Have Attorney's Fees Risen in Class Action Settlements?

February 20, 2004

Unedited transcript prepared from a tape recording

Proceedings:
PROFESSOR EISENBERG:  [In progress.]  But we don't have a high threshold.  We just want a table.  Some numbers, other than page numbers, have to be in it, and this is the first issue, and one of the reasons I've been in the Times is I've actually been trying to get the journal's name out, so we can get subscriptions and papers submitted.

We're in very good shape.  The first year is basically full.  If everything goes according to schedule, we'll publish about 28 articles in our first year by very distinguished people from around the world.  So I would encourage you, if the paper is of policy interest to either you, as a practitioner or an academic, to go online, BlackwellPublishing.com/journals, and you can read all about it and subscribe.

And without subsidies from places like AEI or people showing that the market works and we really are interested in consuming empirical data about the legal system, we're going to keep publishing whether you buy it or not, but we can do a whole lot better work if the publisher is convinced that there's a market for what we are doing, and I think this group is one of the prime markets.  All right.  So much for the commercial.

Maybe the place to start is just with basic descriptive data.  Maybe the place to start is to start with the study we thought we were doing and not the study, not the views of it that probably brought me here.  Geoff Miller and I wanted to write a paper that would help judges assess fees in common fund cases, where there's a large fund recovered by the clients, and the attorneys always petition for fees, and the judge, a reasonably conscientious judge, might read three, four, five cases out there and say, okay, that's the going rate on these kind of cases.

We thought, by systematically reading and coding for factors about all of the cases that we could get our hands on for 10 years, we would be doing something of a public service to both sides in the litigation and especially the courts who would then have a benchmark against which to assess the fee requests in the case.

So our universe of cases is basically what we believe is the same universe of cases that judges have ready access to, and that is what they can find by doing diligently the research.  So, in that sense, for some purposes, people will try and use these data for things that they may or may not support.  I think for the purpose of finding the going rate in cases available to judges, this is the right universe.  This is the cases the judges will read.

It's a little bit like reading what should you read if you're interested in legal doctrine and the latest twist in legal doctrine at a high level.  Well, you don't have to read the massive cases, you have to read the opinions that shape legal doctrine.  Although published opinions are, in some senses, a nonrandom sample of cases, they are certainly the sample for many purposes, including learning about legal doctrine.

So that was our purpose and, of course, along the way when you're looking, we even put a table in the paper.  I asked Jonathan to hand out what I thought would be the thrust of my remarks, but in the full article we sort of put a look-up table for judges, where they can say, okay, here's the amount of the award, here's the basic going rate fee for that kind of award, and we hope to do that service.

In the course of doing that, of course, you find out other things.  One of the other things we found out relates to time trends and the like and loadstar versus percent recovery, and so those are some of the things I'll be working through.

But I think the place to start actually is with Page 8, maybe a comment on the low-tech performance, and that is Jonathan kindly offered to put a PowerPoint up on the screen if I sent them a disk.  I'm never that far prepared in advance.  I would have had to send that Tuesday, before leaving town for Atlanta, so I never got around to a PowerPoint.  My counteroffer was to do a series of transparencies, which are, in fact, a lot of the content of what you have in the sheets with the transparencies.   However, as disabled as I was at doing PowerPoint in advance, AEI was unable to produce an overhead projector.  So we're sort of even on the tech, and maybe you guys are the beneficiaries because you actually get something to take away rather than just staring at overhead transparencies.

But I would start with what's labeled AEI Handout Page 8, which is on the Eisenberg and Miller Page 51 of the article, as published in the journal just to get a feel for what's going on.

If you look at--maybe one of that things is should say is we use two major data sets.  One is our own, based on Geoff Miller's reading of every case he could get a hold of for 10 years, and the other is a group called Class Action Reports, which has been in the business apparently of collecting class action data for many years.  They, in one of their printed publications, just printed an enormous table of 1,122 class action awards, and with enough man and woman power, I had those data entered twice and then compared for mistakes, and so we have captured all of their data, we hope, for the same period we studied.  Their data actually goes back further than 1993, but I wanted sort of comparable years.

