ORDER DENYING MOTION FOR PRELIMINARY APPROVAL OF CLASS ACTION SETTLEMENT
Re: Dkt. No. 169
VINCE CHHABRIA, United States District JudgeThe plaintiffs in this case — three people who have worked as Lyft drivers — move for preliminary approval of a class settlement agreement they reached with Lyft. Five other Lyft drivers, along with the Teamsters — a labor, union that is seeking to organize Lyft drivers — have filed objections to the agreement, and urge the Court to deny the motion for preliminary approval.1
The motion for preliminary approval is denied because the settlement agreement does not fall within the range of reasonableness. Most glaringly, counsel for the plaintiffs pegged the $12.25 million settlement figure primarily to the estimated value of the drivers’ claim for mileage reimbursement. But the lawyers estimated the value of the reimbursement claim to be $64 million, ‘when in fact, using their own methodology, it is worth more than $126 million. The drivers were therefore short*932changed by half on their reimbursement claim alone. Moreover, counsel’s treatment of the drivers’. claim for penalties under the Private Attorneys General Act (“PAGA”) was arbitrary, and may have shortchanged the State of California (not to mention the drivers) even more. The modest nonmonetary relief set forth in the agreement does not come close to making up for these serious defects in the monetary aspect of the settlement.
However, many of the Teamsters’ objections to the agreement are not well-founded. In particular, the Teamsters argue that the Court should not merely reject this agreement, but any agreement that fails to reclassify Lyft drivers from “independent contractors” to “employees.” The Teamsters’ position is based largely on policy arguments better made to the legislative and executive branches. And it disregards the risks the drivers would face if they took their case to trial. Accordingly, if the parties wish to negotiate a new agreement that addresses the defects identified in this ruling, the Court would, at least on the current record, preliminarily approve that agreement even if it fell short of requiring Lyft to classify its drivers as employees.
I. BACKGROUND
Patrick Cotter and Alejandra Maciel used to drive for Lyft in California. They filed this suit contending that Lyft violates California law by classifying its drivers as “independent contractors” rather . than “employees.” This distinction matters a great deal, because under California law, employees get a number of benefits and protections that independent contractors don’t. For example, California law guarantees employees a minimum wage, extra pay for working overtime, and workers’ compensation benefits. The original theory behind this distinction, in large part, was that independent contractors don’t need those kinds of across-the-board safeguards, because their special skills give them bargaining power and the ability to negotiate their own set of contractual benefits and protections. Employees, on the other hand, need a minimum floor of legal safeguards to prevent employers from taking undue advantage of their inferior bargaining position.
As applied to a case involving drivers, perhaps the most significant legal protection enjoyed by “employees” but not “independent contractors” is the right to be reimbursed for expenses incurred in performing the work. The law prevents companies from passing on to employees the cost of doing business. Lyft drivers spend a lot of money on gas, and their vehicles undergo significant wear and tear. If the drivers are employees, the law would require Lyft to reimburse them for these expenses. If the drivers are independent contractors, they would be left to negotiate their own terms with Lyft regarding reimbursement (if, that is, they have any negotiating power to speak of).
Cotter and Maciel brought their suit as a proposed class action. They originally sought to represent everyone throughout the country who has ever driven for Lyft since the company’s inception in May of 2012. However, California’s wage and hour laws don’t apply to people who work exclusively in other states, so the Court ruled that the plaintiffs may only seek to represent drivers who have worked for Lyft in California. Dkt. No. 51; Cotter v. Lyft, Inc., 60 F.Supp.3d 1059 (N.D.Cal.2014). On behalf of that proposed class of California Lyft drivers, Cotter and Maciel seek injunctive relief — -namely, a court order requiring Lyft to classify its drivers as employees, and to provide the drivers with all the benefits and protections that California law confers upon employees. Cotter and Maciel also sought various forms of monetary relief for the drivers — primar*933ily reimbursement of expenses the drivers had incurred while working for Lyft, but also damages and penalties for things like failure to pay the minimum wage, failure to pay overtime, failure to provide paid meal and rest breaks, and failure to pass on tips received from riders.
The parties and the Court agreed on a case schedule that is somewhat different from the typical class action. In most class actions, the first important milestone is a motion by the named plaintiffs for class certification. In adjudicating a motion for class certification, a court must decide whether the case is appropriate for class action treatment, and whether the named plaintiffs are qualified to represent the class in the litigation. Then, assuming a class is certified, the case moves on to the “merits stage,” where the parties litigate (and the court or a jury decides) who wins. In this case, the parties agreed that the Court should first consider whether the named plaintiffs themselves, Cotter and Maciel, were employees or independent contractors as a matter of law. A legal determination that they were independent contractors would end the case, avoiding the need to put Lyft through costly and time-consuming class certification proceedings. A determination that Cotter and Maciel were employees would require the case to continue, with the Court deciding whether the case could proceed on a classwide basis (and if it could, with the Court simply entering judgment in favor of the whole class and against Lyft).
