Disrupting Work Law: Arbitration in the Gig Economy

I.     Disrupting Precedent: Individual Arbitration for Gig Economy Workers

A.    Individual Arbitration Clauses: The Critique

As arbitration has become an increasingly prevalent mechanism of dispute resolution—and especially as it has made inroads into the consumer and employment arena—scholars and advocates have constructed a multi-part critique charging that arbitration systematically disadvantages plaintiffs and the public. This section teases apart the various strands of that critique, separating the arguments that go to the difference between arbitral and judicial forums from those that go to the disaggregation of claims in arbitration that would otherwise be brought on a class or collective basis. This section provides background necessary to evaluate the likely effects of IACs in the gig economy, the prevalence of which I discuss in Part I.B.

1.    Arbitration vs. Litigation

Jean Sternlight, a leading scholar of arbitration, argues that employment (or consumer) arbitration creates both private and public harms, writing that “mandatory arbitration is problematic for two fundamental reasons: lack of consent and lack of public scrutiny.”3 Thus, for Sternlight, there are two distinct sets of harms associated with arbitration: those primarily suffered by individuals who have their claims sent to arbitration; and those that affect the public in general.

The individual-level critique has at least two prongs, roughly focused on the likelihood that people will agree to arbitrate disputes without knowing or understanding what they are agreeing to; and the likelihood that the arbitral process itself will turn out to be unfair or will lead to worse outcomes for employees. As to the first, many arbitration agreements in consumer or work contracts are easy to overlook; they are printed in small type, or they are buried in dense contractual language in a lengthy contract or employee handbook or—more likely in the computer and smartphone age—behind a link that the signer is unlikely to click.4 Moreover, in a study in which researchers asked consumers about typical contracts containing arbitration clauses, only a minority of the respondents who recognized the presence of arbitration clauses understood that they would preclude signatories from proceeding in court, and a large majority thought signatories could participate in a class action despite a class waiver.5 And even those intrepid and motivated employees (or consumers) who both read and understand an arbitration clause before signing it may lack the background knowledge of arbitral and judicial processes required to make an informed choice, or may feel that an attempt to negotiate over the terms would be futile. Even if neither is the case, a new employee or job applicant may be unwilling to negotiate employment terms in general, and particularly reticent to raise terms that will become important only in the event of a dispute with a new employer.6

Second, arbitration may be difficult to access or skewed towards employers at least as compared to litigation, although the evidence on this point is not conclusive. Arbitration critics focus largely on the possibility that the arbitral forum will be more expensive to access or farther away from the plaintiff ’s home or work than a judicial forum, that discovery will be limited or unavailable, and the chance that the “repeat player” phenomenon will erode workers’ or consumers’ recoveries.7

Nonetheless, some employees may genuinely prefer arbitration to litigation on the ground that it takes less time to get a decision and—at least where the employer pays the arbitral forum fees—can be cheaper than litigation.8 Further, arbitrators’ relative freedom to deviate from legal rules can sometimes benefit plaintiffs,9 and the lack of motions practice in most arbitrations means that plaintiffs are more likely to get a chance to tell their story to the decision-maker—a possibility that has intrinsic value for some plaintiffs.10

Unfortunately, empirical evidence about how workers fare in arbitration is inconclusive.11 Some evidence suggests that employees have a lower win rate, and also a lower median recovery when they do win,12 in contrast to earlier studies that arrived at a more positive view of arbitration.13 And even assuming it is true that employees are less likely to win in arbitration, that data does not reveal why this variation occurs. Perhaps claims that reach arbitration are, on average, weaker than cases that reach litigation, among other possibilities.

Next, there is the public-focused critique of arbitration. Arbitration is, by definition, a private process; arbitration awards will not be made public without the participants’ consent.14 Thus, it is difficult to shine light on arbitral processes, either to cover high-profile disputes in the media, or to study the effects of arbitration.15 Further, arbitration does not result in precedential decisions, public oversight and attention of arbitral processes is at best difficult to achieve, and arbitration (especially individual arbitration) lacks class litigation’s deterrent effect on enterprises. In her book Boilerplate,16 Margaret Jane Radin put this critique in terms of “democratic degradation,” writing that “[w]hen a firm’s mass-market boilerplate withdraws a number of important recipients’ rights . . . it is displacing the legal regime enacted by the state with a governance scheme that is more favorable to the firm.”17 Those more favorable and more private regimes reduce the cost of lawbreaking, even taking into account that most class actions settle.18 Finally, Myriam Gilles warns of knock-on effects on future litigants when workers are unable to bring their claims in court, writing that “[w]hen judges are no longer confronted regularly with the civil claims of the poor . . . they will become unversed in and desensitized to the underlying factual issues that affect lower-income groups.”19

2.    Aggregated vs. Individual Resolution

Assume for a moment that arbitration and litigation are equally promising avenues for workers, or that arbitration is more promising for at least some types of disputes brought by some workers. To put it another way, even if employees are unlikely to prevail in arbitration, courts might be just as bad; indeed, many scholars who criticize arbitration for its effects on workers and consumers also heavily criticize federal litigation.20

But even if that assumption is correct, workers who have signed arbitration clauses will still be at a disadvantage if they lack meaningful access to an arbitral forum in disputes where, as a practical matter, a judicial forum could be available. Here, a key consideration for workers with relatively low value claims will be whether they will be less likely to be able to aggregate their claims in arbitration than in litigation.

Two recent Supreme Court decisions make it far less likely that workers (or consumers) will be able to arbitrate their claims on a class or collective basis. First, in AT&T Mobility v. Concepcion,21 the Court held that the Federal Arbitration Act preempted a California rule that IACs were unconscionable when they precluded the aggregation of small-value claims.22 Then, in American Express v. Italian Colors Restaurant,23 the Court held that the undisputed fact that the costs of prosecuting the plaintiff ’s statutory claim in arbitration were prohibitive did not provide grounds under the FAA to refuse to enforce an IAC.24 As Professor Martin Malin has argued, these decisions largely eliminate judicial supervision of IACs by taking a cramped and formalistic view of the effective vindication doctrine—the principle that arbitration agreements will be enforced only when they provide an avenue for plaintiffs to effectively vindicate their rights.25 Thus, a worker who signs an agreement containing an IAC will not be able to argue that she should be able to pursue her claims on an aggregate basis because it is cost-prohibitive to pursue them on an individual basis.26 The obvious advantages of such clauses to corporate defendants have led scholars to predict a swift end to much aggregate claims resolution in the consumer and employment contexts. As Brian Fitzpatrick put it, there is “every reason to believe that businesses will eventually be able to eliminate virtually all class actions that are brought against them.”27

Even where companies commit to picking up arbitral forum costs (a common—though not ubiquitous—term in gig economy agreements, as discussed in the next section), workers are still unlikely to arbitrate their low-dollar claims for a list of reasons, including difficulty retaining counsel. Data bear out this hypothesis. Arbitration provisions in employment contracts have become ubiquitous in the last twenty-five years. Whereas a “1991 survey found fewer than four percent of firms requiring arbitration in employment,” by 2008, “a quarter or more of all non-union employees in the U.S. . . . were covered” by arbitration agreements,28 and another study showed that figure was yet higher by 2014.29 Yet, “[b]etween 2010 and 2013, the [American Arbitration Association] reported 1,349 to 1,599 filings nationwide under employer-promulgated arbitration obligations”30 —a shockingly low number.

Finally, there is little reason to believe that outcomes in the gig economy will be any different. “Low-dollar” is a fitting term to describe many gig workers’ claims; for example, a California Uber driver who convinced her state’s Department of Industrial Relations that she was an employee was awarded $4,152.20 in reimbursable expenses and interest.31 To be sure, workers could represent themselves in arbitration, as this Uber driver did before the Department of Industrial Relations—and some proponents of arbitration tout that process’s informality as friendly to pro se litigants32 —but as a practical matter, many workers will be unable to competently represent themselves, or unwilling to make the attempt.33 Pro se representation will be especially difficult for prospective plaintiffs who are pressed for time, including those who are augmenting their income by driving for Uber.

B.    Individual Worker Arbitration in the Gig Economy

To evaluate IACs in the gig economy worker contracts, I began by collecting a set of fourteen work agreements.34 Some of these were readily available on company websites; many others were attached to motions to compel arbitration filed by companies in response to worker lawsuits. In particular, the O’Connor v. Uber35 case, which involved protracted wrangling over both the enforceability of Uber’s individual arbitration clauses and the extent to which Uber may communicate with class-member drivers, yielded a series of contracts that changed over time. I analyzed each of these agreements as part of this project. I further indicate where I am counting multiple Uber agreements.

Accordingly, the discussion that follows does not reflect either a random sampling of contracts or the total universe of gig economy contracts. For example, I have been unable to locate employment agreements for the handful of gig economy enterprises that classify their workers as employees, so it is possible that these companies do not impose individual arbitration on these employees. Instead, the discussion emphasizes enterprises that have become embroiled in misclassification litigation. Still, it bears emphasis that the agreements discussed below cover the largest gig economy enterprises, and a large majority of gig economy workers.36 Thus, even if there exists more variation in enterprises’ approaches than the following discussion suggests, it is nonetheless true that the large majority of gig economy workers are covered by contracts containing IACs, as discussed below.

