Click here to view the image.
In recent weeks, two courts ruled on motions to dismiss the first wave of class action lawsuits based on alleged price optimization of auto insurance rates. In both Stevenson v. Allstate Ins. Co., No. 15-cv-04788 (N.D. Cal. March 17, 2016), and Harris v. Farmers Ins. Exch., No. BC579498 (Cal. Super. Jan. 25, 2016), the courts invoked the “primary jurisdiction” doctrine to stay the litigation, pending further proceedings by California regulators. Yet both courts also issued rulings on important elements of the plaintiffs’ claims—holding, among other things, (1) that insureds whose rates are affected by price optimization suffer an “injury in fact”; (2) that failure to disclose price optimization can make advertisements “false and misleading”; and (3) that insurers using price optimization may have been unjustly enriched. These rulings raise more questions than they answer about the exposure insurers now face for their actual practices—especially because the two courts came to opposite conclusions about what it is the defendants were accused of doing in the first place.
What Is Price Optimization, And Why Do They Hate It?
In principle, property insurance prices reflect “an actuarially sound estimate of the expected value of all future costs associated with an individual risk transfer.” Insurers arrive at that estimate by (1) calculating a base rate, derived from the total expected cost of insuring a group of policyholders, and (2) defining a set of rating variables or classes—characteristics (such as driving history, age or location) that might increase or decrease the expected cost of insuring an individual policyholder, as compared with other members of the group. Each rating variable is assigned a numerical value, known as a “rating factor” or “relativity.” Individual premiums are calculated by multiplying the base rate by the various rating factors applicable to any given policyholder.
Ideally, insurers would charge the same premium to all individuals who represent the same degree of risk. In reality, the rating variables insurers use cannot capture every circumstance or characteristic that might affect future costs. What insurers can do is equalize (within limits) the premiums of insureds who have the same “risk profile”—i.e., who belong to the same set of rating classes used in the insurer’s rating plan.
In business, the term “price optimization” (“PO”) means setting prices at levels that will maximize profit. When applied to insurance rates, the term generally describes processes that further insurers’ business objectives, by modifying cost-based rates on the basis of information about consumer demand and competitive conditions. PO is commonly associated with such objectives as retaining existing customers and penetrating new markets. As broadly defined, PO is not new. Insurers have historically applied business judgments about consumer demand and competitive conditions to cost-based pricing—for example, by “capping” severe price increases that might otherwise be implemented in response to updated loss information. PO has recently become topical, because those judgments are now being made with the kind of “sophisticated statistical analysis” that is often associated with “big data.”
When regulators and consumer advocates discuss PO, they focus primarily on insurers’ consideration of “the willingness of certain policyholders to pay higher premiums than other policyholders.” Early PO critics invoked scenarios in which insurers might use PO “to determine how much of a premium increase the policyholder will tolerate.” The resulting rate structure might be “unfairly discriminatory”—and, therefore, prohibited under state insurance laws —because it could “result in two insureds with similar risk profiles being charged different premiums.” Of particular concern was the possibility that lower-income consumers might exhibit the smallest elasticity of demand—i.e., the smallest likelihood of responding to a price increase by changing insurers—and so might be singled out to pay higher prices.
In the UK, where it is lawful to do so, some personal lines insurers have, in fact, implemented PO on an individual basis. But US insurers have applied market considerations to the rating process in more complex ways. Some, for example, practice what one vendor calls “ratebook optimization.” This process uses statistical models of cost and consumer demand to modify rating factors—the numerical values assigned to the different rating variables the insurer is already using to price policies. Ratebook optimization does not generate different premiums for policyholders with the same risk profile. And if the modifications to rating factors are within a range that is actuarially justified, the resulting premiums will be neither “excessive, inadequate [n]or unfairly discriminatory.”
Furthermore, evidence that low-income consumers are less likely to shop for insurance has not materialized. One survey conducted by the Insurance Information Institute found that 68 percent of consumers with incomes under $35,000 had shopped for insurance at the time of their last renewal, compared with 61% of consumers earning $100,000.
Consequently, recent criticisms of PO have narrowed their scope. Rather than challenge all use of the practice, they now warn against its being applied in some cases with excessive “granularity.” They contend, for example, that some insurers have divided policyholders into an unreasonable number of sparsely-populated segments, and that this allows the insurers to apply estimates of elasticity of demand at something approaching an individual level. They also assert that insures can manipulate the margin of error for actuarial estimates to justify prices that are unrelated to cost.