The parties thus filed cross-motions for summary judgment on the question whether Cotter and Maciel were employees or independent contractors as a matter of law. The Court denied both motions and ruled that, because there is no clear legal answer to this question, it would be for a jury to decide whether Cotter and Maciel were employees or independent eontrac-tors. Dkt. No. 94; Cotter v. Lyft, Inc., 60 F.Supp.3d 1067 (N.D.Cal.2015). The Court explained that although Cotter and Maciel were like independent contractors in some ways (for example, they could control their own schedules), they were like employees in other ways (for example, Lyft retained the right to control how Cotter and Maciel performed their jobs when they did choose to work, and Lyft retained the right to terminate them for any reason). In this ruling, the Court also flagged a key issue likely to affect a jury’s decision about how to classify a driver— whether he drove regularly and full time, or sporadically and part time. A person who drives regularly and full time, the Court noted, is likely relying on Lyft as his primary source of income, and therefore looks more like the type of person California’s wage and hour laws were designed to protect. A person who drives sporadically for Lyft, as one of many odd jobs for different clients, arguably looks more like an independent contractor.
Following this ruling, the parties decided to participate in a series of settlement conferences with a magistrate judge of this court. The Court referred them to Magistrate Judge Donna Ryu, who is an expert in wage and hour law. And the Court agreed to delay class certification proceedings while the parties negotiated. Meanwhile, the plaintiffs filed an amended complaint in which they added another driver, Jeffrey Knudtson, to the lawsuit as a named plaintiff. In contrast to Cotter and Maciel, who drove for Lyft sporadically and part time before being terminated, Knudtson continues to work for the company and alleges he drives full time.
II. THE SETTLEMENT AGREEMENT
After several settlement conferences, the parties reached an agreement, subject *934to approval by this Court. The agreement contemplates that' a class will be certified (consisting of all people who have driven for Lyft at least once in California), that notice of the settlement will be sent to the class, and that any class member who does not opt out of the settlement will be bound. The terms of the agreement are discussed more fully in Sections IV and V, but they are summarized briefly here.
• The agreement has a prospective component that would somewhat alter the relationship between Lyft and its drivers. Currently, Lyft preserves the right to terminate its drivers (that is, deactivate them from the Lyft platform) for any reason and at .any time. Under the proposed settlement, there would be a laundry list of restrictions and requirements imposed on drivers by contract, and Lyft would have the right to terminate drivers only for breaching the contract. In certain circumstances, a driver would be entitled to receive notice from Lyft of the termination decision and given an opportunity to respond. If Lyft is unsatisfied by the response and maintains its decision to terminate the driver, the driver could take the breach-of-contract dispute to arbitration, with Lyft paying all arbitration-specific fees. ■ And even in termination decisions where the contract does not assure a driver notice and an opportunity to respond, the driver may challenge the termination decision - in arbitration, again with Lyft paying all arbitration-specific fees.
There is also a monetary component to the agreement that would compensate drivers in connection with work they’ve already performed. Specifically, Lyft agrees to pay $12.25 million to a settlement fund, and drivers may collect from that fund based on the estimated number of miles they have 'driven in the past. As counsel for the plaintiffs explain, this $12.25 million- settlement figure was derived almost exclusively from their estimate- of the maximum amount for which Lyft would be liable to the drivers for the plaintiffs’ reimbursement claim. The maximum value plaintiffs’ counsel assigned to the reimbursement claim when they reached the agreement was $64 million.
The agreement also establishes two groups of drivers for settlement purposes: full-time drivers and part-time drivers. Full-time drivers are those who worked more than 30 hours per week at least half the time they worked for Lyft. Part-time drivers are those who worked less than this amount. Full-time drivers receive a higher per-mile settlement payment than part-time drivers, on the assumption that the full-time drivers have a stronger claim for employment status. The premium to be received by the full-time drivers is fifty percent. To use a simple example, if a part-time driver were eligible to collect $100 based on the number of miles he drove, the full-time driver would be eligible to collect $150 for driving that same number of miles. Of the roughly 150,000 proposed class members, only 755 are full-time drivers.
In exchange for this prospective and monetary relief, the class members agree to waive any. legal claims they may currently have against Lyft pertaining to their wages or their employment status. The waiver applies to claims that.may have arisen during the “class period,” which begins at Lyft’s inception in 2012 and ends on the date of preliminary approval of the class settlement by the Court.
The plaintiffs have now moved for -preliminary approval of the settlement. If the Court granted the motion for preliminary -approval, the next step would be for the drivers to receive notice of the settlement. They would then have an opportunity to object to the terms of the settlement, as well as to opt out of it. The Court would then hold a “fairness hearing” *935at which it would consider the ultimate question whether to give final approval to the agreement.
III. LEGAL STANDARD
The Court’s review of this proposed class action settlement is governed by Rule 23(e) of the Federal.Rules of Civil Procedure. That rule generally requires the Court “to determine whether a proposed settlement is fundamentally fair, adequate, and reasonable.” Hanlon v. Chrysler Corp., 150 F.3d 1011, 1026 (9th Cir.1998) (citing Class Plaintiffs v. City of Seattle, 955 F.2d 1268, 1276 (9th Cir.1992)). “It is the settlement taken as a whole, rather than the individual component parts, that must be examined for overall fairness.” Id. (citing Officers for Justice v. Civil Serv. Comm’n of S.F., 688 F.2d 615, 628 (9th Cir.1982)).