1.    Features of gig economy arbitration agreements

My analysis of gig economy IACs revealed several common terms, but also significant contrasts. These features are key to comprehending the various impacts of IACs, and are dissected further below.

a.    Similarities among agreements

i.     Individual arbitration clauses and class action waivers

Each contract I reviewed contained a clause stating clearly that disputes between signatory workers and the company would be resolved in arbitration. Most agreements also added plain statements precluding any class or collective arbitration,37 though those statements were unnecessary in light of the principle that class arbitration must be explicitly authorized.38 In addition, about half of these agreements also waived a jury trial in the event that a court struck down the arbitration clause (or the parties agreed to waive it).39

ii.    Severability and the possibility of class or collective litigation

Each agreement I collected contains at least one severability clause that covered the whole agreement, with most also adding an additional severability clause within the IAC itself.40 These clauses have proven critical to the enforcement of IACs containing unenforceable terms.41 Thus, while severability clauses are surely intended at least in part to guard enterprises that draft IACs against reasonable errors in legal judgment or changes in the law, they also have a more troubling effect: they reduce enterprises’ incentives to diligently remove unenforceable provisions from IACs once they have been identified, because their presence will not invalidate the entire IAC. In fact, in several cases, enterprises have simply agreed to waive problematic provisions when they were challenged in court without attempting to defend them.42 That defense lawyers did not even attempt to argue for these provisions’ validity suggests that the IACs were, at minimum, not carefully vetted by lawyers at the time of drafting. Yet, a worker reviewing an IAC without the advice of counsel may simply assume that the entire agreement will be enforced, and decide not to arbitrate because she cannot travel to the west coast or advance significant arbitration forum fees.43

There is an important exception to the severability clauses contained in these agreements: most direct that, in the event that a class or collective action waiver is deemed unenforceable, the case must return to court.44 Such clauses are common in pre-dispute arbitration agreements; for example, the Consumer Financial Protection Bureau’s 2015 study of consumer financial agreements found that while nearly every agreement precluded class litigation or arbitration, most also “contain an ‘anti-severability’ provision stating that if the no-class arbitration provision were to be held unenforceable, the entire arbitration clause should be deemed to be unenforceable as well.”45

In other words, gig economy enterprises have clear preferences: individual arbitration is their first choice; aggregated litigation is their second; and aggregated arbitration is heavily disfavored. This might seem counterintuitive; surely the benefits of arbitration—simpler, more streamlined procedures—are not lost simply because a single arbitration will involve an aggregated set of claims. But arbitration awards are subject to judicial review only in extremely limited circumstances,46 and parties may not expand the grounds on which a court may refuse to enforce an arbitral award by contract.47 Thus, even with the possibility of an appeal within the arbitral forum48 —an option present in two of the agreements I reviewed49 —general counsel may reasonably conclude that they are unwilling to sacrifice appellate review in what could become a “bet the company” proceeding.50 Indeed, the Concepcion Court identified the lack of appellate review as a reason for rejecting California’s rule permitting the aggregation of low-value claims in arbitration.51

Thus, the agreements I reviewed are similar in key ways. However, they also revealed some more granular differences, some of which have important effects on workers’ attempts to challenge their employment status.

b.    Differences among agreements

i.     Arbitral procedures

In addition to the possibility of appeals from arbitral awards discussed above, the contracts take different approaches to arbitral processes, including:

Pre-arbitration negotiation: Four agreements require a period of negotiation before proceeding to arbitration;52 a fifth encourages negotiation by partially waiving an arbitral filing fee for drivers with claims under $5,000.53

Discovery: Most agreements allow for limited discovery.54 However, one agreement allows the same discovery that would be available in a judicial forum.55

Arbitration based on documents alone: Four agreements allow for the possibility of arbitration based on documents alone, or for phone or online arbitration in at least some circumstances.56

Exhaustion of administrative remedies: Ten agreements state that the IAC does not waive drivers’ rights to make complaints with administrative agencies.57 Nine of those agreements also make failure to exhaust administrative remedies a defense in arbitration, and one purports to preclude drivers from receiving damages in an agency proceeding.58

Location: While most agreements indicate that arbitration will occur at a mutually agreeable location or in the locality where the worker operates, three agreements require arbitration to occur in the enterprise’s home city—Seattle in one instance, and San Francisco in two.59 However, courts have held that the latter terms are invalid as to workers who live and work far from the designated cities.60

ii.    Arbitral forum fees

A majority of IACs indicate that the enterprises will pay all or nearly all of the forum fees in at least some circumstances. The GrubHub agreement is the most unequivocal, stating simply that it will “pay all of the Arbitrator’s fees and costs.” Four other companies state that they will pay the forum fees and costs either “unless law requires otherwise,”61 or unless the complaint is ultimately adjudged to be frivolous.62

Other agreements put drivers on the hook for at least a portion of arbitral forum fees, which can include both the arbitrators’ hourly rate and any travel costs; the costs of renting a conference room in which to hold the arbitration; and potentially other costs, such as hiring a stenographer. Some agreements limit these costs to the amount the driver would have had to pay in order to file a complaint in court,63 but others go farther. For example, Instacart’s IAC called for the parties to “equally advance all of the arbitrator’s expenses and fees.” Similarly, Uber’s agreements prior to December 2015 required drivers to split forum costs with Uber unless governing law forbade that arrangement.64 At least for workers with low claim values, these clauses are likely unenforceable under what remains of the “effective vindication” doctrine after Italian Colors.65 Recognizing that possibility, gig economy enterprises with fee splitting provisions in their IACs have declined to defend these clauses, and instead committed to paying drivers’ forum costs.66

iii.   Representative claims, including claims under California’s Private Attorney General Act

The majority of agreements waive workers’ rights to bring “representative” claims more broadly in addition to class or collective actions. However, several agreements also treat separately a specific form of representative action: claims under California’s Private Attorney General Act (PAGA). PAGA allows employees to step into the shoes of the state to sue their employers on behalf of themselves and other employees for workplace violations; successful PAGA plaintiffs can recover statutory penalties, with three-quarters of any penalty going to the state.67 These claims are discussed in more detail below; for now, sufficed to say that about half of agreements recognize that PAGA claims may not be waived, and call for such claims to be litigated rather than arbitrated.68

iv.   Opt outs

One of the most important differences among the IACs I reviewed is whether they offer signatories an opportunity to opt out of arbitration. Perhaps surprisingly, all but six agreements offered an opportunity to opt out.69 Of those, one nonetheless offered an email address to which workers who wanted to “negotiate” contract terms could write.70

The remaining agreements offered workers a 30-day window in which to opt out of the IAC. Most agreements allowed workers to opt out by sending an email to a designated address. While this process is relatively straightforward, it is still more difficult than accepting the worker agreements containing IACs in the first place—acceptance of the overall agreement usually requires workers to simply click a button, or in some cases, continue using the platform.71

Finally, to the extent a trend can be gleaned from a relatively small number of agreements, the trend is towards an increasing right to opt out of IACs. For example, both GrubHub and Handy (formerly Handybook) changed their IACs to permit opt outs.72 And while each Uber agreement in my set allowed drivers to opt out of arbitration, it shifted from requiring overnight or hand delivery to permitting drivers to opt out by email, albeit in response to a judge’s order.73

Enterprises have a clear incentive to offer workers a chance to opt out of IACs: doing so makes it very likely that a court will uphold the IAC as a whole. Courts will generally enforce IACs unless “the party resisting arbitration . . . [proves] that the claims at issue are unsuitable for arbitration,”74 or that the arbitration contract is invalid under generally applicable contract law,75 which in many cases is that of California. Seeking to invalidate IACs under California contract law, plaintiffs often argue that the IACs are both procedurally and substantively unconscionable, and therefore unenforceable.76 And a key mark of procedural unconscionability is the imposition of the IAC via adhesion contract.77 Thus, several courts evaluating gig economy IACs have found the presence of an opt out to be a controlling factor in their decision to enforce the IAC;78 as one court put it in the context of a motion to compel arbitration filed by Uber, “the [IAC] provisions were not unconscionable because the plaintiffs had the right to opt out from those provisions.”79

Moreover, the costs to gig economy enterprises of offering an opportunity to opt out of IACs seem to be small, as anecdotal evidence suggests that few workers actually opt out. For example, in Tan v. Grubhub, the court denied class certification after Grubhub showed that only two delivery drivers of thousands had opted out of the IAC during the months after Grubhub added its opt out clause.80 Similarly, only 270 drivers out of more than 160,000 opted out of one of Uber’s IACs, leading a district judge to write that “the opt-out right . . . was essentially illusory and ineffective”81 —though that conclusion was later reversed by the Ninth Circuit in Mohamed v. Uber Techs., Inc.82 Of course, while many drivers probably did not consider opting out because they did not notice or fully comprehend the arbitration clause and its opt out provision, it is conceivable that some drivers decided not to opt out based on their assessment that they might have individual claims that they would prefer to arbitrate than litigate—employment discrimination claims, for example, might fall into this category. However, as is discussed in greater detail below, far more than five percent of drivers would have been well-advised to opt out of the IAC.

v.     Subject matter carve-outs

More than half of the agreements in my set contain a carve-out for at least some types of claims. Significantly, three agreements exclude intellectual property disputes from their IACs; perhaps it is telling that these enterprises apparently prefer arbitration for claims drivers are likely to bring, but litigation for claims they are likely to bring.83 The remaining subject matter exclusions cover topics such as workers compensation, disability insurance, and unemployment insurance.84

vi.   Amending the IAC

Half of the agreements I collected contained a process for amending the agreement. Of those agreements, the majority created a one-sided amendment process wherein enterprises could change terms at any time, and workers who continued to use the platform would be deemed to have consented to any changes. Moreover, these agreements did not include a mechanism for informing workers that changes had been made; instead, they directed drivers to periodically review the terms and conditions listed on the enterprise’s website.85 A minority of agreements stated that amendment would take affect only upon affirmative consent by workers.86

2.    Effects of gig economy arbitration agreements: successful enforcement actions and lowered settlements

Courts have already had occasion to consider the enforceability of gig economy IACs in a set of cases. This issue typically arises after a worker or potential class of workers files a lawsuit in court, and the enterprise moves to compel arbitration. Often, these workers are represented by competent and knowledgeable counsel—in particular, the prominent Boston plaintiffs’ attorney Shannon Liss-Riordan has filed a significant number of the cases discussed in this article87 —so it is not necessarily the case that the plaintiff was unaware of the arbitration clause at the time of filing. Rather—and consistent with the critique of individual arbitration described above—these plaintiffs have evidently concluded that their best chances lie in incurring the litigation expenses associated with attempting to invalidate IACs in the hope of being able to proceed on a class or collective basis.

