Tax Notes logo

Predicting Worker Classification in the Gig Economy

Posted on Dec. 20, 2021

Benjamin Alarie is the Osler Chair in Business Law at the University of Toronto and the CEO of Blue J Legal Inc. Kathrin Gardhouse is a legal research associate at Blue J Legal.

In this article, Alarie and Gardhouse examine the classification of workers in the gig economy and use machine-learning models to evaluate the legal factors that determine their categorization as employees or independent contractors for federal income tax purposes.

Copyright 2021 Benjamin Alarie and Kathrin Gardhouse.
All rights reserved.

As of 2015, 16 percent of workers in the United States considered themselves to be independent contractors or self-employed rather than employees.1 Given the tax and other advantages that can arise from not classifying workers as employees, it is likely that a significant proportion of those workers were not correctly categorized according to the applicable legal tests. Indeed, it is likely that many of those workers would be treated as employees if their statuses were formally contested. In a hand-collected data set of worker classification decisions in the tax context from 1927 to 2021, we found that more than half of the cases — 50.7 percent — were resolved in favor of finding workers to be employees.2

The issues involving worker classification go back decades. Given the rise of the gig economy, these issues have resurfaced with urgency, with Uber and other gig economy service providers making headlines for the past several years for allegedly misclassifying workers. The debate about whether Uber food delivery and ride-share workers are independent contractors (as Uber would have it) has thus far been concentrated in the employment law context. But the federal tax implications of worker classification are equally important. This article examines the worker classification issue in the gig economy in the realm of U.S. federal tax law.

Rev. Rul. 87-41, 1987-1 C.B. 296, sets out a 20-factor test that the IRS uses for worker classification. Each factor is associated with one of three categories of consideration: (1) behavioral control, (2) financial control, or (3) type of relationship. In any given set of real-world facts and circumstances, some factors may clearly weigh in favor of characterization as an employee or independent contractor, while other factors may not be so obvious in their influence. Moreover, it is difficult to gauge the relative importance of each of the 20 factors. In general, the factors are permitted to vary in importance based on all the surrounding circumstances. If a case goes to court, it is vital for the parties to know which factors to focus on to develop the most effective litigation strategy regarding the evidence and possible concessions as to various facts. Likewise, if a client requires advice on how to structure a contractual relationship with workers without risking unexpected tax liability, a tax professional should be able to point with confidence to the most relevant considerations and be able to provide an opinion about the most likely characterization for tax purposes.

Machine-learning techniques can help tax practitioners identify previously decided cases with similar facts and circumstances and, based on an analysis of all the case law, predict the outcome of a case if it were to go to court. Also, machine-learning tools permit tax practitioners to perform sensitivity analysis and scenario testing by changing each of the 20 IRS factors individually — recalculating the confidence in the predicted outcome each time. This scenario testing allows for an accurate estimate of the relevance of each factor based on a given set of circumstances. One such machine-learning tool, the Blue J Tax “worker classification” module, leverages a machine-learning model based on federal tax cases involving worker classification decided between 1927 and 2021. Over the entire database of decisions, the Blue J Tax worker classification machine-learning model has a 97 percent agreement rate with the courts in mapping facts and circumstances to court-determined worker classification outcomes.

In previous installments of Blue J Predicts, we examined the strengths and weaknesses of ongoing or recently decided tax cases, provided machine-learning-generated insights concerning key factors driving the courts’ decisions, and predicted how outcomes could change depending on competing findings of fact. In this month’s column, we look at employment cases involving workers in the gig economy that were decided in 2021. We take into account recent developments in the law that may point toward trends of how workers are classified as well as important changes to Uber’s contracts, and we predict how judges would likely classify Uber drivers for federal tax purposes in different scenarios.