“District courts have interpreted Rule 23(e) to require a two-step process for the approval of class action settlements: ‘the Court first determines whether a proposed class action settlement deserves preliminary approval and then, after notice is given to class members, whether final approval is warranted.’” In re High-Tech Emp. Antitrust Litig., No. 11-CV-02509-LHK, 2014 WL 3917126, at *3 (N.D.Cal. Aug. 8, 2014) (quoting Nat’l Rural Telecomms. Coop. v. DIRECTV, Inc., 221 F.R.D. 523, 525 (C.D.Cal.2004)). At the final approval stage, it is well-established that the Court must balance the following non-exhaustive factors to evaluate the fairness of the proposed settlement: “the strength of the plaintiffs’ case; the risk, expense, complexity, and likely duration of further litigation; the risk of maintaining class action status throughout the trial; the amount offered in settlement; the extent of discovery completed and the stage of the proceedings; the experience and views of counsel; the presence of a governmental' participant; and the reaction of the class members to the proposed settlement.” Hanlon, 150 F.3d at 1026 (citing Torrisi v. Tucson Elec. Power Co., 8 F.3d 1370, 1375 (9th Cir.1993)).
It is less clear what factors should guide the Court’s evaluation of the proposed settlement at the preliminary approval stage. “Some district courts . .■. have stated that the relevant inquiry is whether the settlement ‘falls within the range of possible approval’ or ‘within the range of reasonableness.’” In re High-Tech Emp. Antitrust Litig., 2014 WL 3917126, at *3 (quoting In re Tableware Antitrust Litig., 484 F.Supp.2d 1078, 1079 (N.D.Cal.2007)) (citing Cordy v. USS-Posco Indus., No. 12-553, 2013 WL 4028627, at *3 (N.D.Cal. Aug. 1, 2013)). In determining whether the proposed settlement falls within the range of reasonableness, perhaps the most important factor to consider is “plaintiffs’ expected recovery balanced against the value of the settlement offer.” Id. (quoting In re Nat’l Football League Players’ Concussion Injury Litig., 961 F.Supp.2d 708, 714 (E.D.Pa.2014)); see also Nielson v. Sports Auth., No. C-11-4724-SBA, 2012 WL 5941614, at *6 (N.D.Cal. Nov. 27, 2012), Determining whether the settlement falls in the range of reasonableness also requires evaluating the relative strengths and weaknesses of the plaintiffs’ case; it may be reasonable to settle a weak claim for relatively little, while it is not reasonable to settle a strong claim for the same amount. See In re High-Tech Emp. Antitrust Litig., 2014 WL 3917126, at *4.
Where, as here, the parties reach a settlement before class certification, the district court must apply a “higher standard of fairness.” Hanlon, 150 F.3d at 1026. This additional scrutiny is needed to ensure that the interests of the class are adequately protected, because the agreement has “not [been] negotiated by a *936court-designated class representative.” Id.
IY. PROSPECTIVE RELIEF
A. Whether the Settlement Must Include a Reclassification
As a preliminary matter, the Teamsters contend the Court should reject not only this settlement, but any settlement that doesn’t confer employee status on Lyft drivers going forward. The Teamsters note that the named plaintiffs originally sought reclassification on behalf of the drivers, and they argue that an agreement which retains the drivers’ current classification is too inconsistent with the purposes of the lawsuit. “[M]inimal settlements that abandon injunctive relief or a resolution of the essential issue for a modest payout,” the Teamsters argue, “encourage a ‘race to the bottom’ and added financial burdens on the State and the public.” They observe that continuing to classify Lyft drivers as independent contractors rather than employees results in a lower tax base and causes a “drain on public finances and the taxpayers.” And the Teamsters contend the drivers will continue to bear an unfair share of the costs of running Lyft’s business going forward— for example, the drivers’ continued classification as independent contractors means Lyft is not required to reimburse them for expenses or provide workers’ compensation benefits.
These are good arguments, but they are mostly policy arguments best directed to the legislative or executive branches. The Court’s job is not to decide whether it would be better for society if Lyft drivers were classified as employees. The Court’s job is to assess whether the settlement falls within a range of fair outcomes for the class members, considering the risks they would face if they took the case to trial. And the Teamsters, in objecting to the settlement, largely ignore those risks. It is true (as discussed more fully in Section V.E below) that plaintiffs’ counsel, in their effort to obtain preliminary approval, overstate some of the risks they would face on the way to trial. But at the trial itself, the plaintiffs face a significant risk that a jury would find the drivers to be independent contractors under California law. As discussed in the Court’s prior summary judgment ruling, the jury would be required to apply California’s multi-factor test for distinguishing between employees and independent contractors. Some factors point one way, some factors point the other way, and many factors seem outdated and irrelevant. Dkt. No. 94; Cotter, 60 F.Supp.3d at 1076-81.