One possible explanation for this choice could be that lawyers—not clients—are the primary beneficiaries of class or collective action procedures, because they stand to win hefty fee awards as part of even modest settlements. Perhaps, then, it is worth it to lawyers to roll the dice on even unlikely litigation strategies, because the rewards associated with convincing even one enterprise to settle are stratospheric. Indeed, the final settlement in Cotter v. Lyft88 included a fee award of $3.65 million, and a proposed (but rejected) settlement in O’Connor v. Uber89 included a fee request of between eleven million and fifteen million dollars.90 But that is not the only story to tell about lawyers’ incentives to take (or not take) cases: as discussed above, where workers’ claims are worth too little to attract counsel, the choice is aggregated litigation or nothing. Even then, as discussed below, IACs depress the settlement value of workers’ class actions, because of the high likelihood that a court would ultimately enforce IACs. Indeed, courts have done just that in a series of cases.91

The settlement in Cotter and proposed settlement in O’Connor shed light on how IACs reduce the value of workers’ claims. First, in Cotter, the plaintiffs were three drivers who sought to represent a class of about 163,000 Lyft drivers in California, alleging that they were misclassified employees who were entitled to reimbursement of tips and expenses incurred while driving for Lyft.92 Before seeking class certification, the parties filed cross-motions for summary judgment on the ultimate question of whether the plaintiffs were either independent contractors or employees.93 The district court denied both motions, and held that the classification issue should be decided by a jury.94 However, before that could happen, the parties reached a settlement agreement as to the entire putative class. That settlement included a $12.25 million pot, as well as limited non-monetary relief,95 but did not include the re-classification of any drivers as employees.96 In exchange, drivers waived any and all claims related to their wages or employment status through the date of settlement.97

The district court rejected this settlement, in part due to a calculation error that resulted in the total settlement amount falling short of the parties’ stated understanding of what drivers would receive,98 and in part because the parties attributed only $122,250 of the total settlement amount to the drivers’ claims under California’s PAGA statute.99 Shortly thereafter, the parties returned with a new proposed settlement, providing $27 million in monetary relief to the class, with $1 million allocated to the PAGA claim, as well as similar non-monetary relief as in the original settlement. This time, the district court approved the settlement.100 That approval, which works out to an average of about $150 per driver, was based on two threats to the plaintiffs’ case: first, the possibility that they would lose on the merits, because a jury would conclude that the drivers were actually independent contractors; and second, the possibility that Lyft would successfully compel the drivers to arbitrate their claims individually.101 That the court did not rate as even greater the chances of a successful motion to compel arbitration (such that it might have approved even the lower amount proposed during the parties’ first attempt at settling the case) was based on two factors. First, the court observed that Lyft arguably waived its right to compel arbitration by litigating the case in court through the summary judgment stage without attempting to compel arbitration. The court asserted that Lyft could not engage in gamesmanship by proceeding in litigation unless and until the case began to go badly for the company, and then parachute into arbitration.102 Second, the district court cited a National Labor Relations Board rule barring IACs in employment contracts, observing that “there is at least some authority suggesting the arbitration provision is unenforceable entirely, because it violates the National Labor Relations Act (NLRA).103

The story of O’Connor is similar to Cotter in some key ways. First, the underlying claims are similar—on behalf of about 385,000 drivers in California and Massachusetts, plaintiffs sought to show that drivers were employees who were entitled to tips and expenses reimbursements.104 Likewise, the district court denied cross-motions for summary judgment, and set the case for a jury trial.105 But unlike Cotter, the O’Connor court also invalidated a series of Uber’s IACs, and ultimately certified a large class of drivers.106 The O’Connor court’s decisions invalidating Uber’s IACs rested on two main factors: first, the existence of an early agreement without an opt-out provision, which the district court found was unconscionable; and second, later agreements that containing terms waiving drivers’ rights to bring a PAGA claim, which the court found were both invalid as a matter of public policy and non-severable from the remainder of the IAC.107

While Uber petitioned for an interlocutory appeal of the district courts’ class certification decisions—which was eventually granted—the parties reached a proposed settlement agreement. That agreement created a settlement fund of $84 million, or an average of $218 per driver, which would increase to $100 million if Uber had an initial public offering that reached about $93.75 billion.108 Of this amount, $1 million was allocated to the PAGA claim.109 In addition, and similar to Cotter, the agreement contained some non-monetary terms, but did not re-classify the drivers as employees.110 Finally, in exchange for these benefits, class members were required to waive all of their misclassification-related claims, withdraw any complaints pending before administrative agencies, and resist cooperating in future attempts by agencies to investigate whether drivers were misclassified.

The district court refused to approve the settlement, but not because of the relatively low amounts of money that drivers could expect to receive. To the contrary, the district court concluded that drivers faced significant risks that justified significantly reducing the potential value of the case; among them, “the most obvious risk . . . is that the Ninth Circuit will uphold the validity of the [IACs].”111 (Indeed, that view was prescient, as the Ninth Circuit did exactly that in Mohamed v. Uber less than one month later.112 ) Instead, the district court denied preliminary settlement approval because of the meager allocation to the PAGA claims: whereas the plaintiffs argued the PAGA claim could result in $1 billion in statutory penalties (with seventy-five percent of any recovery going to the state), “[p]laintiffs propose settling the PAGA claim for 0.1% of its estimated full worth.”113

As of the time of this writing, O’Connor is still pending. The O’Connor class is facing the possibility of being mostly disbanded if class counsel cannot convince the Ninth Circuit to change its view of the enforceability of Uber’s IACs; if it cannot, then all that will remain of the certified class will be drivers who opted out of the IAC. In addition, drivers will be able to bring their PAGA claims in court, because those claims cannot be waived, and the relevant agreements direct that those claims be brought in court rather than in arbitration.

This section has sought to serve two functions. First, it illustrated the ubiquity of IACs in gig economy work contracts. Second, it argued that the practical effects of these IACs thus far have been to close off the possibility of resolving workers’ misclassification claims on an aggregated basis, with two effects: making it impractical for most workers to bring their claims in any forum; and impeding the resolution of the key question about gig economy enterprises’ relationship to their workforces. In the next section, I discuss whether anything can be done to facilitate the resolution of workers’ misclassification claims in litigation.

II.    Reestablishing Work Law?

Even if one agrees that IACs harm workers’ abilities to enforce their legal rights, one might respond with a shrug: the FAA has been law since 1925, and Congress shows no signs of moving to repeal or modify it. True, the Supreme Court’s expansive view of it has been a much more recent development.114 But even if future justices take a more circumscribed view of the FAA, they might not reverse existing precedent. Thus, Margaret Radin wrote in seeming despair that “because the US Supreme Court has interpreted a federal statute, the Federal Arbitration Act, to preempt states from adjudicating or legislating limits on arbitration, the fix would have to be accomplished by Congress.”115

But perhaps all is not lost. This section illustrates that state and federal agencies, and state legislatures, can play a role in limiting the more damaging aspects of IACs in work contracts. It considers what those bodies are already doing, and what more they can do, to protect workers’ abilities both to aggregate their claims and to pursue them in a judicial forum. Finally, it suggests a role for worker advocacy groups and labor unions in fighting the application of IACs.

A.    Agencies

1.    The NLRB and mandatory individual arbitration

Since 2012, the National Labor Relations Board (NLRB) has taken the view that IACs violate employees’ rights under two foundational labor law statutes by denying them at least one forum in which to aggregate their legal claims against their employer: first, the NLRA, which protects workers’ rights to engage in collective activity for “mutual aid or protection;”116 and second, the Norris-LaGuardia Act, which renders “yellow dog” contracts unenforceable.117 According to this rule, which is often called the D.R. Horton rule after the case in which it was first announced,118 employees must have at least one forum in which they can attempt to redress workplace grievances on an aggregated basis, though that forum may be arbitral or judicial. For workers who qualify as statutory employees, the NLRB’s rule prohibits mandatory IACs. And employers’ demonstrated preference for class litigation over class arbitration means that, as a practical matter, the D.R. Horton rule could effectively reinstate class litigation as the primary way to resolve claims like those in O’Connor and Cotter.

