collective bargaining agreements (CBAs). Absent the EO, these employees and agencies would be subject to the Federal Service Labor-Management Relations Statute (FSLMRS), a statute passed in 1978 as part of civil service reform legislation. According to the EO, the President’s action was justified by a statutory provision to protect the nation’s national security.

Under 5 U.S.C. § 7103(b)(1), the President can exclude federal agencies from application of the FSLMRS if two things are true: a primary function of the agency involves national security work, and if union representation would not be “consistent with national security requirements and considerations.”

The purpose of this column is to consider faulty empirical evidence that has been offered by the Administration to support the EO, and to offer an alternative perspective on the FSLMRS and arbitration.

The Administration contends that union representation is inconsistent with and undermines national security because CBA terms and arbitration decisions interfere with managerial prerogatives to run government agencies without delay or obstruction. In a memo from the Office of Personnel Management (OPM), also issued on March 27, the Administration transmitted guidance on the EO to affected agencies, advising them, among other actions, to disregard CBA provisions on reductions-in-force, grievance participation, and deduction of union dues.

In addition, in a Fact Sheet issued on March 27, the White House charged that some federal sector unions “have declared war on President Trump’s agenda,” noting that one union has been “widely filing grievances to block Trump policies.” The Fact Sheet contrasts these reasons for the EO with the President’s support for “constructive partnerships with unions who work with him.” Responding to the EO, several unions sued President Trump seeking injunctive relief, alleging that the EO violates the FSLMRS and constitutes retaliation for union activity protected under the First Amendment.

The administration’s opposition to one of the union lawsuits elaborated on its national security rationale:

Employee performance is also critical in agencies with important national security roles. Many provisions in the Defendant agencies’ CBAs make it more difficult to remove employees who perform poorly. For example, CBAs often require “performance improvement periods” (PIPs) of at least 60 days before agencies can propose removing an employee for poor performance…Even after that process, CBAs allow unions to grieve dismissals of poor performances to binding arbitration, with arbitrators overturning approximately three-fifths of removals they hear.

The AFPI Report

In submitting the opposition brief quoted above, the administration cited a research report from the America First Policy Institute (AFPI), Union Arbitrators Overturn Most Federal Employee Dismissals. The sponsor, as the name suggests, is not an academic institution engaged in disinterested social science research, but a policy proponent. According to the AFPI’s mission statement, it:

…exists to advance policies that put the American people first. Our guiding principles are liberty, free enterprise, national greatness, American military superiority, foreign-policy engagement in the American interest, and the primacy of American workers, families, and communities in all we do.

The AFPI’s values aside, it is appropriate to closely examine the report as it is the sole empirical support offered by the Administration for its assertion that the grievance and arbitration process has a workplace impact that undermines national security.

As the AFPI report explains, non-unionized employees may object to dismissals before the Merit Systems Protection Board (MSPB). Unionized employees, if they wish, can opt to arbitrate under a CBA’s grievance procedure, subject to review by the Federal Labor Relations Authority (FLRA), the administrative agency overseeing federal sector labor relations. The legal standard to decide whether dismissal is justified applies equally to the MSPB and in arbitration, and case law has developed to assess individual cases in both forums. The Supreme Court has affirmed this approach.

Using MSPB decisions from 2011 to 2016, and a sample of 435 arbitration awards from 2018 to 2020, the AFPI report finds the following:

The AFPI report attributes the difference in total managerial vindication—72 percent versus 42 percent—to two main causes: the incentive for unethical arbitrator decisions to curry favor with the prevailing party, and deficiencies in the arbitral selection process.

Of the first, the report acknowledges that the tripartite Code of Professional Responsibility for Arbitrators of Labor-Management Disputes prohibits the compromise of a decision, a so-called “split decision,” when “made for the sake of attempting to achieve personal acceptability.” The AFPI report points to the 20 percent rate of split decisions in arbitration as evidence that arbitrators act unethically.

Of the second cause for the differential, the report posits that unions are repeat players selecting arbitrators drawn from a master list maintained by the Federal Mediation and Conciliation Service (FMCS), but an agency “goes to arbitration only a few times a year.” From this premise, the AFPI concludes that unions track arbitrators for their win/loss performance while agencies do not, and that “arbitrators who frequently rule against unions get little federal arbitration work.”

Reasons for Skepticism

Before addressing the AFPI report’s explanation of its findings, there are reasons to be skeptical of its data. It is estimated that there are about 2.3 million full-time federal employees, with over a million—a number that has been increasing—being represented by unions and having access to arbitration. The report notes that, according to OPM records, thousands of federal employees with permanent civil service status are dismissed each year. Only a small fraction of dismissals lead to administrative proceedings. For cases of unacceptable performance, notice and an opportunity to improve typically are required by statute as a pre-condition to discharge, thereby affording employees a means of preserving employment.

The first problem with the AFPI data is that it does not provide the number of MSPB decisions it relies on, only a percentage. We simply cannot tell if the number of cases in the MSPB category is statistically meaningful. A related complication is that we do not know the union status of an employee with an MSPB case, unlike employees with arbitration cases brought by unions.