Critics also express concern about the types of consumer information that enter the rating process through the integration of consumer demand models. These models might take account of data—such as magazine subscriptions, or whether a consumer uses bottled water—that have no demonstrated correlation to the insured risk. Critics argue that price determinations based on considerations unrelated to risk will make insurance less effective at pooling risks, and will make insurance prices less effective at incentivizing risk-averse conduct. They also warn that consumers will have no opportunity to detect and correct price determinations made on the basis of erroneous data.
To date, the insurance departments of 19 states have issued bulletins dealing with PO. The earliest bulletins purported to define and prohibit the practice. California’s Department, for example, offered an exceptionally broad definition of PO as “any method of taking into account an individual’s or class’s willingness to pay a higher premium relative to other individuals or classes.” It then declared that
any use of Price Optimization in the ratemaking/pricing process or in a rating plan is unfairly discriminatory in violation of California law.
More recent statements are more circumspect; they articulate concerns about PO, without prohibiting any particular practice.
Price Optimization Class Actions
Since early 2015, a single firm has filed four lawsuits challenging insurers’ alleged use of PO. The first two cases—Slocombe v. The Allstate Corp., No. 15-2-03508-8 (Wash. Super. Ct.), and Durham v. The Allstate Corp., No. BC 571810 (Cal. Super. Ct.)—were voluntarily dismissed shortly after they were filed. The courts in the other cases—Harris v. Farmers and Stevenson v. Allstate—recently ruled on motions testing the validity of the complaints.
All four cases have been pleaded in essentially the same way. In each case, the plaintiffs are longstanding policyholders of the defendant insurer. They seek to represent a state-wide class, consisting of all customers of the defendant who “were charged or paid a higher premium than the risk-based premium.” In amended complaints, this definition was narrowed to include only those insureds who paid more than a risk-based premium, and for whom “elasticity of demand” had been “used as a rating factor.”
To establish that Farmers and Allstate are using PO in their rate structures, the plaintiffs rely on a variety of public statements. Most of these appear to come from the LinkedIn profiles of current and former employees, who reported having worked with PO and PO products during their time with one of the defendants. Allstate also cited “price optimization” as one of its “goals” in a 2011 securities filing, and it thereafter reported that it was using “increasingly sophisticated pricing models” in some (unidentified) states. For Farmers, plaintiffs note that an actuary employed by Farmers has defended PO in presentations to the Casualty Actuarial Society and the National Association of Insurance Commissioners.
The California complaints purport to state multiple claims under that state’s Unfair Competition Law, Cal. Bus. & Prof. Code § 17200, and under Section 1861.10 of the California Insurance Code. They allege that the use of PO is “unlawful,” “unfair” and “fraudulent,” because elasticity of demand is not a permissible rating variable under California law; because the defendants allegedly used PO “in bad faith”; and because the insurers’ failure to disclose their reliance on PO was allegedly misleading. Plaintiffs also asserted tort claims for unjust enrichment.
Slocombe, the Washington suit, asserted a single claim under the Consumer Protection Act, Wash. Rev. Code Wash. §19.86 et seq., and an additional claim for unjust enrichment.
What The Lawsuits Say About Price Optimization
In describing PO, the class action complaints echo the earliest criticisms of the practice. Insurers should take note of the fact that they do so, in large part, by quoting from marketing materials of companies that sell PO products.
Language from these materials is used to suggest that PO represents a significant departure from cost-based pricing, because “ratemaking based on risk and cost alone is no longer sufficient.” It is also used to suggest that the goal of PO is simply to extract higher payments from consumers, because the “cost plus profit approach leaves a lot of money on the table,” and “[t]here are cases in which consumers may be willing to pay a higher price than what insurers are charging.” (The latter advertisement goes on to say that “there are many other cases where insurers ask more than what customers are willing to pay,” but that statement was not included in the complaints.)
As for how PO actually works, the complaints provide little detail. The first round of pleadings described a simplified version of ratebook optimization. The plaintiffs alleged:
Defendants … increas[e] the relativities [i.e., the values assigned to different rating variables] for rating characteristics that are associated with a willingness to tolerate a price increase. The rates thereby produced exceed the risk-based rates that those policyholders would pay absent Defendants’ use of [PO].