The risk is particularly acute for people who drive only sporadically and part time for Lyft. In contrast to people who drive full time and rely on their work for Lyft as their primary source of income, a jury might be reluctant to conclude that Lyft “employs” people who only give occasional rides when their schedules permit. Cf. Bowerman v. Field Asset Servs., Inc., No. 13-CV-00057-WHO, 2014 WL 4676611, at *11-12 (N.D.Cal. Sept. 17, 2014); Narayan v. EGL, Inc., 285 F.R.D. 473, 479 (N.D.Cal.2012). For example, during his deposition testimony, Patrick Cotter described the occasional rides he gave for Lyft as a hobby he pursued when he was not busy with his full-time job at Face-book. Lyft would presumably make much of this at trial, possibly to great effect. And recall that part-time drivers constitute a large percentage of the proposed class. Only 755 members of the 150,000-person class are full-time drivers, meaning they drove at least 30 hours a week in at least half of the weeks they worked for Lyft. And more than two-thirds of the class members — over 100,000 people— have driven less than 60 hours in total for Lyft. Of course, the sporadic nature of the part-time drivers’ work is not disposi-*937five, and a jury could find that even drivers who worked very little for Lyft were employees, not independent contractors. See Cotter, 60 F.Supp.3d at 1081 n. 7; Burlingham v. Gray, 22 Cal.2d 87, 137 P.2d 9, 16 (1943); cf. Brock v. Superior Care, Inc., 840 F.2d 1054, 1060-61 (2d Cir.1988). But the outcome either way is far from assured.
It should thus be clear that this case is quite different from In re General Motors Corp. Pick-Up Truck Fuel Tank Products Liability Litigation, 55 F.3d 768 (3d Cir.1995), on which the Teamsters rely. That case involved an allegedly life-threatening defect in the fuel tank design of pickup trucks. The Third Circuit held that the district court erred in approving a settlement agreement that: (i) gave the plaintiffs coupons that had no actual cash value but' could be applied toward the purchase of new pick-up trucks; and (ii) didn’t require the manufacturer to fix the allegedly dangerous defect in existing trucks. See id. at 806-07, 811. The Teamsters argue that a “defect” is also not being fixed here, in that Lyft will continue to treat the drivers as independent contractors. But that argument presumes that classifying the drivers as independent contractors is a “defect” in the first place — a presumption with which a jury could well disagree.
There is another, related problem with the Teamsters’ insistence that the Court reject any settlement that doesn’t require the drivers to be reclassified. The Teamsters use rhetoric that seems designed to imply that the proposed settlement agreement carves the drivers’ status as independent contractors into stone going forward. That is not correct. Under the settlement, class members are only required to release claims against Lyft that accrued between May 2012 and the date of preliminary approval of the agreement. Thus, there would be nothing to stop drivers from pursuing future claims against -Lyft relating to their employment status.2 This includes claims under California’s Private Attorneys General Act to enforce the state’s wage and hour laws, which are not subject to individual arbitration. See Sakkab v. Luxottica Retail N. Am., Inc., 803 F.3d 425, 427 (9th Cir.2015); Iskanian v. CLS Transp. L.A., LLC, 59 Cal.4th 348, 173 Cal.Rptr.3d 289, 327 P.3d 129, 152 (2014), cert. denied, — U.S. -, 135 S.Ct. 1155, 190 L.Ed.2d 911 (2015). Perhaps more importantly, nothing in this agreement would preclude the appropriate state or federal executive branch agencies from taking enforcement actions against Lyft if they determine that Lyft has misclassified its drivers under state or federal law. For that matter, nothing in this agreement would prevent’ the California Legislature (or Congress) from offering a legislative solution to the- problem of how to classify workers in the “gig economy.”
B. Whether the Proposed Changes Are Meaningful
Putting aside the Teamsters’ argument that the Court should reject any settlement that doesn’t involve reclassification, the question remains whether the prospective relief contemplated by the agreement is of any real value to the drivers. As mentioned previously, the current contract between Lyft and its drivers allows Lyft to terminate drivers “at will” — that is, at any time and for any reason. Drivers have no formal right to seek reconsideration of any termination decision short of arbitration. Under the settlement, Lyft will only be allowed to terminate drivers for breach of contract, and the newly-proposed contract *938sets out a laundry list of misconduct that constitutes- a breach. Moreover, the proposed contract creates an official, internal appeal process whereby, in many -instances, a driver will receive notice of the termination decision, and will be given an opportunity to argue that Lyft made a factual error or was too harsh in its decision to terminate. And in all cases, if a driver is terminated, he may pursue a claim in arbitration, and Lyft will pay all arbitration-specific fees.
It is difficult to tell, on the current record, just how significant these changes would be. Although Lyft could no longer terminate drivers at will, the list 'of ways a driver can breach the proposed contract is quite long, and it’s not clear that Lyft has ever terminated a driver in the past for a reason not listed in the newly-proposed contract. On the other hand, one potentially significant change is that under the proposed contract (in contrast to the current contract), Lyft may not terminate drivers for declining ride requests while they are signed in to the platform. And more generally, the drivers would now have a contractual commitment from Lyft that it couldn’t terminate them for reasons that are arbitrary and unrelated to the performance of their contractual obligations.