One of the I think comforting things about the study is Class Action Reports, which doesn't purport to limit itself to published opinions, but like the rest of us is probably going to get more of them than the unpublished variety.  Most of the results we find are pretty consistent, with their awards being a bit higher, their sample being a bit different, more dominated by securities cases than ours is.

But if we look at the actual numbers, Table 1, Panel A, one perhaps interesting thing worth noting is the totals row, and I think my remarks today will be mostly limited to the nonfee-shifting class of cases.

Fee shifting tend to be civil rights or consumer cases, employment discrimination cases where federal statutes or state statutes dictate the method of computing the fee.  And they also tend to be cases with much lower awards than the mass class action cases, and so the fees tend to be a much higher percent, and you can see that if you look around the totals row, that the mean fee in a fee-shifting case, in the right-hand columns, is 37 percent, the mean fee in a nonfee-shifting case is 21.9 percent.  So there's sort of heterogeneity in the data that require one to separately account for the two classes of cases.

Looking at the nonfee-shifting cases, we see that there is very little support for the sort of standard one-third notion of a fee going to a contingency fee lawyer.  The mean is 21.9.  The median is 23.2, and there is no class of cases, except the two civil rights cases, in which fees get to even the 30-percent range; that is, I think a fee of a quarter or a fifth of the client recovery would more accurately reflect what's going on in the world rather than the one-third.  Now, the one-third I'm sure gets paid a lot in smaller litigation, but these are big cases, and there's good reason to think that the percent fee would go down in these larger cases.

We also don't know what percent it is of the actual recovery to the client, since the fee is set before the case ends, and so you do or do not know--the judge has to set it based on the best information available, and perhaps sometimes with an opening-up clause if things don't go as expected.

You can see a pattern across different case categories.  I'm not sure to make too much of that because in the models we did I don't think that the case category played a major role in explaining the size of the fee award.  So that's a basic description of what's going on.  As you can see, securities dominate, in our sample, 142 of 303 were security cases, in the Class Action Reports' data, 483 of 630 were securities cases.  So it makes some argument for maybe studying securities separately from others, just as in tort I think it often makes sense to study automobile cases separately from the rest of tort cases because automobile cases are a huge fraction of the action, and yet there are institutional things in place that make automobile quite different than other tort cases.

On to page--the cover page is just the abstract.  A result that certainly caught the eye of the Times--I think they even bothered to reprint our graphs, which is sort of, if you do data, to have your graphs reprinted in the Times is sort of, wow.  It felt very good.  And what you see on AEI Handout Page 2 is the time trend in recoveries.  So the top graph is the client recovery, at least the nominal recovery, as reported by the Court as best we can read it.  We did not follow these cases to see what actually got paid.

And you'll see a much smoother line in the CAR data, in the Class Action Report data, that dashed line, because again they have a more homogeneous class of cases because it's dominated by securities, a few more peaks in our data.  I've forgotten the details, but the peaks actually represent a rather small number of cases which of course are important, but they are, in one of the peak years, it was one case with a multi-billion award and one of the peak years was about four cases that led to the peak.

And you can see the peaks become quite a bit diminished when you move to the second graph on Page 2, and that's the median award, and there aren't such sharp peaks.  And there's sort of an upward trend from, say, '95 to '99.  On the other hand, the median recovery in millions in 2002 was the same as it was in 1994.  So there's not the sort of monolithically increasing thing that one might have expected to see.  These are, I hope, correctly inflation-adjusted dollars that we're looking at.

Again, the CAR data and our data tell a similar story, a pretty flat trend over time in terms of client recoveries.  In some sense, this paper, talking about client recoveries is not the thrust of this paper, but it turned out if you want to explain the lawyer's fee, the single most important thing was the client recovery because the lawyers are getting paid roughly in proportion to the amount they recover from a client.