I. Background

Courts across the world have come to differing conclusions about whether a particular type of worker is classified as an employee or independent contractor, depending on varying definitions of what constitutes an employee relationship in different jurisdictions.3 California, Uber’s home jurisdiction, is a particularly interesting battleground because it is notoriously unclear what the applicable worker classification test is in the gig economy context. New employment legislation that took effect on September 4, 2020, codified the governing case law establishing an “ABC test” for worker classification purposes. This test stipulates that a worker is to be considered an employee unless all of the following factors can be shown by the hirer:

  1. that the worker is free from the control and direction of the hiring entity in connection with the performance of the work, both under the contract for the performance of the work and in fact;

  2. that the worker performs work that is outside the usual course of the hiring entity’s business; and

  3. that the worker is customarily engaged in an independently established trade, occupation, or business.

This test was developed in Dynamex,4 and compared with other U.S. jurisdictions, it is one of the most difficult hurdles to overcome for hirers that would like to see their workers classified as independent contractors. It appears that Uber may find it challenging to meet these stringent requirements.5 For example, Uber Eats drivers perform the core function of Uber Eats’ business — that is, food delivery — and are therefore likely going to be considered to be performing work in the usual course of Uber Eats’ business under the B portion of the test. Anticipating the challenges, Uber invested millions to support Proposition 22 in California, which took effect on December 16, 2020. Proposition 22 stipulates an exception to the employee classification for app-based workers under some conditions — for example, granting the worker flexibility for the hours worked and permitting them to work for competing companies.6 A recent case that addressed, but doesn’t definitively decide, the applicability of Proposition 22 over the ABC test is Hassell.7 Further, in January 2020 Uber changed its contract with Uber ride-share drivers to give them more flexibility and independence, a move that it claims allows Uber to meet the requirements of the new law.8 Uber’s strategy to set out and change the labor law does not stop in California. The company is making similar efforts across the United States and Canada.9

In the following sections, we will look at the dispute from a federal tax perspective rather than an employment law perspective and outline what is at stake for Uber and its workers. We will use the facts set out in James — a case involving Uber ride-share drivers decided in January in California under Dynamex — to engage in scenario testing and determine the relative effect of the factors offered in support of the claim that Uber drivers should be classified as employees for employment law purposes. This case was chosen because the factual considerations under Dynamex are comparable to those under the IRS 20-factor test. But first we will outline the general federal tax consequences of worker classification.

II. (Mis)classification Consequences

As the pandemic has made clear, the most problematic consequence of classifying a worker as an independent contractor rather than an employee is the former’s lack of access to employer-funded unemployment insurance. Unlike employees, independent contractors pay a self-employment tax of 15.3 percent under section 1401, which functions as a self-employed person’s contribution to Social Security and Medicare.10 A self-employed person can deduct an employer’s equivalent of the self-employment tax when calculating their adjusted gross income.

If an employee is wrongly classified as an independent contractor and this classification goes unchallenged, employers save significant expenses. These avoided expenses could include the employer’s half of the Social Security pension contribution and the Medicare tax (together 7.65 percent of the gross income)11 and state and federal unemployment insurance tax. Section 3509(b) and section 6672 impose a penalty on the employer if a worker is intentionally misclassified. It is also possible to bring a personal action against the business owner or officers. The employer is specifically prohibited from recovering the penalty amount from the worker. Moreover, under section 6721, the employer must pay a minimum of $50 per misclassified employee per year — that is if there is no intent, and a timely correction is made. If fraud is at issue, section 7202 provides for imprisonment for up to five years and an additional penalty payment of up to $10,000.

To summarize, from a tax perspective, much can be gained from the proper classification of a worker as an independent contractor, and much can be lost if there has been a misclassification, particularly if this misclassification is done intentionally and fraudulently. We will now look at the factors relevant for the worker classification in the James case using Blue J’s prediction module to determine whether under the IRS test James and the purported members of his class action would likely be classified as independent contractors or employees.

