By motivating lawyers to handle class actions, fee awards enable millions of people to obtain access to justice every year and strengthen the effect of regulatory laws. But the process by which judges decide how much to pay lawyers has long been a black box. Settlements go in one side; fee awards come out the other. Researchers have studied the inputs and the outputs, but not the process that connects the two. Consequently, it is difficult to know why judges award the amounts they do.

Our new study, to be published this fall in the Columbia Law Review, seeks to close that gap. We assembled a dataset of 434 securities class actions that settled in federal district courts from 2007 through 2012. We scoured the dockets to gather materials covering all of the major points at which federal judges are likely to consider issues relating to fees, mainly the competition among lead plaintiff candidates for control of the litigation and the preliminary and final reviews of settlements and fee awards. These data enable us to paint a picture of the fee-setting process that is unusually detailed and nuanced. Our findings are extensive, and the full paper is available here. For this blog post, we decided to highlight just three findings and talk about the normative conclusions that we think flow from them.

Settlement Size and Judicial Experience

One of our most striking findings relates to the famous “increase/decrease” rule, according to which an increase in the size of a settlement predicts a decrease in class counsel’s fee percentage. The existence of this rule is one of the most robust findings of empirical research on class actions, and commentators have implicitly assumed that all districts and judges adhere to it. They don’t. We found that the rule is followed only by judges who see lots of securities class action settlements themselves or sit in courts that handle these cases frequently. Other judges tend to award constant fee percentages, regardless of settlement size. For a settlement one standard deviation above the mean ($216.8 million), fee awards in low volume districts averaged about 29% compared to 25% in high volume districts. This is an average difference of about $10 million in the few awarded, an amount that is both statistically significant and economically meaningful.

How to explain this result? One possibility is that plaintiffs’ attorneys exploit opportunities to apply for and receive larger fee awards from judges with less experience and information. Whether or not that is true, the difference across courts and judges certainly indicates that the PSLRA is not working as intended. Both the plaintiffs’ securities bar and the investment markets operate nationwide; any plaintiffs’ law firm and any investor can appear in a securities class action in any federal district court. If lead plaintiffs consistently shopped for lawyers and drove hard bargains when negotiating ex ante fee agreements, one would expect to find fairly uniform fee requests and fee awards across courts after controlling for relevant case characteristics. In fact, there are statistically significant variations in fee requests and fee awards across courts. This suggests that experienced judges are playing active roles. They are applying downward pressure on fees while their less experienced colleagues are not.

The Role of Ex Ante Fee Agreements

We know from prior studies that cases led by public pension funds tend to have higher recoveries and lower attorneys’ fees than other securities class actions. Even in cases led by other kinds of plaintiffs, attorneys’ fees have declined significantly over time, suggesting that competition among lawyers for clients has reduced fees in securities class actions across the board. Most commentators assumed this was because these funds bargained hard with the attorneys’ representing them, but no one had yet tested that assumption.

What we found was surprising. In the vast majority of cases, there is no evidence that ex ante agreements exist. They appeared in less than 18% of the cases we studied. Evidence of those agreements was more frequent in cases with public pensions, but they were by no means universal.

The scarcity of ex ante agreements should raise concerns because the impact of those agreements on fee requests is substantial. At the mean settlement value, the average fee request in a case without an ex ante agreement is about 31% compared to just 25% in cases with an agreement, a nearly 20% reduction in the requested fee. In fact, the decline in fee requests when there is an ex ante agreement is about the same as when there is a public pension lead plaintiff. Each of these two variables is measuring the influence of an active, sophisticated lead plaintiff on fee requests. Together they suggest that this aspect of the PSLRA is working largely as Congress intended. Although public pension funds appear as lead plaintiffs less often than they could—they take the helm in only about one-third of the cases—when a real plaintiff plays a meaningful role in selecting and retaining counsel, some of the agency costs typically associated with securities class actions are reduced. And judges appear to respect those negotiations. All else equal, we found that the presence of ex ante fee agreements or a public pension lead plaintiff made judicially imposed fee cuts less likely. Ex ante fee agreements not only reduce fees, they make them more predictable.

Judicial Fee Cuts

Some level of certainty is crucial because judicial fee cuts are difficult to predict. We tried to do that with a logit model that used objective case characteristics, but we couldn’t. The easiest way to see that is to look at well our model predicted the cases where fee cuts actually occurred. Overall, we predicted 85.17% of the cases correctly. In the abstract that sounds like a powerful predictive tool, until one realizes that there were no fee cuts in 85.38% of the cases in the database. We would have achieved virtually identical results if we had simply guessed that there were no fee reductions in any of our sampled cases. Our model correctly predicted only one of the sixty cases in which reductions actually occurred. In other words, when it came to identifying the situations in which judges did cut fees, the objective case characteristics used by our model had no predictive value more than 98% of the time.