Similarly, although Lyft drivers would have a formal process under the proposed settlement to appeal many termination decisions to the company before resorting to arbitration, as a practical matter, drivers may already be able to “appeal” a termination decision by sending an email to the company. On the other hand, creating a formal' process like this gives drivers a degree of assurance that any dispute they may have with the company will not -be ignored.3
Finally, although Lyft would be required to pay arbitration-specific fees if a dispute is not resolved with a driver internally, California law already reqúires a company to pay most or all of those fees in disputes relating to employment. Armendariz v. Found. Health Psychcare Servs., Inc., 24 Cal.4th 83, 99 Cal.Rptr.2d 745, 6 P.3d 669, 687-88 (2000). On the other hand, the scope of disputes for which Lyft would be compelled to pay arbitration-specific fees under the new agreement may be broader than that called for by California law. For example, if a driver were to claim that Lyft reneged on a promise to give commissions for successfully referring hew drivers (a claim that would not obviously fall under the umbrella of employment disputes envisioned by Armendariz), under the new agreement, the driver would be able to arbitrate that claim at Lyffis expense.
The Teamsters complain that the protections for drivers under the settlement agreement would be weaker than the protections often given to unionized employees who work pursuant to the terms of a collective bargaining agreement — protections such as the right to a hearing about any disciplinary action and thé obligation by the company to impose “progressive discipline.” That is correct, but .the comparison is inapt. Even if Lyft drivers were indisputably “employees” under California law, they would not have a legal right to the' kinds of protections labor unions are sometimes able to negotiate for the people they represent. The true question is whether the prospective relief contemplated by the agreement improves job security for Lyft drivers, compared to the *939situation they face now. And although this question can’t be reliably answered on this record, the Teamsters’ assertion that the changes are “largely illusory” is probably an overstatement. The changes certainly aren’t revolutionary, but they are not nothing, either.
Y. MONETARY RELIEF
A. The Settlement Calculation
As mentioned in Section II, the agreement requires Lyft to pay $12.25 million into a settlement fund. The agreement contemplates that plaintiffs’ counsel will, at the final approval stage, seek thirty percent of this amount to cover costs and attorneys’ fees. The agreement also assumes that class notice and administration costs will total $120,000. After subtracting the State’s portion of the PAGA penalties, this would leave $8.36 million to be claimed by the class members, who would be eligible for payments based on the amount of driving they did during the class period. Assuming all drivers claim their money, this would result in an average payout of $53.02 to part-time drivers and an average payout of $676.19 to full-time drivers. These figures take into account the fifty percent premium that full-time drivers would receive — a premium that represents an effort to acknowledge the relative strength of the full-time drivers’ claims. As a practical matter, many drivers who file claims will actually receive more, because some drivers won’t claim money from the fund (as happens in all class actions), and the unclaimed money will be redistributed to the drivers who do make claims. Again, the class would consist of people who drove for Lyft in California between the company’s inception in 2012 and the date of preliminary approval of the settlement.
According to plaintiffs’ counsel, this settlement figure was reached by estimating a maximum recovery for the class in the event it prevailed at trial, and then -discounting that amount to account for the risks the class would face along the way. In estimating the maximum recovery, and in reaching the ultimate settlement amount, the primary factor for plaintiffs’ counsel was the drivers’ claim for mileage reimbursement. Counsel used the mileage reimbursement rate adopted by the Internal Revenue Service. This accounts for the cost of gas, as well as vehicle wear and tear. Plaintiffs’ counsel obtained data from Lyft about the approximate number of miles covered by drivers in picking up passengers and then taking them to their destinations, from the time the company started up in May of 2012 until June 2015. Multiplying the estimated number of miles driven during this period by the IRS rate, plaintiffs’ counsel came up with a figure of $64 million for the class.
With respect to the other claims for damages contained in the lawsuit, plaintiffs’ counsel assigned a minimal value. Counsel valued the claim for overtime pay at $900,000, and the claim for tip reimbursement at $250,000. Counsel valued the remaining claims for damages- (based on alleged failure to pay minimum wage, failure to give meal and rest breaks, failure to provide accurate pay records, and failure to timely provide full pay upon an employee’s departure) at zero. Thus, counsel’s total maximum damages estimate was $65.15 million.
Plaintiffs’ counsel then allegedly attempted to estimate a maximum realistic award of PAGA penalties. Because courts have significant discretion to reduce the amount of PAGA penalties imposed on an employer, see Cal. Lab. Code § 2699(e)(2), plaintiffs’ counsel assumed that the Court, in this case, would be unwilling to assess penalties in an amount higher than one-third of the total maximum damages recovery. One-third of $65.15 million is $21.72 *940million. But then, counsel reduced this figure further to account for the fact that the State of California must receive seventy-five percent of any PAGA penalties recovered by a private plaintiff. Cal. Lab. Code § 2699(i). Thus, the “maximum” PAGA penalty became $5.43 million. Adding this $5.43 million to the estimated $65.15 million in maximum damages, counsel came up with a figure of $70.58 million as an estimate of the maximum amount of recovery if the drivers prevailed in the litigation.
