Amazon is currently defending itself in a class action alleging racial discrimination. In 2016, the company implemented a stricter background check policy for subcontracted delivery drivers. As a result of the change, several existing drivers—many of whom were black and Latino—became ineligible for continued employment due to criminal offenses that had previously been overlooked. Amazon directed the termination of these drivers despite their satisfactory performance records. Six of these drivers responded by alleging that Amazon’s actions were discriminatory under a disparate impact theory of liability.

To bring a successful disparate impact claim, plaintiffs must show that an employer’s facially neutral policy has an unjustified and disproportionate effect on members of a protected class. In such cases, an absence of discriminatory intent is irrelevant. If the plaintiff can show that the employer’s practice is “fair in form, but discriminatory in operation,” the burden shifts to the employer to prove that the challenged practice is justified by business necessity. In Griggs v. Duke Power Company, the Supreme Court explained that employers must demonstrate that the policy “bear[s] a demonstrable relationship to the successful performance of the job.” Even if the employer meets this burden, the plaintiff can still prevail by showing that the employer refused to adopt a less discriminatory, yet equally effective alternative to the challenged practice.

The EEOC acknowledges that “National data . . . supports a finding that criminal record exclusions have a disparate impact based on race and national origin.” The extent to which racial bias pervades all levels of the criminal justice system has been well documented. As a result, black men are six times more likely to be incarcerated than white men. This general trend persists even after researchers control for background factors, such as parental marriage rates, education, and wealth. For example, black men who grew up in the top 1% of household incomes are incarcerated at the same rate as white men who grew up in families making $36,000 per year.

In light of the presumption that employers’ use of criminal records will produce a disparate impact on racial minorities, the EEOC elaborated on the burden of proof required for a business necessity defense in a 2014 guidance document—a nonbinding, legally persuasive explanation of how the agency interprets its authorizing statutes. As a preliminary, the guidance reaffirmed that an employer may not apply a “blanket policy” of denying employment to any applicant with a criminal conviction. Beyond that, an employer has two options for demonstrating that a criminal conduct screen is sufficiently job-related. First, the employer can produce scientific studies quantifying a relationship between criminal conduct and future performance in a particular position—however, the EEOC recognized that such data is rare. Second, the employer can develop a “targeted screen” based on factors first articulated by the 8th Circuit in Green v. Missouri Pacific Railroad: the nature of the offense, the time elapsed since the offense, and the nature of the position. Furthermore, if such a screen would exclude an applicant, the employer must also provide an opportunity for an individualized assessment.

It seems clear that Amazon won’t be able to satisfy its burden under either standard. Despite claims that the background check process is “focused on job-related criminal and motor vehicle convictions,” the employees appear to have been denied any individualized assessment of their unique circumstances and their demonstrated ability to perform the jobs in question. Even if these particular plaintiffs succeed, the current system of case-by-case litigation and inadequately protective laws isn’t doing enough to protect workers. According to one survey, up to 92% of employers subject some or all job candidates to criminal background checks. If one in three black men born in 2001 is statistically likely to be imprisoned at some point in his life, then a substantial percentage of the workforce that anti-discrimination laws aim to protect face significant and disproportionate burdens to obtaining employment—particularly in positions which are more likely to encourage upward socioeconomic mobility. Therefore, policymakers should consider ways to limit the detrimental effects of past criminal conduct on future employment.

One option would be to reduce employers’ incentive to inquire about applicants’ criminal histories. Studies estimate that over half of employers rely on background checks for the express purpose of avoiding liability for negligent hiring (or negligent retention, as would be the case with Amazon’s delivery drivers). Litigation over such issues often focuses on the adequacy of pre-employment investigations into an employee’s fitness. Policymakers could restrict employers’ overall exposure to liability or limit the admissibility of evidence pertaining to employers’ knowledge of past criminal conduct. However, such a change may not do enough to discourage background checks. After all, preventing lawsuits is only one of several reasons why employers are concerned about employees’ criminal records.

Another option would be to limit employers’ ability to inquire into past criminal conduct at all. Policymakers could preclude any consideration of criminal histories, perhaps with limited exceptions for narrow categories of positions and crimes (e.g., childcare workers and sexual offenses). Critics might challenge this as an incursion into employer autonomy, particularly given the background rule of employment at will. However, as anti-discrimination laws demonstrate, employers’ right to determine the organization of their workforce is not absolute. Employer interests can be outweighed by countervailing interests of social and racial equality. Still, such a severe curtailment of employer rights would be a tough sell.

A possible middle ground could formally require employers to articulate background screening policies based on the Green factors. The court’s intuition that employers can estimate an applicant’s risk level has some support from empirical data. Although research on recidivism at work is lacking, actuarial analyses have quantified the point at which past criminal records cease to have any predictive value on future misconduct generally. However, even assuming employers can accurately (and consistently) adhere to a Green-style balancing test, such policies would still systematically exclude individuals with criminal records from employment for at least some period of time. The Green factors are an imperfect solution to the disparate impact problem, but may be the best way to reconcile the inherent tensions between several conflicting interests. Furthermore, a commitment to rigorous enforcement may also serve a symbolic function by signaling a policy preference for employment equality.

Some might question whether employment law is an appropriate mechanism for addressing disparities caused by criminal law. However, the root issue of structural racism is both systemic and cumulative. In The New Jim Crow, Michelle Alexander invokes Iris Marion Young’s famous birdcage metaphor: “If one thinks about racism by examining only one wire of the cage, or one form of disadvantage, it is difficult to understand how and why the bird is trapped. Only a large number of wires arranged in a specific way, and connected to one another, serve to enclose the bird and to ensure that it cannot escape.” Racial inequality in employment and racial inequality in the criminal justice system are two sides of the same coin. The underlying problem is so ubiquitous that piecemeal solutions to each phenomenon in isolation have proven inadequate. Instead, policymakers should aspire to remedy structural racism through comprehensive reforms within all areas of law, including employment.