Why are fee cuts so hard to predict? One possibility is that judicial fee reductions are, for all intents and purposes, random events. The second is that other unobserved (or perhaps unobservable) case or judicial characteristics play the dominant role in determining whether the court will cut fees. Both possibilities raise serious questions about how fees are set in class actions. Some judges may believe they have the ability to identify and accurately reduce “excessive” fee requests, but if they do they are relying on criteria other than those we have tested. The standards courts explicitly employ contemplate that judges will assess ex post a vast array of factors, including how hard the attorneys worked, how much risk they faced, the quality of the results they obtained, and the relationship of the fee request to similar cases. Perhaps long experience with any given case or with securities class actions more generally provides judges information that is more nuanced than we can capture in our analysis. Indeed, that might partially explain why fees in high volume districts differ so dramatically from fees in low volume districts.

But we suspect that something more is going on. We speculate that fee reductions are primarily the product of subjective assessments by judges based on their own idiosyncrasies, biases, and heuristics rather than the objective facts of any given case. Temperament, judicial philosophy, experience with class action litigation, personal wealth, pre-judicial work history, personal earnings history, political ideology, and other individual and largely unobservable judicial characteristics may well—consciously or unconsciously—drive judges’ decisions. Indeed, the importance of such variables might partially explain our finding that in one-third of the fee-reduction cases the court offered no justification or explanation at all for its decision. When courts do offer an explanation, the most common one is simply that the requested fee is “too large.”

Is there a Way to Address these Concerns?

The unpredictability of fee cuts implies that, to a significant extent, fee awards are unpredictable too. This is likely true for all class actions, not just securities cases, because the doctrines and procedures that govern fee awards are largely trans-substantive. Uncertainty regarding fees is a significant problem because it likely discourages lawyers from investing optimally in class actions. This uncertainty ultimately harms both class members (by impairing the quality of the representation they receive) and the general public (by weakening the deterrent effect of the law).

In sum, there is good news and bad news to report. The good news is that by peering inside the black box, we learned that the participation of public pension funds and the use of ex ante fee agreements reduce agency costs in securities fraud class actions. They enable investors to get more “bang for the buck” in terms of the fees that class action lawyers are paid. The bad news is that institutional investors are too passive on whole: they assume the lead plaintiff role less often than they might, and when they do step up to the plate, they often fail to obtain written fee terms up front, even though ex ante fee agreements have been proven to benefit such plaintiffs.

These findings led us to propose a set of procedural reforms that would make securities class actions more efficient and bring their operation into closer conformity with the PSLRA.

1. The lead plaintiff should negotiate a fee when retaining counsel to handle the case;
2. The lead plaintiff should disclose the terms of the negotiated fee to the district court in camera when offering a law firm for appointment as class counsel;
3. The district court should review the negotiated fee terms before appointing class counsel and should uphold them unless they are clearly unreasonable or not the product of arm’s length negotiations; and
4. When reviewing class counsel’s request for a fee award at the end of litigation, the district court should apply the agreed terms unless unforeseen developments have rendered those terms clearly excessive or unfair.

We believe that judges could implement these reforms immediately, without further action by Congress or the Federal Rules Advisory Committee.

The only procedural step that entails any possible difficulty is the requirement that courts review fees when they appoint class counsel. Because plaintiffs have legitimate reasons for wanting to prevent defendants from acquiring information about their fee terms, this review would have to occur in camera. It would also require judges to consult evidence about fees prevailing in the legal services market. This could be provided by means of affidavits from pension funds’ general counsel describing the number of law firms that monitor the fund’s portfolio for signs of fraud, the number of firms that competed for the opportunity to represent the fund in the current litigation, the range of the competing proposed fees, and the reason(s) supporting the choice of the law firm offered to the court as counsel for the class. The affidavit might also describe the relationship between the fund and the law firm, including any political contributions that lawyers associated with the firm may have made to anyone associated with the fund, along with the fees the fund agreed to pay outside counsel when suing on its own in other matters. Information about fees could also be obtained from experts.

Over time, the body of evidence bearing on the reasonableness of ex ante fee contracts would grow, and judges’ initial fee assessments would become increasingly reliable. In cases with competing lead plaintiff candidates, judges could require all contenders to submit their fee agreements. This would increase the amount of information available and pressure all lawyers involved to offer market rates. Judges would also have a firm basis for cutting the fees of lead plaintiffs who failed to bargain hard by referencing fees negotiated in comparable cases. If this evidence were readily available, it is reasonable to expect that it would have the greatest impact on, and utility for, judges in low volume districts or judges with little or no prior experience with securities class actions.

Lynn A. Baker holds the Frederick M. Baron Chair in Law at the School of Law at the University of Texas at Austin. Michael A. Perino is the Dean George W. Matheson Professor of Law at St. John’s University School of Law. Charles Silver holds the Roy W. and Eugenia C. McDonald Endowed Chair in Civil Procedure at the School of Law at the University of Texas at Austin.