At bottom, the Board’s reasoning in D.R. Horton and later cases is straightforward. First, it is a longstanding principle of labor law that “the NLRA protects employees’ ability to join together to pursue workplace grievances, including through litigation.”119 Second, employers may not demand that employees waive their rights under the NLRA as a condition of employment.120 Moreover, the Board concluded that there was no conflict between its construction of the NLRA and the FAA, because the Supreme Court has already held that the FAA “may not require a party to ‘forgo the substantive rights afforded by the statute’” and the NLRA’s substantive guarantee is precisely the right of collective action.121 Thus, argument that IACs do not waive substantive statutory rights because those rights can be vindicated in individual arbitration simply does not make sense in the context of the NLRA, where the right on offer is the right to redress workplace grievances collectively. Or, as Judge Wood put it in a decision upholding the D.R. Horton rule: “just as the NLRA is not Rule 23, it is not the ADEA [Age Discrimination in Employment Act] or the FLSA [Fair Labor Standards Act]. While the FLSA and ADEA allow class or collective actions, they do not guarantee collective process. The NLRA does.”122

Like the Seventh Circuit, the Ninth Circuit has also upheld the NLRB’s D.R. Horton rule in Morris v. Ernst & Young.123 Both circuits concluded that the NLRA’s statutory language clearly encompassed collective litigation or arbitration, without resorting to Chevron step II.124 However, three other circuits to consider the D.R. Horton rule have rejected it. In refusing to enforce the Board’s D.R. Horton decision itself, the Fifth Circuit held that there was no substantive right to use class action procedures under the NLRA, and that in any event, the Board’s rule ran afoul of Concepcion by disfavoring arbitration.125 As to the latter point, the Fifth Circuit agreed that the NLRB’s rule was facially neutral between litigation and arbitration, but nonetheless held that the fact that employers would lose their incentive to use arbitration was enough to trigger the Concepcion rule.126 The Eighth Circuit has also rejected the D.R. Horton rule, based on similar reasoning.127 Finally, the Second Circuit has joined the Fifth and the Eighth circuits in rejecting the rule, albeit in a brief footnote that was devoid of explanatory reasoning.128

In January 2017, the Supreme Court granted certiorari in Epic Systems, Ernst & Young, and Murphy Oil, and oral argument took place on October 2, 2017; as of this writing, the cases are still pending. In the meantime, the NLRB’s D.R. Horton rule is playing a key role in the O’Connor plaintiffs’ bid to convince the Ninth Circuit to affirm the district court’s class certification decision. As discussed above, the Ninth Circuit already held in Mohamed v. Uber that Uber’s IAC is enforceable.129 However, the Mohamed court paid little attention to the D.R. Horton argument, in large part because the plaintiffs in that case failed to raise the argument until briefing was nearly complete.130 Seizing on this small opening, the O’Connor plaintiffs’ answering brief on appeal focused primarily on the D.R. Horton rule, and the NLRB filed an amicus brief in support of its rule.131 Thus, if the O’Connor class survives intact, it will be thanks to the D.R. Horton rule—though at least two doctrinal hurdles as well as the likelihood that the Supreme Court will reject the D.R. Horton rule call this outcome into doubt.

First, the O’Connor plaintiffs will have to prevail upon the Ninth Circuit to reverse its current view that the D.R. Horton rule does not apply when employees have an opportunity to opt out of an IAC. In Johnmohammadi v. Bloomingdales, Inc., the Ninth Circuit held that a non-adhesion contract did not on its face interfere with, restrain, or coerce employees into giving up their rights to engage in protected concerted activity, and therefore did not violate their rights under Norris-LaGuardia or the NLRA.132 Thus, the Johnmohammadi court, in addition to deeming the D.R. Horton argument waived, went on to observe that “[e]ven if the argument had been properly raised, . . . the option to opt out meant that” Uber’s IAC was not mandatory.133 Accordingly, the Ninth Circuit’s current view that an opt out is sufficient to save an IAC could doom the O’Connor plaintiffs’ D.R. Horton Hail Mary, and seriously limit that doctrine’s potential to help gig economy workers in general.

But here too, the NLRB may have a role to play. In 2015, the Board considered an IAC with an opt out, and held that the same core NLRA principles that support the D.R. Horton rule also preclude employers from requiring that employees take affirmative steps to preserve their rights to engage in collective action.134 And indeed, if one accepts that the NLRA protects the right to redress violations of the law on a collective basis, then it follows that employers cannot limit their exercise through a pre-dispute opt out procedure. Were it otherwise, many more employers would secure their employees’ advance promise to refrain from striking or soliciting coworkers to join a union though a similar pre-dispute procedure.

Second, there is another catch: the NLRA applies only to “employees,” and not to independent contractors,135 so there would have to be a preliminary determination of precisely the question at the heart of many worker complaints—whether gig economy workers are actually employees, albeit under the NLRA rather than the statutes on which the workers’ claims are based. The NLRA determination will not bind other tribunals or control determinations under other statutes—even where the applicable tests are very similar—meaning that workers could be independent contractors for some purposes, but not others. For example, while the list of factors that control the independent contractor/employee determination under California law are very similar to the factors that the NLRB applies,136 the California factors are applied in a way that makes it more likely that drivers will be deemed employees. Moreover, the O’Connor and Cotter courts each denied cross motions for summary judgment, and also held that, under California law, a jury was entitled to make the final call;137 of course, juries can be unpredictable. Still, in addition to its effect on IACs, an NLRB finding that certain workers qualify as employees is not irrelevant either—at a minimum, it would be persuasive authority, and such a finding could provide leverage for plaintiffs’ attorneys to achieve more favorable settlements. It may even shift public opinion about whether drivers are employees.138

While the NLRB’s D.R. Horton rule has the most potential to upend IACs in the gig economy, other state and federal agencies with law enforcement responsibilities may also play a role in pushing for certainty as to drivers’ status. So far, a handful of state unemployment offices have issued decisions regarding Uber drivers. Many of these decisions have not been made public, but Uber claims that it has prevailed before agencies in Georgia, Pennsylvania, Colorado, Indiana, Texas, New York, Illinois, and California.139 In addition, in a published decision, the Florida Department of Economic Opportunity held that an Uber driver was an independent contractor mostly on the strength of the clause in the relevant driver agreement declaring the driver to be an independent contractor.140 Conversely, California and New York agencies have found at least some drivers to be employees.141 While many of these decisions are not made public, making them a poor substitute for a public litigation process, agencies including the EEOC and NLRB may yet weigh in on gig economy workers’ statuses in public processes.

B.    Private Attorney General Statutes

A second path forward could involve representative actions, such as those that workers can bring under California’s PAGA statute. As the discussion of Cotter and O’Connor in the previous section suggests, California’s PAGA statute continues to play a significant role in gig economy misclassification cases. The interplay between PAGA and these and other gig economy misclassification cases suggests a path forward for states seeking to preserve and even enhance the public-facing benefits of litigation.

PAGA’s role in gig economy misclassification cases is twofold. First, the California Supreme Court has held that workers cannot waive their rights under PAGA—contrary to the terms of several gig economy IACs—because PAGA claims allow employees to step into the shoes of the state, and the state is not a signatory to the IAC.142 To be sure, the fact that PAGA claims cannot be waived does not mean that they cannot be arbitrated; that issue remains open, though several courts have held that PAGA claims are arbitrable, or at least that parties to IACs may delegate the arbitrability of PAGA claims to an arbitrator.143 However, that issue will often be irrelevant, because, as discussed above, enterprises often prefer to litigate PAGA claims than arbitrate them.144 Thus, as long as the U.S. Supreme Court does not hold that California’s rule against waiving PAGA claims is preempted by the FAA, PAGA will likely provide a path to court for workers who claim they have been subjected to low-value but widespread violations of California employment law.

Second, PAGA claims make cases harder to settle, and PAGA’s representative nature demands close judicial scrutiny when parties do reach settlements. First, as Cotter and O’Connor demonstrate, it is difficult to settle PAGA claims because of the structure of the statute’s damages provision. Specifically, PAGA authorizes courts to award a penalty of $100 per aggrieved employee per pay period, with that amount doubling in the case of repeat offenders.145 However, seventy-five percent of any PAGA award goes to the state of California,146 so that a negotiated settlement that releases a PAGA claim will simultaneously exhaust the defendant’s willingness to pay while doing relatively little to satisfy the plaintiffs’ demands. Further, as the O’Connor district court observed, PAGA settlements bind non-party employees without providing an option for those employees to opt out, so there is even greater imperative for district courts to scrutinize the fairness of PAGA settlements than there is as to monetary class action settlements more generally.147 To be sure, there are tradeoffs involved in making settlement more difficult,148 but for those who worry about the effects of quick class settlements, PAGA offers a partial response; moreover, increased judicial (and possibly public) scrutiny of PAGA settlements could make the public benefits of litigation discussed in Part I more robust.

This is not to suggest PAGA is a panacea. First, the same award-splitting provisions that deter settlements can also make PAGA cases unattractive to plaintiffs’ lawyers. Moreover, courts often significantly reduce PAGA awards below the maximum statutory penalty, further reducing PAGA’s deterrent effect.149 And where PAGA claims are not intertwined with non-PAGA claims for damages, the dynamics described in the previous paragraph will not apply. Still, PAGA provides a useful model for states interested in promoting judicial resolution of work law disputes.