A second data problem is that the AFPI report’s analysis of dismissals that went to arbitration not only is based on an unstated number of cases, but the volume analyzed—435 over two-plus years—includes disciplinary suspensions along with dismissals, as well as alleged violations of contract terms. This mix of data is a confusing jumble, and not illuminating.

A third data deficiency is that the AFPI report focuses on the government’s supposed inability to remove employees for poor performance as an impediment to protecting national security. But the AFPI data treats all discharges alike without separating discharges based on poor performance. The agglomeration of all discharge cases tells us nothing about whether employees dismissed for poor performance in the federal sector are treated disparately by arbitrators compared to the MSPB, or whether the number of discharges for poor performance—which the AFPI report does not disclose—are too few to be statistically significant.

Last, the only examples of actual cases cited in the AFPI report are dismissals based on employee misconduct; for example, harassment, drug possession, accepting gifts, and patient misconduct. The report does not offer a single instance of a termination for poor performance, much less one adversely affecting national security.

We believe that these methodological shortcomings have led the Administration to make an illogical leap in citing the AFPI report to support the EO.

Assuming arguendo that the AFPI’s raw data prompts a measure of concern, even if inchoate, the relevant question is this: What does actual experience tell us about the role of arbitration and its alleged interference with national security? The final section of this column will directly address this question, but first the report’s explanations are considered.

The Report’s Explanations

There are solid scholarly studies of arbitration, most recently a comprehensive, rigorous, and insightful work drawn from more than 2,000 discipline and discharge decisions filed with a Minnesota labor relations agency. The Minnesota study shows that it is not unusual for the results in discipline cases to vary based on a number of variables. The study found, for example, that arbitrators in Minnesota upheld private sector discharges in full 49 percent of the time, but 56 percent of the time in the public sector. By the AFPI’s reasoning, arbitration must be biased to favor public sector employers over those in the private sector. Q.E.D.

Outcomes also vary by the cause given for discharge. Arbitrators in Minnesota upheld discharge for poor performance about 65 percent of the time, and about 45 percent of the time for on-the-job misconduct. But, as noted, the AFPI report aggregates all federal sector discharges even though the report’s singular concern is performance, not misconduct.

There is no need to unpack all the APFI report’s explanatory failings, even assuming its statistical sample is adequate. Three reasons alone suffice.

First, union represented federal employees who face discharge may pursue either the contractual grievance procedure or an appeal to the MSPB. They cannot do both. As a practical matter, unions commonly retain the power to decide whether to file a formal grievance and control the forum for disposition; that is, whether without settlement, it will take the case to arbitration. In the ordinary course, a union will conduct a preliminary investigation into an employee’s complaint before deciding whether to grieve. If a union believes the employee to be wrong or the case too weak, it is unlikely to grieve. If the union does not file a grievance, the employee has recourse to the MSPB.

Although the AFPI report tells us nothing about the union status of the employees in its sample, it remains likely that employees who are not in a union-represented bargaining unit, and who have recourse to the MSPB, will not have had their case screened by union representatives knowledgeable about the possible outcome. It also is likely that unionized employees appearing before the MSPB without a union representative will have had their cases rejected after screening by their unions, had they sought union assistance. Under these conditions a disparity in result, if there is one, should not be surprising, especially if an employee is appearing pro se without any representation, as the MSPB permits.

Second, is the supposed “repeat player” effect described in the AFPI report and the alleged advantage for arbitrators known to unions. Awards in the federal sector are provided in the comprehensive CyberFEDS database available by subscription to those handling federal cases. A subscriber can learn an arbitrator’s decision history by the click of a mouse. Yet the report would have the reader believe that federal agencies have been too ignorant, indifferent, or lethargic to inquire.

CyberFEDS access aside, the AFPI report errs in its description of the role of the FMCS. True, the FMCS maintains an arbitrator roster with hundreds of names, and it can provide a list of several names from the roster for parties to consider in a single case. However, many if not most agencies and their union counterparts agree on a standing panel of arbitrators to hear all of their cases with the results known to both parties. For example, the Internal Revenue Service and the National Treasury Employees Union have a well-developed contractual system for panels of arbitrators for national and local disputes. In this system, both sides are repeat players. The reference to “union arbitrators,” as in the title of the AFPI report, is a misnomer.

Third, the AFPI report is misleading in describing the significance of what it refers to as split decisions. The MSPB is permitted to render partial decisions with neither side prevailing completely. As noted above, the AFPI’s reported numbers conflate MSPB split decisions with those MSPB decisions fully sustaining an employee appeal. The AFPI report attributes the prevalence of split decisions, of whatever number, to arbitral corruption akin to self-dealing, even as the report, by aggregating the data, declines to inform the reader of how often the MSPB makes split decisions.