III. Factors in James

James was decided by the U.S. District Court for the Northern District of California on January 26. Uber driver Christopher James asked the court to certify a class action including all Uber drivers who have worked in California and have not released their claims against the company. In the course of its decision on the class action certification, the court had to consider the facts pertaining to the A and B prongs of the Dynamex test. The C prong was disregarded for the class action certification because the court found that it required individual consideration for each driver and therefore did not lend itself to a determination for the entire class.

For purposes of the A and B prongs, however, the court considered the following facts that will also prove relevant for worker classification under the IRS test:

  • Some drivers worked full time, some part time.

  • Uber exercised no control over the drivers’ schedules other than requiring at least one ride-share service per month (before the contract change considered below).

  • In January 2020 Uber changed its contract to permit drivers:

    • to develop relationships with particular riders who can specifically request them;

    • to use the Uber application infrequently without penalty;

    • to simultaneously use competing applications like Lyft;

    • to contract directly with passengers outside of Uber;

    • to hire employees to drive on their behalf;

    • to decline ride requests without facing any penalties (no more one-drive-per-month requirement); and

    • to set their own fares.12

  • Some drivers have taken advantage of the 2020 contract and drive for the competition, and among those drivers, some own independent driving businesses in the course of which they employ subcontractors or employees.

  • Uber can discharge drivers at will.

  • Uber requires drivers to maintain a minimum rating.

  • Uber does not guarantee minimum wage and overtime pay.

  • As of December 16, 2020, if Uber wants to take advantage of Proposition 22 and have its drivers classified as independent contractors under section 7451 of the California Business and Professions Code, it must also comply with section 7453 of the same code that requires Uber to guarantee a “net earnings floor.” This compensates drivers at a rate that is 120 percent of the minimum wage for hours during which the drivers are actively engaged on the app and provides vehicle expense compensation of 30 cents per mile.

The court in James did not consider this, but sections 7454 and 7455 also provide for healthcare benefits and loss and liability insurance. In the following section, we will probe the relevance of some of these factors for a worker classification under federal tax law.

IV. Blue J’s Predictions

For each of the following scenarios, we assume that the workers provide their own vehicles at their own cost and are not otherwise at risk of any loss. Moreover, we assume workers must maintain a minimum rating and must abide by specific standards that Uber sets. We also assume that there is a continuous relationship between Uber and its workers. Further, we will alter each scenario reflecting the outcome for part-time and full-time work, as these are both common approaches drivers take that have a significant effect on the predicted outcome (the influence ranges between 14 and 19 percent). Whether drivers also work for competing service providers is a similar factor that can have a significant effect on the outcome, and it is present for many but not all workers. We changed this factor to be in sync with the part-time factor, as it seems plausible that a part-time Uber driver may work for another ride-share service provider during the remaining time.

A. Scenario 1

In Scenario 1, we predict the worker classification of Christopher James for federal tax purposes before January 2020. James was working full time and exclusively for Uber. He received no minimum base pay or any benefits and did not have his own business, assistants, or the opportunity to profit on his own initiative, especially when it came to setting the price for the ride. Blue J predicts that James would be an employee for tax purposes with 69 percent confidence. However, if he only worked part time and for Uber’s competitors, he would likely be an independent contractor (with 60 percent confidence).

B. Scenario 2

Scenario 2 predicts the outcome for the James case for those workers of the purported class action that were governed by the January 2020 contract with Uber and made full use of the flexibility the new contract provides. We assume, however, that Proposition 22 does not yet apply. Hence, this scenario sets out the situation many California Uber drivers were in from January 2020 to December 16, 2020, or rather until Uber complied with the requirements as set out in the amended California Business and Professions Code. Thus, the drivers in our Scenario 2 are not entitled to a minimum payment and compensation for their expenses, and they receive no healthcare benefits. Yet they are allowed to (1) develop relationships with particular riders who can specifically request them; (2) use the Uber application infrequently without penalty; (3) simultaneously use competing applications like Lyft; (4) contract directly with passengers outside Uber; (5) hire employees to drive on their behalf; and (6) may decline ride requests without facing any penalties.13 We capture these six factors under “Own driving business” in the table. The January 2020 contract also allows drivers to deviate from the suggested price, which we interpret as an opportunity to profit from one’s own management or initiative. We assume first that the drivers are working for Uber only part time. In this scenario, Blue J predicts with 83 percent confidence that the workers would be classified as independent contractors for federal tax purposes.