C.    Organizing & Opt Outs

Finally, there may be effective strategies for avoiding IACs that gig economy workers can pursue on their own, possibly with the encouragement of worker advocacy groups or labor unions that have begun focusing on organizing gig economy workers.150 First, it might be tempting to say that workers should simply be more diligent in preserving their rights to a judicial forum by opting out of IACs. Thus, a campaign by worker advocates to encourage opt outs might reach enough workers who are newly signing up to work for platforms like Uber to make a difference; of course, if only a handful of workers opt out, then class-based resolution still will not be available as either practical or a legal matter.151 To be sure, some gig economy workers are already attentive to the merits and mechanics of opting out of individual arbitration clauses. For example, there exist lengthy threads on the merits of opting out of IACs on internet discussion boards for Uber drivers, driven in large part by media coverage of O’Connor v. Uber.152 However, the substance of these discussions suggests that even with the benefits of press accounts of O’Connor and district court vetting of Uber’s communications about its IAC, some drivers still feel they lack sufficient information. Drivers participating in these discussions are often aware that the choice whether to opt out of an IAC is important, but have questions about the best course of action and the process for opting out. However, because the opt-out period is relatively short, drivers who do not receive timely and accurate information may nonetheless end up sleeping on their rights and recreating the status quo.

A second possibility for collective action that suffers from fewer coordination problems involves drivers simply taking gig economy companies up on their offer to pay arbitral forum costs, and filing lots of arbitration demands, forcing the companies to advance forum fees that are likely to approach the amounts that drivers could potentially recoup on the merits. This strategy is not unknown to plaintiffs’ lawyers; for example, Professor Martin Malin offers this first-hand account of events that followed a court decision enforcing an IAC in an employment contract in the “traditional” economy:

The employee then filed his arbitration demand individually . . . Over the next few months, approximately forty current and former employees filed similar arbitration demands . . . AAA rules require employers to pay all arbitrator fees . . . I estimate that, for the more than forty individual arbitrations, the employer had to deposit between $500,000 and $1,000,000 in up-front arbitration fees. The parties reached a global settlement resolving all of the claims.153

It may be that gig economy employers would respond to such a tactic by fighting harder to justify forum cost-splitting between themselves and workers who file for arbitration. Indeed, cost-splitting would be consistent with the enterprises’ views that the workers are operating their own small businesses.154 However, as discussed above, this tactic might trigger the effective vindication doctrine, and send the cases back to court anyway. Thus, despite obvious practical hurdles associated with this strategy, it should remain in worker advocates’ arsenals.

Conclusion

The gig economy offers an important opportunity to grapple with the effects of IACs on workers’ and consumers’ low-value claims. So far, the results are troubling: while it is too early to say what is happening to drivers who pursue arbitration, it is apparent that IACs are impeding the development of answers to questions about drivers’ employment status, and significantly reducing the value of workers’ claims in litigation. But all is not lost: agencies, states, and workers themselves may be able to bring pressure to bear on IACs, at least partially restoring the ideal of a “day in court.”