Contrary to the charge made by the AFPI, there is nothing inherently nefarious or corrupt in a split decision. It is unethical only when compromise is made, “…for the sake of attempting to achieve personal acceptability.” As the Minnesota study recognizes, split decisions are ethical when “an arbitrator finds the employee engaged in inappropriate conduct, but also concludes that the employer’s discipline was excessive.” In such a case, rather than discharge, the arbitrator can award a suspension with back pay from the end of the suspension to reinstatement, or can award reinstatement with no back pay at all. As these researchers find,

In a close case requiring significant discipline but not the ultimate sanction of discharge, the reinstatement without back pay directive permits the arbitrator to assign the economic burden to the party whose misconduct triggered the need for discipline in the first place—the employee.

The emphasized clause reminds us that arbitration is the end product of a multi-step grievance procedure in which the union and agency representatives have jointly examined the facts and issues. It is not surprising that those disputes most intractable of settlement—“close cases”—are those most likely to be arbitrated and to result in split decisions.

In sum, whatever its rhetorical value, the AFPI report is not social science.

Federal Sector Labor Law and National Security

Rather than rely on findings and explanations that are not supported by data, we can better understand the impact of arbitration on national security from the text of the FSLMRS and from cases applying it. The EO’s abrogation of collective bargaining and grievance arbitration is predicated on the claim that, in an agency with a primary function in national security, even if not the dominant function, the obligation to bargain collectively with unions imposes inflexibility that threatens national security, irrespective of the work these employees do. However, by operation of law, the FSLMRS in 5 U.S.C. § 7112(b)(6) already prohibits a bargaining unit from including “any employee engaged in intelligence, counterintelligence, investigative, or security work which directly affects national security.” The FLRA has provided extensive interpretation of the meaning of subsection (b)(6)’s exemptions.

An early FLRA decision, Oak Ridge, determined that the statutory exclusion from union representation should apply narrowly to work that “directly affects national security,” citing as key concerns “espionage, sabotage, subversion, foreign aggression, and any other illegal acts which adversely affect the national defense.” In Oak Ridge, the FLRA relied on Cole v. Young, a 1956 decision of the Supreme Court rejecting summary discipline based on a national security claim. For the Supreme Court, an “indefinite and virtually unlimited” understanding of national security would create an exception to general personnel laws that “could be utilized effectively to supersede those laws.” The same reasoning can apply to the EO. Oak Ridge remains controlling precedent under the FSLMRS.

An employee doing any national security work in the departments exempted by the EO—Veterans Affairs, for example—is already precluded from collective bargaining. Building on the express exclusion of a handful of agencies in the FSLMRS, presidents beginning with Jimmy Carter have excluded a small fraction of the workforce in several agencies and departments based on national security concerns, but, as observed in one of the EO injunction decisions, no president, until now, has sought to remove most federal employees from the FSLMRS.

So understood, the EO’s reliance on the national security exclusion in 5 U.S.C. § 7103(b)(1) is based on the blanket assumption that because a primary function of a portion of these units is devoted to national security, the exacting, job-by-job bargaining unit determinations required by 5 U.S.C. § 7112(b)(6) need not be undertaken for the agency as a whole. This reflects neither a fair reading of the text of the statute, nor practice.

Beyond unit determinations, employees without national security functions in these agencies can be exempted from collective bargaining pursuant to 5 U.S.C.§ 7103(b)(1) if, but only if, collective bargaining for them would be inconsistent with national security. This is the second part of the statutory test established by Congress. The Administration claims that this condition is satisfied because agency management is hindered by alleged bargaining delays and contract inflexibility leading to performance impediments that are enforced in arbitration. The argument does not withstand scrutiny; the logical leap too great.

Under 5 U.S.C. § 7122(a)(1), the FSLMRS permits either party in an arbitration to challenge an award on the ground that it is “contrary to any law, rule, or regulation.” This provision includes any law, rule, or regulation dealing with national security functions. Given this limitation, the FSLMRS itself has an appellate mechanism that prevents any arbitration award from interfering with national security.

Moreover, actual data demonstrates that national security is a subject rarely touched upon in federal sector arbitration. Using the advanced search function of the FLRA’s comprehensive compilation of all awards appealed under 5 U.S.C. § 7122 from 1980 through 2025, the search identified 2,324 challenges. Of these, a search for keywords related to the FSLMRS’s national security exemption found only five awards in the past 45 years that have possible relevance; that is, 0.00215 percent of all awards.

From the FLRA decision search, three awards deal with “security clearances:” one involving the loss of clearance for a temporary guide; another for denial of sick leave after loss of security clearance; and, a third about barring union communication with an employee during a security clearance investigation. A search for awards involving “internal security practices” found two: one affirming the right of management to assign an employee displaced after loss of a security clearance; and, a second about designating employees for random drug testing based on internal security practice. This sparse arbitral record should not be surprising. Under 5 U.S.C. § 7121(c)(3), the FSLMRS expressly excludes from arbitration the dismissal or suspension of an employee covered by 5 U.S.C. § 7532; that is, an employee subject to management’s disciplinary authority “in the interests of national security.”

When the empirical evidence is properly considered, the public record demonstrates that arbitration under CBAs neither interferes with nor constrains agency action based on national security. The EO cannot be supported by relying on faulty empirical evidence.