Plaintiffs’ counsel then discounted this figure based on a variety of risks they believed the class would face on the road to recovery, coming up with the settlement figure of $12.25 million. Of this, $122,250 is allocated to PAGA penalties, with the State of California slated to get seventy-five percent of that amount.
B. The Reimbursement Claim
There is one clear problem (and it is a major one) with the way plaintiffs’ counsel reached this figure. The problem is, using the approach devised by plaintiffs’ counsel themselves, they shortchanged the drivers at least by half — because a proper estimate of the maximum reimbursement claim is at least twice the amount plaintiffs’ counsel based the settlement on. Counsel assigned a maximum value of $64 million to the drivers’ reimbursement claim. That figure was based on data provided by Lyft about the approximate number of miles logged by drivers from the company’s inception through June 2015. But the class period does not end at June 2015 — it proceeds through the date that the settlement receives preliminaiy approval. In response to questions posed by the Court, the plaintiffs and Lyft have now revealed that by February 14, 2016 (more than a month prior to the hearing on the motion for preliminary approval), Lyft drivers had logged roughly double the amount of miles, so that the maximum value of the reimbursement claim was actually $126 million as of that date.
Counsel for the plaintiffs reached their agreement with Lyft in November 2015. They did not base that agreement on updated mileage figures as of November 2015. Nor did they attempt to estimate what the mileage figures would be around the time of preliminary approval, which they knew would not likely occur before Spring 2016. Instead, the plaintiffs’ lawyers used outdated information, from June 2015. The lawyers correctly assert that the claim for reimbursement is by far the most valuable, and they emphasize that the settlement figure they reached (and identified as fair) in November 2015 was “driven primarily by mileage reimbursement.” Reply in Support of Motion for Preliminary Approval, Dkt. No. 188, at 11 n.8. But 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% percent of the reimbursement claim. This is major defect in the agreement.
C. The PAGA Penalties
There is a second problem with the way counsel for the plaintiffs reached their settlement figure — a problem that would likely defeat preliminary approval even if the reimbursement error were corrected. This problem involves their calculation of the maximum PAGA penalty, along with their seemingly arbitrary reduction of that penalty to a miniscule portion of the settlement amount — $122,250, which is less than one percent of the total.
Plaintiffs’ counsel assumes that the Court would impose PAGA penalties in an amount equaling one-third of the damages recovery. For that proposition they cite Amaral v. Cintas Corp. No. 2, 163 Cal.App.4th 1157, 78 Cal.Rptr.3d 572 (2008), *941which they portray as requiring a trial court to limit PAGA penalties to one-third the amount of damages. But Amaral says no such thing. The California Court of Appeal merely held that the trial court did not abuse its discretion, on the facts of that case, .in declining to reduce PAGA penalties to an amount lower than approximately one-third of the total damages award. Id. at 618. This indicates a trial court also has discretion to impose a higher penalty if called for by PAGA. And it suggests nothing about the amount of PAGA penalties a court should impose on different facts.
In any event, if plaintiffs’ counsel were correct that the one-third figure is appropriate on these facts, the maximum reasonable PAGA penalty would need to be one-third of the maximum damages recovery after correcting the error plaintiffs’ counsel made in reaching that figure. Furthermore, in estimating the maximum reasonable PAGA penalty, there is no basis for slashing it by seventy-five percent (as plaintiffs’ counsel did here) on the ground that the State of California would be entitled to recover that portion. Just because the State would receive it doesn’t mean it’s not part of the penalty. To the contrary, a PAGA plaintiff stands in the shoes of the State in enforcing its wage and hour laws. See, e.g., Achal v. Gate Gourmet, Inc., 114 F.Supp.3d 781, 806 (N.D.Cal.2015). So it makes no sense to simply erase the State’s portion of the recovery when estimating the maximum value of the claim. Nor, for that matter, do plaintiffs’ counsel provide any clue as to how they ultimately attributed one percent of the settlement amount to the PAGA claim, if they truly believed the PAGA claim was worth one-third of the damages. If the case went to- trial, and if the plaintiffs won, and if the Court imposed PAGA penalties in an amount equaling one-third of damages, the State would get seventy-five percent of that amount. So if the plaintiffs wish to assume a maximum PAGA penalty of one-third the amount of damages, they must do so, and not reduce that figure arbitrarily so as to shortchange the State of the amount it would be entitled to recover under the settlement.
On the other hand, there is an argument, given the facts of this case, that counsel for the plaintiffs are using an overly aggressive approach in assuming the maximum reasonable PAGA penalty would be one-third of damages. Trial courts have the discretion to reduce PAGA penalties when it would be “unjust” or “oppressive” to impose the full amount. Cal. Lab. Code § 2699(e)(2). This does not appear to be a case in which a company deliberately sought to evade California’s wage and hour laws by classifying workers as “independent contractors” when it knew they were really “employees.” How to classify Lyft drivers under California’s archaic law is a difficult question, and it does not seem that Lyft made its initial classification decision in bad- faith. It would therefore be “unjust” or “oppressive” to impose full PAGA penalties, and it may well be unjust to merely reduce them to one-third of damages.