• 3Jean R. Sternlight, Creeping Mandatory Arbitration: Is it Just?, 57 Stan. L. Rev. 1631, 1635 (2005).
• 4Id. at 1648–49.
• 5Jeff Sovern et al., “Whimsy Little Contracts” With Unexpected Consequences: An Empirical Analysis of Consumer Understanding of Arbitration Agreements, 75 Md. L. Rev. 1, 4 (2015) (stating that “[m]any [consumers] expect to have access to the judicial system and class actions regardless of what they sign”).
• 6See Cynthia Estlund, Just the Facts: The Case for Workplace Transparency, 63 Stan. L. Rev. 351, 388 (2011) (“Asking about mandatory arbitration may signal that one could imagine suing the employer someday.”).
• 7The repeat player phenomenon, for which there is some empirical evidence, posits that a defendant who arbitrates repeatedly may be able to achieve certain advantages over plaintiffs; these range from greater expertise in navigating the arbitral forum, to the risk that arbitrators may skew their results toward repeat players in order to maintain “acceptability” to those enterprises. Sternlight, supra note 3, at 1650–51; see, e.g., Lisa B. Bingham, Employment Arbitration: The Repeat Player Effect, 1 Emp. Rts. & Emp. Pol’y J. 189, 191, 208–12 (1997) (finding evidence of repeat player effect); Alexander Colvin, An Empirical Study of Employment Arbitration: Case Outcomes and Processes 18–20 (2011), http://www.employmentandlaborinsider.com/wp-content/uploads/sites/328/2014/05/Blog.5.16.14.Cornell-Study-on-Arbitration.pdf, https://perma.cc/6KMP-LHKJ. The Colvin study found that “employers were more successful in cases involving repeat employer-arbitrator pairings.” Id. at 14, 22.
• 8Colvin, supra note 7, at 12 (concluding that “the time it takes to obtain a resolution after a hearing is about half as long in arbitration as in litigation”); see also Samuel Estreicher, Saturns for Rickshaws: The Stakes in the Debate Over Predispute Employment Arbitration Agreements, 16 Ohio St. J. on Disp. Resol. 559, 563 (2001) (arguing that “without employment arbitration as an available option, we would essentially have a ‘cadillac’ system for the few and a ‘rickshaw’ system for the many . . . a properly designed[] system . . . can do a better job of delivering accessible justice for average claimants”); Theodore J. St. Antoine, Mandatory Arbitration: Why It’s Better Than It Looks, 41 U. Mich. J.L. Ref. 783, 796 (2008) (concluding that “for most lower-paid workers, [arbitration] . . . may in fact be their only feasible option”).
• 9See George Padis, Arbitration Under Siege: Reforming Consumer and Employment Arbitration and Class Actions, 91 Tx. L. Rev. 665, 692 (2013).
• 10See Martin H. Malin, Privatizing Justice but by How Much? Questions Gilmer Did Not Answer, 16 Ohio St. J. Disp. Resol. 589, 609 (2001).
• 11Jean R. Sternlight, In Defense of Mandatory Binding Arbitration (If Imposed on the Company), 8 Nev. L.J. 82, 88 (2007) (discussing confidentiality of arbitration awards and noting that “arbitration providers have for the most part not been willing to open their files to researchers”).
• 12See Colvin, supra note 7, at 1, 9–10 (finding an employee arbitration win rate of 21.4%; median award amount of $36,500; and mean award of $23,548, and noting that these rates are significantly lower than those found by prior, more limited studies).
• 13St. Antoine, supra note 8, at 793 (reviewing studies of employment arbitration awards).
• 14Judith Resnik, Diffusing Disputes: The Public in the Private of Arbitration, the Private in Courts, and the Erasure of Rights, 124 Yale L. J. 2804, 2853–54 (2015) (discussing multifaceted confidentiality of arbitration processes, in which arbitrators, clients, and awards are all kept confidential).
• 15Id.; Arthur R. Miller, The Preservation and Rejuvenation of Aggregate Litigation: A Systematic Imperative, 64 Emory L.J. 293, 301 (2014) (arguing that “[t]he invisible character of arbitration results in far less deterrent effect than does the public nature of class litigation, which often is accompanied by media attention”).
• 16Margaret Jane Radin, Boilerplate: The Fine Print, Vanishing Rights, and the Rule of Law (2013).
• 17Id. at 33; see also David Horton, Mass Arbitration and Democratic Legitimacy, 85 Colo. L. Rev. 459, 464 (2014) (arguing that “the pervasiveness of arbitration clauses in consumer and employment contracts vividly illustrates the wear on democratic ideals that Radin describes”).
• 18Brian T. Fitzpatrick, The End of Class Actions?, 57 Ariz. L. Rev. 161, 194­–96 (2015).
• 19Myriam Gilles, Class Warfare: The Disappearance of Low-Income Litigants from the Civil Docket, 65 Emory L.J. 1531, 1561 (2016).
• 20E.g., Miller, supra note 15; Judith Resnik, Whither and Whether Adjudication?, 86 B.U. L. Rev. 1101, 1103 (2006).
• 21563 U.S. 333 (2011).
• 22Id. at 340, 352 (describing California rule).
• 23133 S.Ct. 2304 (2013).
• 24Id. at 2310–11.
• 25Martin H. Malin, The Three Phases of the Supreme Court’s Arbitration Jurisprudence: Empowering the Already-Empowered, 17 Nev. L.J. 23, 59–60 (2017).
• 26The Italian Colors Court left open a small window to invalidate IACs when the forum costs—rather than the costs of proving one’s case—make arbitration cost-prohibitive. Italian Colors, 133 S.Ct. at 2310–11 (Court’s “effective vindication” doctrine “would perhaps cover filing and administrative fees attached to arbitration that are so high as to make access to the forum impossible”); see also Green Tree Financial Corp.-Ala. v. Randolph, 531 U.S. 79, 90 (2000). For reasons discussed below, this exception is unlikely to be relevant for workers who have signed the agreements discussed in this article.
• 27Fitzpatrick, supra note 18, at 163.
• 28Resnik, supra note 14, at 2872; see also Sternlight, supra note 3, at 1639–40 (“With respect to employment, while the percentage of employees required to arbitrate future disputes is probably lower than one-third, it is rising.”).
• 29Jean R. Sternlight, Disarming Employees: How American Employers Are Using Mandatory Arbitration to Deprive Workers of Legal Protection, 80 Brook. L. Rev. 1309, 1310 (discussing survey of general counsel finding that use of IACs more than doubled, from 16.1% in 2012 to 42.7% in 2014).
• 30Resnik, supra note 14, at 2907. Consumers are in a similar—and probably worse—situation. Predictably, “public records indicate that almost no individual consumers use arbitration.” Id. at 2900.
• 31Berwick v. Uber Techs. Inc., No. 11-47739EK (Labor Comm’r of Cal. 2015).
• 32See Colvin, supra note 7, at 24 (“One of the possible benefits of employment arbitration is that the relative[] simplicity of the forum might make self-representation by employees more plausible than in litigation.”).
• 33Id. at 17 (finding that “employment arbitration appears to be a dispute resolution system predominantly based on employee representation by counsel, as is the case with litigation”); see also Martin H. Malin & Jon M. Werner, 14 Penn Plaza LLC v. Pyett: Oppression or Opportunity for U.S. Workers; Learning From Canada, 2017 U. Chi. L. Forum 347 (2017) (finding that discrimination claimants were more likely to succeed in grievance arbitration than in proceedings before the Human Rights Tribunal of Ontario, and that this difference was attributable to fact that claimants were represented by attorneys in arbitration).
• 34These were agreements that I was able to locate on enterprise websites or in litigation dockets. In addition to these methods, I emailed some enterprises to request copies of their agreements, but without success. As I discuss below, this means that the set of agreements I analyze are not a random sample. However, I do include agreements drafted by many of the major players in the platform economy, including Uber.
• 3582 F.Supp.3d 1133 (N.D.Cal. 2015).
• 36Aaron Smith, Shared, Collaborative and On Demand: The New Digital Economy, Pew Research Center 6 (2016), http://www.pewinternet.org/2016/05/19/the-new-digital-econ
  omy/, https://perma.cc/5FZ7-JJQ8.
• 37The exceptions were Maplebear (Instacart), and Mechanical Turk. However, the presence of an arbitration clause made the inclusion of a specific waiver of the right to litigate on a class or collective basis mostly superfluous; such a clause would have become relevant only in the event that a court struck down the arbitration clause itself.
• 38Only one agreement did not contain an explicit individual arbitration clause, and in subsequent litigation, a district court stated that because class arbitration must be explicitly authorized, the mandatory arbitration clause alone was enough to preclude arbitration on a class or collective basis. Cobarruviaz v. Maplebear, Inc., 143 F.Supp.3d 930 (N.D. Cal. Nov. 3, 2015).
• 39The exceptions were GrubHub, Caviar, Uber (Dec. 11, 2015 agreement), Handy, DoorDash, and Mechanical Turk.
• 40Most agreements include separate severability clauses as to the individual arbitration clauses themselves; the list of agreements including only general severability clauses covering the whole agreement includes Maplebear (Instacart), Caviar, and Mechanical Turk.
• 41E.g., Bynum v. Maplebear, 160 F.Supp. 3d 527, 537 (E.D.N.Y. 2016) (noting that the venue clause was invalid, and severing it); Levin v. Caviar, Inc., 146 F.Supp.3d 1146, 1155 (N.D. Cal. 2015) (severing unconscionable Private Attorney General Act (“PAGA”) waiver).
• 42Bynum, 160 F.Supp. 3d at 537; Levin, 146 F.Supp.3d at 1155.
• 43Cf. Sovern et al., supra note 5.
• 44Most agreements are explicit that any class or collective action must proceed in court instead of arbitration, including each Uber agreement I reviewed, as well as the agreements of Lyft, Postmates, Handy, and DoorDash. Alternatively, the GrubHub agreement states that if the class waiver is deemed unenforceable, “the Arbitration provision is otherwise silent as to any party’s ability to bring a class, collective, or representative action in arbitration.” Because class or collective arbitration must be specifically authorized, this approach is also very likely to result in any class or collective action returning to court.
• 45Consumer Fin. Prot. Bureau, Arbitration Study: Report to Congress, Pursuant to Dodd-Frank Wall Street Reform and Consumer Protection Act § 1028(a) 10 (2015), http://files.consumerfinance.gov/f/201503_cfpb_arbitration-study-report-to-congress-2015.pdf, https://perma.cc/KJ82-U5XQ.