In plaintiffs’ account, then, PO is used exclusively to increase prices, rather than (for example) to lower rates that would otherwise impair customer retention or competition for new business. They also assert that the process identifies the elasticity of demand of specific rating classes that are already in use. According to plaintiffs’ original complaints, PO simply jacks up the prices for the most vulnerable rating classes. The plaintiffs contended that this practice enables insurers to “conceal their use of [PO]” from regulators that approve the rating plan.
These contentions exposed two potential flaws in the plaintiffs’ claims. First, because the complaints failed to identify the specific rating classes for which PO allegedly produced increased rates, they were unable to substantiate the allegations that the named plaintiffs suffered any actual injury. This problem was not addressed by the recent decisions in the two cases. It is likely to resurface if the lawsuits continue, but it will be tied up with questions about whether plaintiffs have accurately described the way PO works.
The second problem with plaintiffs’ description is that it concedes an important point: the rates the defendants charge—including the rates paid by the plaintiffs—were filed with, and approved by, California’s Department of Insurance. In Stevenson, the defendants seized on this fact, moving to dismiss the complaint under the “filed rate doctrine.” This is a rule which provides that “rates duly adopted by a regulatory agency are not subject to collateral attack in court.” MacKay v. Superior Court, 188 Cal.App.4th 1427 (Cal. Ct. App. 2010).
In response, the plaintiffs amended their complaints in both Stevenson and Harris. The amended complaints appeared to change plaintiffs’ position, alleging that the defendant insurers “use … elasticity of demand as a rating factor.” (They also alleged that this “use” of elasticity was not disclosed to regulators, and so that “the Department did not approve” it.) During hearings in both cases, however, plaintiffs appear to have disclaimed any suggestion that “elasticity of demand” was used as a separate rating variable from the ones the regulators had approved. That is, plaintiffs appear to have reverted to their original description of PO.
What The Courts Had To Say: Primary Jurisdiction
Farmers demurred to the Amended Complaint in Harris, and Allstate moved to dismiss the Amended Complaint in Stevenson. Both motions were granted in part and denied in part.
The most important ruling in both cases was a decision to stay each action under the “primary jurisdiction doctrine.” This is a doctrine that
comes into play whenever enforcement of [a] claim requires the resolution of issues which, under a regulatory scheme, have been placed within the special competence of an administrative body. …
[I]n such a case the judicial process is suspended pending referral of such issues to the administrative body for its views.
United States v. Western Pacific Railroad Co., 352 U.S. 59 (1956).
In Harris, the Superior Court applied the doctrine to the allegation that Farmers used elasticity of demand “as a rating factor”:
Defendants contend that they did not use elasticity of demand as a rating factor. Thus, evaluating Plaintiffs’ claims would necessarily involve a technical analysis of the rating factors and formulas used by Defendants in order determine whether or not elasticity of demand was taken into account. In such a situation, … ‘it seems clear that the Insurance Commissioner … is best suited initially to determine whether his or her own regulations … have been faithfully adhered to … .
In Stevenson, the District Court reached a similar conclusion:
Plaintiff challenges the criteria Defendants take into account when formulating their class plan for approval by the Commissioner. In Plaintiff’s view, Defendants should have disclosed [elasticity of demand] as a rating factor to the Commissioner when they submitted their class plan. These are precisely the types of claims that implicate ‘questions involving insurance rate making [that] pose issues for which specialized agency fact-finding and expertise is needed … .’
As a result of these rulings, there will be no further proceedings in the Harris and Stevenson cases until the California Department addresses the question of whether each insurer’s alleged use of PO violates state law governing insurance rates.
The Courts Say More: Is There Secondary Jurisdiction?
Although both courts found that plaintiffs’ fundamental claims “necessarily involve a technical analysis” that should be conducted by the Insurance Commissioner, both courts also issued several rulings on issues that the analysis might resolve. In effect, the courts ruled (1) that the conduct which is alleged in the complaint is unlawful; but (2) that the courts need help from the Insurance Commissioner to determine whether the defendants actually engaged in that conduct.
Thus, both courts ruled that the plaintiffs stated a plausible claim for unjust enrichment. In Harris, this ruling was supported by a finding that “Plaintiffs have alleged injury in fact and have stated a cause of action under the [Unfair Competition Law].” The court seems to have been convinced, therefore, that California law prohibits the “use of elasticity of demand as a rating factor”—presumably because (as noted above) the state’s Insurance Department has declared that “any method of taking into account an individual’s or class’s willingness to pay a higher premium” is a form of “unfairly discriminatory” pricing. But the court was not willing to decide whether Farmers has actually “used” elasticity of demand in this way—at least, not without the benefit of some regulatory guidance.