In sum, counsel’s estimate of the value of the PAGA claim, and the portion of the settlement attributed to that claim, is arbitrary. It shortchanges the State .on its share of PAGA penalties, It ties the value of the PAGA claim to the facts of a different wage and hour case (Amaral) that merely happened to result in a one-third penalty amount. And then, after assuming a one-third penalty amount, the settlement inexplicably assigns a value to the PAGA claims far smaller than that.
D. Premium for Full-Time Drivers
There is one additional, albeit easily curable, problem with the settlement agreement. As previously noted, the proposed *942settlement gives full-time drivers (i.e., those class, members who worked at least 30 hours a week in at least half of the weeks they drove for Lyft) a fifty percent premium above part-time drivers in recognition that, as a practical matter, full-time drivers may have stronger claims for employee status at trial. If the parties had proposed a settlement amount tethered to the maximum value of the reimbursement claim as of the date of the preliminary approval hearing, it would have been possible to give the full-time drivers a much larger premium' than fifty percent (by orders of magnitude) while still achieving a greater monetary recovery for all class members. And a much larger premium for full-time drivers is warranted given the acknowledgement by all parties involved (including the Teamsters) that they have a stronger case.
E. The Effect of These Errors
At the hearing on the motion for preliminary approval, both sides seemed to acknowledge the error made by the plaintiffs’ lawyers in tying the settlement amount to an- artificially low estimate of the value of the reimbursement claim. Nor did either side have a rational explanation for the value that was assigned to the PAGA claim, or for why they attributed such a miniscule portion of the settlement to that claim. The parties argue that the proposed settlement nonetheless falls within the range of reasonableness, because other courts have approved settlements with monetary components that represented an even smaller percentage recovery for the plaintiffs. But a trial court must assess the reasonableness of a proposed settlement in light of unique facts of the case at hand, in light of the proposed settlement as a whole, and in light of the particular risks involved. In this case, perhaps such a low monetary recovery would warrant preliminary approval if the prospective relief were more significant (such as, for example, an agreement to reclassify Lyft drivers as employees, or to reimburse them for expenses going forward). But as discussed above, the non-monetary relief, while more meaningful than the Teamsters claim, is not nearly significant enough to excuse these serious monetary defects. Or perhaps the low monetary recovery would warrant preliminary approval if the plaintiffs’ case were exceedingly weak. But as discussed below, while the driver class does face risks in this case, the lawyers for the plaintiffs have overstated some of those risks. There is a reasonable possibility that, if the plaintiffs took the case to trial: (i) the class would be certified; (ii) the class would not be subject to arbitration; and (iii) the drivers would win the case in front of a jury.
First, there is no apparent reason why the proposed class of drivers wouldn’t meet the criteria set out by Rule 23 for certifying a class. The case presents a common question — namely, in light of Lyft’s right to control the work of the drivers, are those drivers employees or independent contractors under California law? This is not merely a question common to all drivers in the case, it is the predominant question in the case. And although the Court previously expressed concern that Cotter and Maciel may not be adequate representatives of people who drove for Lyft full time or more regularly, Jeffrey Knudtson, allegedly a full-time driver, has been added as a plaintiff. Indeed, to the extent (as Lyft argues) the difference between part-time drivers and full-time drivers is significant for class certification purposes, the Court could certify sub-classes, with-one sub-class represented by Cotter and Maciel, and the other subclass represented by Knudtson. And as the previous discussion of monetary relief makes clear, this case is amenable to the creation of a damages methodology that *943could apply in a formulaic way to all drivers.
Second, counsel for the plaintiffs note that even if a class were certified, Lyft would seek to compel each class member to individual arbitration, based on the arbitration provision contained in Lyft’s current contract with the drivers. At times, counsel for the plaintiffs seem to suggest this will be a slam dunk for Lyft. But it is far from clear that Lyft could successfully compel the unnamed class members to arbitrate their claims individually. First, Lyft could be held to have waived its right to compel arbitration as to the class. After all, Lyft explicitly waived its right to assert the arbitration provision against the named plaintiffs. And it did so in an attempt to get a favorable ruling on the merits from this Court before class certification. It would arguably be unfair for Lyft to pursue litigation in a federal class action in hope of obtaining a favorable district court ruling at summary judgment, only to keep an arbitration provision in its back pocket in case things didn’t go as planned. Cf. Edwards v. First Am. Corp., 289 F.R.D. 296, 306-07 (C.D.Cal.2012); Petroleum Pipe Americas Corp. v. Jindal Saw, Ltd., 575 F.3d 476, 480 (5th Cir.2009). In this respect, the case cited by Lyft, Mora v. Harley-Davidson Credit Corp., No. 1:08-CV-01453-AWI, 2012 WL 1189769 (E.D.Cal. Apr. 9, 2012), is distinguishable. Mora did not .involve an attempt by a class action defendant to pursue litigation in court until that forum was no longer to its liking. (Indeed, the named plaintiff in that case wasn’t subject to an arbitration provision at all.) O’Connor v. Tiber Technologies, Inc., No. 13-cv-03826-EMC, is distinguishable for the same reason. In re TFT-LCD (Flat Panel) Antitrust Litigation, No. M 07-1827-SI, 2011 WL 1753784 (N.D.Cal. May 9, 2011) is also distinguishable, because, unlike here, the defendants in that case did not explicitly waive arbitration with the named plaintiffs, and moreover, moved to compel arbitration with certain members of the class before the merits had been determined (though they ’ did wait until after class certification). And the question whether -the defendants had waived their right to compel arbitration by litigating in federal court, which is inherently factual and case-specific, was “extremely close” in that case. Id. at *4.