• 469 U.S.C. § 10(a) (2002) (District courts may vacate arbitration awards when “the award was procured by corruption, fraud, or undue means; where there was evident partiality or corruption in the arbitrators . . . where the arbitrators were guilty of misconduct . . . by which the rights of any party have been prejudiced; or where the arbitrators exceeded their powers.”).
• 47Hall St. Assocs. v. Mattel, Inc., 552 U.S. 567, 586 (2008) (holding that FAA enumerates exclusive grounds for overturning an arbitration award).
• 48See Yes, You Can Appeal an Arbitration Award, Law360 (Jan. 28, 2015), http://www.law360.
  com/articles/614689/yes-you-can-appeal-an-arbitration-award, https://perma.cc/6R4Z-LM93 (detailing requirements of arbitration services for appellate review of arbitrator’s award within the arbitral forum).
• 49The Taskrabbit and Handybook agreements state that the arbitrator must follow the law, and that his award “can be challenged” if he fails to do so. These agreements do not state where that challenge will occur, but the Mattel rule makes the arbitral forum the only possibility. The Maplebear (Instacart) agreement purports to allow appeals to California state courts, despite the Mattel rule.
• 50Id.
• 51AT&T Mobility LLC v. Concepcion, 563 U.S. 333, 350 (2011) (“[W]hen damages allegedly owed to tens of thousands of potential claimants are aggregated and decided at once, the risk of an error will often become unacceptable.”).
• 52Agreements that require negotiation include TaskRabbit, Postmates, and Handy/
  Handybook.
• 53Lyft’s agreement calls for the driver to pay forum fees in an amount equal to the cost of judicial filing fees, but limits that amount to $50 when the driver both participates in negotiation and has a claim worth less than $5,000.
• 54See agreements of Maplebear (Instacart) (sufficient discovery to satisfy due process); GrubHub (sufficient discovery); Uber (all agreements) (adequate discovery); Lyft (reasonable discovery); Handy (arbitrator may allow discovery, taking into account efficient process); DoorDash (arbitrator may allow discovery, taking into account efficient process).
• 55Caviar agreement.
• 56Agreements of TaskRabbit, Lyft, Handy, and Handybook.
• 57That the agreements do not purport to waive workers’ rights to appeal to administrative agencies is consistent with governing law; private enterprises and individuals may not impede agencies’ ability to enforce the law through pre-dispute agreements. See EEOC v. Frank’s Nursery & Crafts, Inc., 177 F.3d 448, 460 (6th Cir. 1999) (reasoning that employee’s complaint to EEOC did not breach arbitration agreement, and employee could not bind EEOC to arbitrate claims against employer).
• 58See agreements of Maplebear (Instacart), Uber (all agreements), Postmates, Handy, and DoorDash (all stating that drivers may bring claims before administrative tribunals, but preserving failure to exhaust defense); see also GrubHub agreement (stating only that drivers may bring claims and be awarded damages in administrative agencies); Lyft (stating that drivers may not receive damage awards through agency proceedings).
• 59Agreements of Maplebear (Instacart), TaskRabbit, and Mechanical Turk.
• 60Bynum v. Maplebear, 160 F. Supp. 3d 527 (E.D.N.Y. 2016) (venue in San Francisco invalid as to plaintiff who lived in New York); cf. Cobarruviaz v. Maplebear, Inc., 143 F. Supp. 3d 930, 941–42 (N.D. Cal. 2015) (venue in San Francisco enforceable as to plaintiff who lived in San Francisco).
• 61Agreements of Postmates, Handy, and DoorDash.
• 62Agreements of TaskRabbit (limiting the guarantee to claims not exceeding $10,000); Handy.
• 63See agreements of Caviar, Uber (Dec. 11, 2015); Lyft (providing that drivers must pay up to the amount of a court filing fee if they refuse to engage in pre-arbitral negotiate with the company).
• 64The Uber (Dec. 22, 2015) agreement limited drivers’ contributions to expenses they would have had to bear in court.
• 65133 S.Ct. at 2310–11 (2013).
• 66Mohamed v. Uber Techs., Inc., 836 F.3d 1102, 1112 (9th Cir. 2016) (stating that “Uber has committed to paying the full costs of arbitration. So long as Uber abides by this commitment, the fee term in the arbitration agreement presents Plaintiffs with no obstacle to pursuing vindication of their federal statutory rights in arbitration. As a result, we decline to reach the question of whether the fee term would run afoul of the effective vindication doctrine if it were enforced as written”); Bynum¸ 160 F.Supp. at 538 (quoting counsel for Instacart, who agreed to waive fee splitting provisions, stating, “I believe they’re invalid”).
• 67Private Attorney General Act of 2004, Cal. Lab. Code §§ 2698–2699.5.
• 68Agreements of TaskRabbit, Uber (all agreements), Lyft, Handy.
• 69Agreements of Maplebear (Instacart), Grubhub (June 2014 agreement), Caviar, Postmates, MTurk, and Handybook.
• 70Postmates agreement.
• 71See, e.g., Levin v. Caviar, Inc., 146 F. Supp. 3d 1146 (N.D. Cal. 2015) (Plaintiff agreed to terms of service via a screen stating “[b]y marking yourself available you agree to Caviar’s Courier terms of service,” with a hyperlink to the terms.).
• 72Compare Handybook agreement of May 31, 2011 with Handy agreement; Declaration of Stan Chia in Support of Defendants’ Motion to Deny Class Certification, Tan v. GrubHub, 171 F. Supp. 3d 998 (N.D. Cal. June 2, 2016) (No. 3:15-cv-05128-JSC) (stating that, beginning in July 2015, delivery partner agreements provided the right to opt out of the IAC).
• 73See O’Connor v. Uber Techs., Inc., No. C-13-3826 EMC, 2013 WL 6407583 (N.D. Cal. Dec. 6, 2013) (“Uber drivers must be given . . . reasonable means of opting out of the arbitration provision within 30 days of the notice.”).
• 74Green Tree Fin. Corp.-Ala. v. Randolph, 531 U.S. 79, 91 (2000).
• 75AT&T Mobility LLC v. Concepcion, 563 U.S. 333, 339 (2011) (The FAA “permits arbitration agreements to be declared unenforceable ‘upon such grounds as exist at law or in equity for the revocation of any contract.’ This saving clause permits agreements to arbitrate to be invalidated by ‘generally applicable contract defenses, such as fraud, duress, or unconscionability,’ but not by defenses that apply only to arbitration or that derive their meaning from the fact that an agreement to arbitrate is at issue.”)
• 76See Pokorny v. Quixtar, Inc., 601 F.3d 987, 996 (9th Cir. 2010).
• 77See Armendariz v. Found. Health Psychcare Servs., Inc., 6 P.3d 669, 689 (Cal. 2000) (adhesion contract is oppressive, as required under California’s test for procedural unconscionability, which focuses on oppression and surprise); see also Levin v. Caviar, Inc., 146 F.Supp.3d 1146, 1158 (N.D. Cal. 2015) (“The arbitration agreement, which is clearly a contract of adhesion, is procedurally unconscionable.”).
• 78See, e.g., Mohamed v. Uber Techs., 836 F.3d 1102, 1111 (9th Cir. 2016) (finding that even onerous opt out process requiring overnight delivery is enough to support conclusion that IAC was not adhesive, and therefore not unconscionable); Sena v. Uber Techs., Inc., No. CV-15-02418-PHX-DLR, 2016 WL 1376445, at *6 (D. Ariz. Apr. 7, 2016) (“[B]ecause Sena was not required to accept the Arbitration Provision, the Delegation Clause is not procedurally unconscionable.”); Micheletti v. Uber Techs. Inc., No. 15-1001, 2016 WL 5793799 (W.D. TX. Oct. 3, 2016) (delegation clause enforceable because IAC gave drivers opportunity to opt out); Varon v. Uber Techs., Inc., No. 15-3650, 2016 WL 3917213 (D. Md. July 20, 2016) (same).
• 79Lee v. Uber Techs., Inc., 208 F. Supp. 3d 886, 892–893 (N.D. Ill. 2016) (granting motion to compel arbitration as to whether IAC was unconscionable).
• 80Tan v. Grubhub, No. 3:15-cv-05128-JSC, 2016 WL 4721439, at *2 (N.D. Cal. Jul. 19, 2016).
• 81Gillette v. Uber Techs., No. C-14-5241 EMC, 2015 WL 4481706, at *4 (N.D. Cal. July 22, 2015).
• 82Mohamed v. Uber Techs., Inc., 836 F.3d 1102, 1111 (9th Cir. 2016) (writing that “an arbitration agreement is not adhesive if there is an opportunity to opt out” and rejecting the district court’s conclusion that the opt-out provision was too difficult to drivers to find and use).
• 83Agreements of Mechanical Turk, Handybook, and TaskRabbit; see also Resnick, supra note 14.
• 84Agreements of Uber (all agreements), Lyft.
• 85Agreements of TaskRabbit, Uber (UberBlack/UberSUV agreements of Jul. 24, 2013 & June 21, 2014), Lyft, Mechanical Turk.
• 86Agreements of Uber (Dec. 11, 2015); Handy.
• 87See, e.g., O’Connor v. Uber Techs., Inc., 201 F. Supp. 3d 1110 (N.D. Cal. 2016); Tan v. GrubHub, Inc., 171 F. Supp. 3d 998 (N.D. Cal. 2016); Cotter v. Lyft, 193 F. Supp. 3d 1030 (N.D. Cal. 2016).
• 88193 F. Supp. 3d 1030 (N.D. Cal. 2016) (Cotter II).
• 89201 F. Supp. 3d 1110 (N.D. Cal. 2016).
• 90In O’Connor v. Uber, the proposed settlement initially called for class counsel to seek an award of up to twenty-five percent of the settlement fund; that fund was at least $84 million, and could have grown to $100 million. However, class counsel later agreed to reduce her fee award by $10 million. Class Action Settlement and Release, O’Connor v. Uber, No. C-13-cv-3826 EMC, at *35 (Apr. 21, 2016) (“Class Counsel agrees not to seek a Fee and Expense Award from the Court in excess of twenty-five percent of the Settlement Fund.”); O’Connor v. Uber Techs., Inc., 201 F. Supp. 3d 1110 (2016); Notice of Motion and Memorandum of Points and Authorities in Support of Plaintiffs’ Motion for Preliminary Approval of Revised Class Action Settlement, Cotter v. Lyft, No. 13–cv–04065–VC, at *1 n.1, *4 n.5 (May 11, 2016) (stating plaintiffs’ counsel would not seek additional fees based on re-negotiation of settlement that more than doubled settlement fund, and that original fee request was $3.675m).
• 91E.g., Mohamed v. Uber Techs., 836 F.3d 1102 (9th Cir. 2016); Bynum v. Maplebear, 160 F. Supp. 3d 527 (E.D.N.Y. 2016) (compelling arbitration event though venue and fee splitting provisions were unenforceable); Moton v. Maplebear, No. 15 Civ. 8879 (CM), 2016 WL 616343 (S.D.N.Y. Feb. 9, 2016) (also compelling arbitration event though venue and fee splitting provisions were unenforceable); Sena v. Uber Techs., Inc., No. CV-15-02418-PHX-DLR, 2016 WL 4064584 (D. Ariz. May 3, 2016) (denying motion for reconsideration of order compelling arbitration); Suarez v. Uber Techs., Inc., No. 8:16-cv-166-T-30MAP, 2016 WL 2348706 (M.D. Fla. May 4, 2016) (granting motion to compel arbitration); Varon v. Uber Techs., Inc., No. MJG-15-3650, 2016 WL 1752835 (D. Md. May 3, 2016) (granting motion to compel arbitration); Lee v. Uber Techs., Inc., 208 F. Supp. 3d 886, (N.D. Ill. 2016) (granting motion to compel arbitration); cf. Razak v. Uber Techs., Inc., No. 16-573, 2016 WL 5874822 (E.D. Pa. Oct. 7, 2016) (denying motion to compel arbitration where plaintiffs had opted out).