This approach carries a risk of sowing confusion, because the courts were unable to settle on a coherent account of just what conduct the plaintiffs actually alleged. The difficulty is illustrated by the courts’ discussion of the filed rate doctrine. In Harris, the court found that the doctrine did not bar the plaintiffs’ claims, because
Plaintiffs allege that in applying the approved rate, Defendants improperly took into consideration elasticity of demand as a rating factor. …
Plaintiffs are not challenging the rate or rating factors filed with the Department of Insurance. Instead, Plaintiffs allege that Defendants used inelasticity of demand as a rating factor without the Department’s approval and as a result charged a rate higher than the approved rate.
Stevenson, on the other hand, held that the filed rate doctrine did apply. Although the Amended Complaints in both cases used identical language to describe the defendants’ conduct, the court in Stevenson concluded that Allstate was not accused of “charging a rate higher than the approved rate”:
The gravamen of Plaintiff’s allegations is a challenge to the approved rates and not the application thereof. … Plaintiff is unable to allege that she paid a premium higher than would be calculated using the rate and class plan approved by the Commissioner.
Based on the same analysis, the two courts also reached opposing conclusions about whether the Insurance Commissioner had exclusive jurisdiction over plaintiffs’ claims, under Cal. Ins. Code § 1860.1.
This lack of clarity makes some of the courts’ other rulings problematic. For example, the ruling on unjust enrichment in Stevenson depended, in part, on the court’s finding that
Plaintiff pleads payment of premiums that were artificially inflated based on … alleged unlawful practices. … Plaintiff’s alleged injury is … that she ‘paid higher prices … than have other insureds’ who were not charged more based on price optimization.
Given the way the Stevenson court interpreted the plaintiff’s case, this conclusion seems premature. If, as the court found, the defendant in Stevenson used PO only to modify the relativities (i.e., the numerical values) that were used in “the rate and class plan [that was] approved by the Commissioner,” then the plaintiff in that case would not have “paid higher prices” than any other insured with the same risk profile. Other insureds might have paid premiums that were not affected by PO, and those premiums might have differed from the plaintiff’s. But those insureds would also have a different risk profile; consequently, the plaintiff’s premiums would still have differed from theirs, even if PO had never been invented. In short, the allegations in the complaint fell short of establishing that PO caused the plaintiff to pay more than other, similarly-situated insureds.
It is also difficult to evaluate the court’s apparent conclusion that PO causes premiums to be “artificially inflated.” For any given relativity used in a rating plan, there might be a range of actuarially-justified values, any of which will generate cost-based premiums. Consequently, the language of Stevenson suggests that even a premium that is justifiable on the basis of cost might still be “artificially inflated”—and therefore unlawful—if the value of a specific relativity was selected with an improper motive. It should be noted that the court’s opinion shares this anomaly with the broad language (quoted above) of the California Commissioner’s bulletin on PO. Neither document gives much guidance about how to distinguish bad motives from lawful ones.
Stevenson also contains a confusing ruling about advertising. The complaint quoted a statement on Allstate’s website, which begins, “The quote you receive is impacted by the following factors,” and which goes on to list eight different variables—none of which is “elasticity of demand.” The court found that this statement could be “false and misleading” under California’s Unfair Competition Law, because Allstate “does not inform customers it uses [elasticity of demand] as a rating factor”—that is, that it “accounts for [elasticity of demand]” when it assigns values to the rating variables that are named in the advertisement.
As discussed above, however, regulators have historically permitted insurers to modify those values for business reasons—such as a desire to limit sudden, large price increases that might cause large numbers of customers not to renew. Thus, the statement cited in Stevenson might have been challenged as “misleading,” even if Allstate did not engage in the conduct alleged in the complaint. But failure to disclose these traditional modifications have not previously given rise to false advertising claims. The ruling in Stevenson means the court thinks some modifications to rating factors are more consequential than others, but, again, it gives little or no basis for identifying the ones that might now be “material.”
The early denunciations of price optimization for personal lines insurance have generally given way to more nuanced and tolerant analyses. The opinions in Stevenson and Harris share characteristics with the earlier discussions, in which details were scarce, and every form of the practice was assumed to be a disreputable and illegitimate departure from traditional practices. The results are unfavorable for insurers, but it is difficult to predict how, if at all, they will affect future litigation.
The good news is that both cases recognize that price optimization raises complex issues which regulators must resolve.