Even beyond the possibility that Lyft has waived the right to force the class members to arbitration, there is at least some authority suggesting the arbitration provision' is unenforceable entirely, because it violates the National Labor Relations Act.' See Totten v. Kellogg Brown & Root, LLC, No. ED CV 14-1766 DMG, 2016 WL 316019, at *16 (C.D.Cal. Jan. 22, 2016). Although this question is not settled, the plaintiffs’ lawyers did not even mention the possibility that they could attack the arbitration 'provision as a violation of the NLRA in their motion for preliminary approval. Finally, as the Teamsters note, PAGA claims are not subject to class action waiver, so at .a minimum the class members could not be forced to arbitrate those claims individually. Sakkab, 803 F.3d at 427; Iskanian, 173 Cal.Rptr.3d 289, 327 P.3d at 152. In all, the plaintiffs’ lawyers have overstated the risk that the unnamed class members could be compelled to individual arbitration following class certification.
Third, as for the trial itself, it’s true that there would be real risks for the drivers. But there would be real risk for Lyft too. As discussed'in the Court’s summary judgment ruling, the primary question for the jury at trial would be whether Lyft has retained the right to “control” the drivers’ work. Not whether Lyft actually exercised control, but whether it retained the right to do so. In many respects, Lyft does indeed retain that right. See Cotter, *94460 F.Supp.3d at 1078-79. The jury would be instructed that, under California law, this is the primary factor in deciding whether a worker is an “employee” or an “independent contractor.” And although a jury could be tempted to conclude that people who drive only sporadically for Lyft should be independent contractors, it seems equally likely that the jury could analogize Lyft drivers to restaurant workers who work in multiple venues, but only occasionally at each particular venue. There is no dispute that, under California law, someone who picks up a couple of restaurant shifts here and there is an “employee” of that restaurant (along with any other restaurants where he works). And if the jury reached a similar conclusion about Lyft drivers, the consequences for Lyft would be enormous.
In light of the foregoing, the Court declines to approve the settlement. Given the defect in the reimbursement calculation, to fall within the range of reasonableness the settlement amount would, at a minimum, need to be adjusted so it represents roughly 17 percent of the maximum value of the reimbursement claim as of the date of the prior preliminary approval healing.4 Moreover, any settlement amount must properly account for the PAGA claim and must not shortchange the State of its portion of PAGA penalties (although the plaintiffs should not assume the Court would impose penalties as high as one-third the amount of damages). Finally, given the relative strength of their case, the full-time drivers would need to receive a much larger premium.
F. Other Objections
The Teamsters object to the decisions by plaintiffs’ counsel to value the drivers’ other claims at such low amounts — for example, the decisions to value the overtime claims at $900,000 and the minimum wage claims at $0. On the surface, it appears that any problems in these valuations would be minor matters of degree, in contrast to the fundamental defects inherent in the reimbursement and PAGA valuations. Nonetheless, if the parties wish to present another agreement for preliminary approval, counsel for the plaintiffs should be prepared to better explain why they valued these claims as they did.
VI. CONCLUSION
The motion for preliminary approval is denied. If the parties wish to present another agreement for preliminary approval that is based on mileage estimates as of the date of the preliminary approval hearing on March 24, 2016 and that is otherwise consistent with this opinion, the motion must be heard by no later than May 2016.
IT IS SO ORDERED.
. For convenience, the Court will refer to this group of objectors as the Teamsters.
. Regarding the class members’ release of claims, the parties may need to alter the language of the proposed notice to the class, because the proposed notice as currently drafted suggests the release is broader than actually set forth in the settlement agreement.
. Incidentally, it isn’t clear why Lyft could not make this pre-arbitration dispute resolution process available in every instance where Lyft has terminated, or has decided to terminate, a driver for breach of contract. In the event of a renewed motion for preliminary approval, the parties should address this question.
. Of course, if a settlement is preliminarily approved, the date of preliminary approval will not have been March 24, 2016. But if the parties choose to negotiate a new agreement consistent with this ruling, they cannot predict with precision when the preliminary approval date would be. Assuming they seek preliminary approval reasonably promptly, data regarding the estimated number of miles accrued by March 24, 2016 (the date of the prior hearing) would provide an appropriate benchmark. Specifically, if the parties notice a hearing for May 2016, this would be sufficient. The Court will entertain a motion to shorten time if necessary to hold a hearing in that time frame.