• 92Cotter v. Lyft, 176 F.Supp.3d 930, 932–33 (N.D. Cal. 2016) (Cotter I).
• 93Id. at 933.
• 94Id.
• 95Id. at 934.
• 96Id. at 934, 937. The non-monetary relief involved Lyft agreeing to limit the grounds on which it would terminate drivers and the creation of an internal appeals process for terminated drivers.
• 97Id. at 934.
• 98Id. 176 F. Supp. 3d 930, 940 (N.D. Cal. 2016). (The court explained that the parties calculated the value of drivers’ claims only through June 2015, rather than through the date of potential settlement approval, likely to occur around spring 2016. Thus, “counsel thought they were getting their clients a settlement that was roughly 17.36% of the maximum value of the reimbursement claim. In fact, they got their clients a settlement that was at most only 8.82% of the reimbursement claim.”).
• 99Id. The significance of the failed attempt to settle the PAGA claim is discussed in greater detail below.
• 100Cotter II, 193 F. Supp. 3d 1030, 1033 (N.D. Cal. 2016); Order, Cotter v. Lyft, No. 13-cv-4065-VC (July 1, 2016) (setting fairness hearing).
• 101Cotter I, 193 F. Supp. 3d at 942–44.
• 102Id. at 944.
• 103Id.; see also infra Part II.A.
• 104O’Connor v. Uber Techs., Inc., 201 F. Supp. 3d 1110, 1113–14 (N.D. Cal. 2016).
• 105Id. at 1114.
• 106Id. at 1113–16 (discussing O’Connor’s procedural history).
• 107Id.
• 108Id. at 1116.
• 109Id. at 1128.
• 110Id.
• 111Id. at 1123.
• 112Mohamed v. Uber Techs., Inc., 836 F.3d 1102 (9th Cir. 2016).
• 113O’Connor, 201 F. Supp. 3d at 1135.
• 114See generally Malin, supra note 25; Burt Neuborne, Ending Lochner Lite, 50 Harv. C.R.-C.L. L. Rev. 183, 203 (2015) (discussing prospects for FAA revision).
• 115Radin, supra note 16, at 224. Other commentators have been similarly pessimistic about the possibility for bodies other than Congress to deter or eliminate mandatory individual arbitration. See, e.g., Javier J. Castro, Employment Arbitration Reform: Preserving the Right to Class Proceedings in Workplace Disputes, 48 U. Mich. J.L. Reform 241, 243 (2014) (arguing that “Congress should enact a statutory amendment to the FAA that bans enforcement of class waiver provisions in mandatory arbitration agreements, which prevent employees from aggregating their claims in any forum”); Fitzpatrick, supra note 18, at 190 (after reviewing possibilities, stating that “there is little in either state or federal law that will stop businesses from taking advantage of the opportunity to bind consumers, employees, and even shareholders to class action waivers in arbitration clauses”).
• 11629 U.S.C. § 157 (2012).
• 11729 U.S.C. § 103 (2012). For a more detailed discussion of the argument that IACs are inconsistent with these statutes, see Ann C. Hodges, Can Compulsory Arbitration be Reconciled With Section 7 Rights?, 38 Wake Forest L. Rev. 173 (2003); Charles A. Sullivan & Timothy P. Glynn, Horton Hatches the Egg: Concerted Action Includes Concerted Dispute Resolution, 64 Ala. L. Rev. 1013 (2013).
• 118D.R. Horton, 375 N.L.R.B. 2277 (2012). The D.R. Horton decision was issued in part by NLRB members whose recess appointments the Supreme Court later found were invalid. NLRB v. Noel Canning, 134 S. Ct. 2550 (2014). The Board later reaffirmed the D.R. Horton rule in Murphy Oil, 361 N.L.R.B. 774 (2014), on similar reasoning.
• 119D.R. Horton, 375 N.L.R.B. at 2278–79 (citations omitted).
• 120Id. at 2280 (citing J.I. Case Co. v. NLRB, 321 U.S. 332 (1944)).
• 121Id. at 2285.
• 122Lewis v. Epic Sys., 823 F.3d 1147, 1161 (7th Cir. 2016).
• 123Morris v. Ernst & Young, LLP, 834 F.3d 975 (9th Cir. 2016).
• 124Id. at 981 (“The intent of Congress is clear from the statute and is consistent with the Board’s interpretation.”); Lewis, 823 F.3d at 1154 (“Congress was aware of class, representative, and collective legal proceedings when it enacted the NLRA. The plain language of Section 7 encompasses them, and there is no evidence that Congress intended them to be excluded.”). It is worth noting that, despite the high-profile nature of this issue, no judge of the Seventh Circuit voted to rehear the case en banc. Id. at 1157.
• 125D.R. Horton, Inc. v. NLRB, 737 F.3d 344, 357, 359 (5th Cir. 2013).
• 126Id. at 359.
• 127Owen v. Bristol Care, Inc., 702 F.3d 1050 (8th Cir. 2013).
• 128Sutherland v. Ernst & Young, LLP, 726 F.3d 290, 297 n.8 (2d Cir. 2013).
• 129Mohamed v. Uber Techs., 836 F.3d 1102 (9th Cir. 2016).
• 130Id. at 1112 n.6.
• 131Plaintiff-Appellees’ Consolidated Answering Brief, O’Connor v. Uber Techs., Inc., No. 15-17420 (9th Cir. Oct. 12, 2016).
• 132755 F.3d 1072, 1075 (9th Cir. 2014).
• 133Id.
• 134On Assignment Staffing Services, Inc., 362 N.L.R.B. 189, at *6 (2015) (“Regardless of the procedures required, the fact that employees must take any steps to preserve their Section 7 rights burdens the exercise of those rights.”). On Assignment Staffing was reversed by the Fifth Circuit, but under longstanding NLRB practice, the Board will continue to apply the rule in other cases, unless the rule is rejected by the Supreme Court or changed by the Board itself. On Assignment Staffing Services v. NLRB, No. 15-60642, 2016 WL 3685206 (5th Cir. June 6, 2016).
• 13529 U.S.C. § 152(3) (2012) (employee shall not include “any individual having the status of independent contractor).
• 136Compare S.G. Borello & Sons, Inc., v. Dep’t of Ind. Relations, 769 P.2d 399 (Cal. 1989), with FedEx Home Delivery, 361 N.L.R.B. 55 (2014).
• 137O’Connor v. Uber Techs., Inc., 82 F. Supp. 3d 1133, 1148, 1153 53 (N.D. Cal. 2015) (holding that neither plaintiffs nor Uber were entitled to summary judgment and that a jury should decide the ultimate question of whether drivers were employees); Cotter v. Lyft, 60 F. Supp. 3d 1067, 1081 (N.D. Cal. 2015) (“[T]he jury in this case will be handed a square peg and asked to choose between two round holes.”).
• 138See Smith, supra note 32 (indicating that sixty-six percent of survey respondents believed that gig economy workers were independent contractors).
• 139Heather Somerville, Former Uber Driver Was an Employee, Rules California Department, Reuters (Sept. 8, 2016), http://www.reuters.com/article/uber-tech-california-ruling-idUSL1N11F1KT20150910, https://perma.cc/78MG-LEJP.
• 140Raiser LLC v. State of Florida Dep’t of Econ. Opportunity, Protest of Liability, No. 0026 2825 90-02 (Dec. 3, 2015).
• 141See Doe v. Uber, Cal. Unemployment Ins. Appeals Bd., Case No. 5371509 (June 1, 2015); Uber Tech. Inc. v. Berwick, Labor Commissioner, State of CA, No. 11-46739 (June 3, 2015); Noam Scheiber, Uber Drivers Ruled Eligible for Jobless Payments in New York State, N.Y. Times (Oct. 12, 2016), http://www.nytimes.com/2016/10/13/business/state-rules-2-former-uber-drivers-eligi
  ble-for-jobless-payments.html?_r=1, https://perma.cc/Z79C-VBF7.
• 142Iskanian v. CLS Transp. L.A., LLC, 327 P.3d 129, 146–47 (Cal. 2014). The Ninth Circuit has upheld the Iskanian rule, rejecting an argument that it is preempted by the FAA. Sakkab v. Luxottica Retail N. Am., Inc., 803 F.3d 425, 434, 436, 438 (9th Cir. 2015); see also Mohamed v. Uber Techs., 836 F.3d 1102, 1113 (9th Cir. 2016) (agreeing that “the PAGA waiver in [Uber’s 2013] agreement was invalid under California law”). This rule is further consistent with Italian Colors, in which the Court held that “a provision in an arbitration agreement forbidding the assertion of certain statutory rights” would be invalid. American Exp. Co. v. Italian Colors Restaurant, 133 S. Ct. 2304, 2310 (2013). The unlawful PAGA waiver that led the O’Connor district court to invalidate IACs covering the large majority of Uber drivers ultimately included in the O’Connor class, although the Mohamed court held that the unlawful PAGA waiver was severable.
• 143See, e.g., Zenilaj v. Handybook, 82 F. Supp. 3d 968 (N.D. Cal. 2015); Levin v. Caviar, Inc., 146 F. Supp. 3d 1146 (N.D. Cal. 2015)
• 144See supra note 68 and accompanying text.
• 145Cal. Labor Code § 2699(f )(2).
• 146Cal. Labor Code § 2699(i).
• 147O’Connor v. Uber Techs., Inc., 201 F. Supp. 3d 1110, 1133 (N.D. Cal. 2016).
• 148See, e.g., Owen M. Fiss, Against Settlement, 93 Yale L.J. 1073 (1984); Geoffrey C. Hazard, Jr., The Settlement Black Box, 75 B.U. L. Rev. 1257 (1995); Kenneth R. Feinberg, Reexamining the Arguments in Owen M. Fiss, Against Settlement, 78 Fordham L. Rev. 1171 (2009).
• 149See Cal. Labor Code § 2699(e)(2) (providing that a court may award less than the maximum civil penalty when “to do otherwise would result in an award that is unjust, arbitrary and oppressive, or confiscatory”).
• 150See Radin, supra note 16, at 243 (discussing the role of non-governmental organizations in educating consumers about contract terms).
• 151See Tan v. Grubhub, No. 15-cv-05128-JSC, 2016 WL 4721439 (N.D. Cal. Jul. 19, 2016).
• 152See, e.g., Drivers’ Last Chance to Opt-out of Binding Arbitration, UberPeople.Net, http://uberpeople.net/threads/drivers-last-chance-to-opt-out-of-binding-arbitration.11099/, https://perma.cc/85AJ-CF78 (last visited Nov. 12, 2016); Harry Campbell, Should Drivers Opt Out of Uber’s New Driver-Partner Agreement?, The Rideshare Guy (Dec. 11, 2015), http://therideshareguy.com/should-drivers-opt-out-of-ubers-new-driver-partner-agreement/, https://perma.cc/78AS-WVMF.
• 153Martin H. Malin, The Employment Decisions of the Supreme Court’s 2012–13 Term, 29 ABA J. Lab. & Emp. Law 203, 213–14 (2014).
• 154Compare AAA Commercial Arbitration Rules, Rule R-53 (Administrative Fees), https://www.adr.org/aaa/ShowProperty?nodeId=/UCM/ADRSTG_004103, https://perma.cc/9NEU-GZHZ (party making claim responsible for paying administrative fees, subject to later apportionment) with AAA Employment Arbitration Rules, https://www.adr.org/aaa/ShowProperty?nodeId=/UCM/ADRSTG_004362, https://perma.cc/52HG-NBUG (employer responsible for paying most arbitral forum costs).