If the workers were to work for Uber part time and for its competitors as well, they would likely still be considered independent contractors, but with only 69 percent confidence (Scenario 2a). Next, we assume there are workers among those described in Scenario 2a that only take advantage of some of the liberties provided in the January 2020 contract. If we change the answer to the question whether a worker hires his own employees to drive on their behalf to “no” (Scenario 2b), our predictor is only 56 percent confident that the workers would be classified as independent contractors.

C. Scenario 3

Scenario 3 reflects the situation of Uber drivers in California taking full advantage of the new flexibility the January 2020 contract provides them, and they are governed by Proposition 22, guaranteeing a minimum base pay and compensation for some expenses. We assume that they are working part time for Uber. In this scenario, we disregard the healthcare benefits that must be awarded under section 7454 to isolate the effect of the benefits factor from the minimum pay and expenses factor. In this scenario, Blue J predicts that the workers would be classified as independent contractors with 67 percent confidence. Changing the scenario to account for part-time workers that also drive for the competition (Scenario 3a), the outcome is inconclusive: Blue J predicts with 51 percent confidence that the workers would be considered employees.

D. Scenario 4

Scenario 4 assumes that Uber complies with all the requirements set out in the amended California Business and Professions Code, including healthcare benefits.14 For part-time Uber drivers, this means that they will likely be classified as employees. Our predictor is 65 percent confident of this outcome. Changing these facts to account for full-time Uber drivers (Scenario 4a), Blue J predicts with 79 percent confidence that those workers will be classified as employees.

Table. Blue J’s Predictions

 

Flexibility

Minimum Pay and Expenses

Part-Time and for Competition

Profit Opportunity

Own Driving Business

Benefits

Outcome

Scenario 1

Yes

No

No

No

No

No

69% employee

Scenario 1a

Yes

No

Yes

No

No

No

60% independent contractor

Scenario 2

Yes

No

Yes

Yes

Yes

No

83% independent contractor

Scenario 2a

Yes

No

No

Yes

Yes

No

69% independent contractor

Scenario 2b

Yes

No

No

Yes

No assistants

No

56% independent contractor

Scenario 3

Yes

Yes

Yes

Yes

Yes

No

67% independent contractor

Scenario 3a

Yes

Yes

No

Yes

Yes

No

51% employee

Scenario 4

Yes

Yes

Yes

Yes

Yes

Yes

65% employee

Scenario 4a

Yes

Yes

No

Yes

Yes

Yes

79% employee

V. Analysis

One striking conclusion from this Blue J Predicts analysis is that there is no such thing as the classification of an Uber driver. With the exception of Scenario 4, individual differences such as working full time or part time and for the competition, and providing the work personally or via assistants, are significant enough to flip the most appropriate classification, or at least to render it an uncomfortably close call.

A further notable conclusion is that Uber’s January 2020 contract has a significant effect on the classification of its workers. The flexibility and independence the contract provides increase the likelihood of an independent contractor classification by 23 percent, comparing part-time drivers who worked for the competition pre- and post-January 2020 (Scenario 1a versus Scenario 2). For federal tax purposes, Uber may well have achieved its goal to make its workers truly independent contractors by implementing this new contract.

