The Pitfalls and Opportunities of Algorithmic Thinking in the Federal Regulatory Process: A Case Study

0
1596

On September 27, 2019, the Department of Labor[1] published its final rule expanding overtime protections for executive, administrative, and professional employees—the culmination of a five year struggle that straddled two administrations and ultimately left 2.8 million workers without the effective pay increase they had initially been promised.[2] It is accurate, though an oversimplification, to describe this controversial rulemaking as turning on the choice between two sub-algorithms. This rulemaking demonstrates how agencies use algorithmic thinking to interpret statutes, and why that approach, without better data analytics, is susceptible to misuse. It also reveals the inefficiencies and inequities in the regulatory process that could be addressed by a targeted application of algorithmic thinking.

Defining and Delimiting Statutory Text with Algorithms

The Overtime Rules were promulgated under an express direction from Congress that DOL “define[] and delimit[]”[3]the scope of an exemption to the Fair Labor Standards Act’s[4] minimum wage and overtime protections. Congress intended to except white collar workers under the theory that because these workers enjoyed better salaries, job security, and opportunities for advancement, they had no need for worker protections. To interpret the meaning of the statutory phrase “any employee employed in a bona fide executive, administrative, or professional capacity,”[5] DOL has generally used the following algorithm, though the agency describes it as a series of tests each of which must be met:

if an employee is paid a fixed salary (the “salary basis test”); and
if an employee’s salary is above a certain level (the “salary level test”); and
if an employee’s duties meet the regulatory definition (the “duties test”); and
then she is a bona fide executive, administrative, or professional worker who is exempt from the FLSA’s minimum wage and overtime protections.[6]

The salary level test and the duties test each consist of sub-algorithms; for decades, these tests were considered together on a sliding scale, under which a more rigorous duties test was paired with a lower salary level threshold, and a less stringent duties test with a higher salary level.[7] The agency, therefore, has traditionally relied on a series of nested conditionals to precisely sort the bona fide white collar employees that Congress intended to exempt from the FLSA’s protections.

The Timing of Regulatory Updates and Its Effect on Stakeholders

The Department issued its first regulations interpreting the FLSA’s exemptions in 1938, and then updated them to account for changing economic conditions in 1940, 1949, 1954, 1958, 1961, 1963, 1967, 1970, 1973, and 1975.[8] Due to shifting political priorities and the considerable expense of promulgating regulations, the Department did not again revise its interpretation until 2004, at which point the salary level the agency had been using to identify “white collar” workers was $8,060 per year—well below the 2004 minimum wage.[9] The Department began its next revision in 2014, resulting in the 2016 Final Rule, under which over 4 million American workers would have been newly eligible for overtime starting on December 1, 2016. After the rule was enjoined shortly before the busy holiday season on November 22, 2016, these workers and their employers, many of whom had been preparing for this change for months, were thrown into limbo. Moreover, the unpredictability of litigation as well as a change in administrations in January 2017 created additional uncertainty and confusion among those most directly affected by the Rule.

A Better Approach to Applying Algorithmic Thinking in the Regulatory Process

Algorithms in software undergo rigorous testing and can be written to be adaptive if certain conditions are met. By contrast, in the regulatory context, they are subject to notice-and-comment rulemaking and are updated with no predictable regularity. Although political objectives are inherent to executive branch functions, automatically generated, regular reports on regulatory effects could achieve two goals: (1) provide objective, publicly available indicators for when a rule needs updating; and (2) impose guidelines for agency discretion, limiting possible changes to a range of acceptable outcomes determined by data analytics. The 2016 Final Rule proposed automatically recalculating the salary level – using the same algorithm adopted in the Rule – with current economic data every three years.[10] Software could further automate and optimize this mechanism. And comprehensive data analysis – in publicly available reports – could keep the agency and stakeholders informed about changing economic conditions and the ongoing relative efficacy of a regulation. Thus, data analytics – not politics – would determine when a model is in need of an update. Using technology to promote more open and informed, and less politicized, decision-making would also enable workers, businesses, and advocates to adapt to regulatory changes and shifting economic conditions more effectively. Of course, no software exists that does not reflect the biases of its creators, so these enhanced tools would need to be developed in a non-partisan way. And unfortunately, political systems do not generally adapt at the same rate at which technology advances

[1] (“DOL” or “the Department”).

[2] I played a small part in this process, as the lead attorney at the U.S. Department of Justice who defended the 2016 overtime rule in a lawsuit brought in the U.S. District Court for the Eastern District of Texas by the state of Nevada and twenty other states challenging this rule under the Administrative Procedure Act and the Tenth Amendment to the U.S. Constitution. See Nevada v. Dep’t of Labor, 218 F. Supp. 3d 520, 534 (E.D. Tex. 2016). I do not discuss the substance of the litigation in this article. Rather, my focus is on the rule and its authorizing statute, as well as its regulatory history, as described in the preamble to the 2016 and 2019 Final Rules.  See Defining and Delimiting the Exemptions for Executive, Administrative, Professional, Outside Sales and Computer Employees, 81 Fed. Reg. 32,391 (May 23, 2016) (“2016 Final Rule”); Defining and Delimiting the Exemptions for Executive, Administrative, Professional, Outside Sales and Computer Employees, 84 Fed. Reg. 51,230 (Sept. 27, 2019) (“2019 Final Rule”) (collectively “the Overtime Rules”).

[3] 29 U.S.C. § 213(a)(1).

[4] The Fair Labor Standards Act of 1938, 52 Stat. 1060, 29 U.S.C. § 201 et seq. (1940 ed.) (“FLSA”).

[5] 29 U.S.C. § 213(a)(1).

[6] See 84 Fed. Reg. at 51,230.

[7] 81 Fed. Reg. at 32,392.

[8] 84 Fed. Reg. at 51,232.

[9] Defining and Delimiting the Exemptions for Executive, Administrative, Professional, Outside Sales and Computer Employees, 69 Fed. Reg. 22,122 (Apr. 23, 2004).

[10] 81 Fed. Reg. at 32,430.

 

This piece was contributed as part of the 2019 Harvard Legal Technology Symposium organized by the Harvard Law & Technology Society. The Symposium was the largest student organized legal technology event in the world. It brought together an interdisciplinary and international community to think deeply about how technology can improve and shape the law.