Yet, while section 7451 of the California Business and Professions Code may well have the effect that Uber drivers are independent contractors under that state’s employment law, the statute’s requirement of a net earnings floor and compensation for vehicle expenses noticeably weigh in favor of an employee classification in the tax context. The data show that Uber may have weakened its own position by successfully lobbying for this legislation. Before taking effect, it was 83 percent likely that a part-time Uber driver with all the liberties under the January 2020 contract would be classified as an independent contractor for federal tax purposes (Scenario 2), while with the new legislation, the same outcome is going to be obtained with only a 67 percent likelihood (Scenario 3). If an Uber driver then decides to work for Uber full time, the classification could go one way or another.

Lastly, assuming, as in Scenario 4, that Uber provides healthcare benefits, all the liberties under the January 2020 contract cannot save Uber’s case arguing for an independent contractor classification. Be it part-time or full-time work, the result is likely going to be an employee classification under federal tax law according to Blue J’s machine-learning model of worker classification.

VI. Conclusion

Machine-learning models can provide invaluable insight for hiring entities in positions similar to those faced by Uber and other gig economy service providers. Our analysis has revealed that the proper worker classification cannot be undertaken for an entire workforce at once, absent legislation that determines the question one way or the other. Instead, the most accurate approach is likely to be one based on a fact-based inquiry that is tailored to the working terms and conditions faced by different kinds of workers (for example, those that work part time, for the competition, on their own, or via assistants, etc.) Equipped with a predictive machine-learning model such as the one we used for our scenario testing, the task of gauging the risk of misclassifying workers can be made simpler and faster, leading to better-informed risk-taking and more effective lobbying, risk mitigation, and tax and employment law litigation strategies.

FOOTNOTES

1 Lawrence F. Katz and Alan B. Krueger, “The Rise and Nature of Alternative Work Arrangements in the United States, 1995-2015,” 72(2) ILR Review 382-416 (2019).

2 In our hand-collected data set of 361 federal cases involving worker classification for tax purposes, there are 183 cases (50.7 percent) that found that the workers should be classified as employees and 178 (49.3 percent) that found that the workers should be classified as independent contractors.

3 Uber BV v. Aslam, EWCA Civ 2748 (2018), (finding that Uber drivers are employees); Federatie Nederlandse Vakbeweging v. Uber BV, 8937120 CV EXPL 20-22882 (2021) (finding that Uber drivers are employees); Cour de Cassation, Chambre Sociale, Ruling no. 374 of March 4, 2020 (19-13.316) (finding that Uber drivers are employees); Razak v. Uber Technologies Inc., Civil Action No. 16-573 (E.D. Pa. 2018) (classifying Uber drivers as independent contractors; however, the Third Circuit remanded the case for further proceedings, as summary judgment was found to be inappropriate); McGillis v. Department of Economic Opportunity, 210 So.3d 220 (Fla. App. 2017) (finding Uber drivers were independent contractors); Marcio Vieira Jacob v. Uber do Brasil Tecnologia Ltda., RR-1000123-89.2017.5.02.0038 (Feb. 6, 2020) (Tribunal Superior do Trabalho (TST), Brazil) (Brazil’s Superior Labor Court held that Uber drivers were independent contractors.).

4 Dynamex Operations West Inc. v. Superior Court of Los Angeles, 416 P.3d 1 (Cal. 2018).

5 Hassell v. Uber Technologies Inc., No. 20-cv-04062, at 4 (N.D. Cal. 2020); Nicolas v. Uber Technologies Inc., No. 19-cv-08228, at 5 (N.D. Cal. 2021).

6 California Business and Professions Code, section 7451.

7 Hassell, No. 20-cv-04062.

8 James v. Uber Technologies Inc., No. 19-cv-06462, at 14 (N.D. Cal. 2021).

12 James, No. 19-cv-06462, at 14.

13 Id.

14 Note that providing healthcare benefits to its workers is something Uber is prepared to do beyond California; see, e.g., Uber Canada, “A Blueprint for a Flexible Benefits Fund for App-Based Workers in Canada” (Aug. 30, 2021).

END FOOTNOTES

Copy RID