And if you turn to the AEI Handout Page 3, you'll see the time trend in the fee recovered by the lawyers, and there there's one peak in 2000.  I think that's a couple of very large cases, but generally no clearly upward time trend in terms of mean fee awards to lawyers.

And if you look at mean CAR fee, Class Action Reports' fee, it's again fairly smooth.  It tracks our data set very closely, except for 2000, where we obviously had a big case or two that they didn't, and that led to the peak.  Otherwise, actually, our lines strike me as remarkable close--the sort of small dashed line of the CAR data and the solid line of our data are sort of within an eyeball of each other for everything but that 2000 peak fee.

Median fee awards, that's AEI Handout Page 4.  Actually, not much going on here in terms of an upward trend.  I'm a little suspicious of the line that gets down so low.  I'd have to think about whether it's mislabeled or not, but I won't worry about that for now.  Anyway, no significant upward trend.

I said, for us, the major story that emerged from the data was, in explaining fee awards rather than time trends, was the incredibly strong relation between the fee amount and the client recovery.  And the scatter plot that is on Page 5 of the AEI handout, to me, is sort of the heart of the paper, that judges seem to be basically looking at the recovery and then picking  a fee based on some notion of the size of recovery.   We have the two regression lines in the picture are the best-fitting line for the nonfee-shifting cases, and the best fitting line for the fee-shifting cases, and you can see the relation between fee amount and client recovery is similar in both and I think a very tight fit.  To the extent there is scatter, you know, sort of isolated points in this figure, they tend to be on the low end; that is, a sort of outlier high award there almost isn't one.  The density of the data are pretty tightly packed above the regression line.  When you get below it, you have a lot of scatter.  So the courts seem to be using things to knock the awards down.  There's very little that knocks an award up so that it's literally off the line.  They're almost all very close to the line.

I thought another important result from the relation between fee percent, fee recovery or fee and client recovery is the scaling effect, which probably comes as no surprise, but that's illustrated on Page 6 of the handout.  If you look  at, say, the solid line, the solid-fitting regression line, the F's are the fee-shifting cases, the N's are the nonfee-shifting cases.  We fit the best line to the data, and one thing that emerges from all three lines, fee-shifting, nonfee-shifting and Class Action Reports' data is there is a discernible downward slope, and what that's saying is, as the client recovery goes up, the percent recovered by the lawyer, and here the Y axis is the percent not the dollar amount, as the client recovery goes up, the percent taken by the lawyers goes down, a very substantial scaling effect in all three subsets of the data we reviewed.

To me, there's no one thing that makes or breaks an area of legal doctrine, but having such a scaling effect I think, in class actions, is either you can view it as verifying the theory of class actions or as just simply a very efficient thing to have, and maybe those are the same thing.  But what it says is, if you aggregate things, aggregate actions that would otherwise be done separately, so that you have a larger recovery for the clients, the lawyers are going to walk away with less.  If you have $100-million claims litigated separately, the lawyers will probably walk away with a fee of 30 or 40 percent.  If you aggregate those $100-million claims into a $100-million claim, the lawyers are going to walk away with 10 or 15 percent.  And so, in that sense, the class action mechanism is working if the goal is to get to the victims the biggest possible piece of the recovery and have the lawyers take less.

Figure 8 is another illustration of the scaling effect.  I think it also gives the fee percent range, plus or minus one standard deviation.  So that, for example, if your recovery is in the $20-million range, which would be 15 to 22 on the X axis of the first part of the figure, if your recovery is in the $15- to $22-million range, as I read the graph, your fee should be about 22 percent, with a fairly substantial range, but that would be the midpoint of the fee recovery.

If you look at the low end of the axis, the low end of the X axis, where recoveries are what we label less than $1.4 million, fees are at 30 percent.  If you look at the high end of the axis where recoveries are what we label greater than $190 million, the fee award is getting close to 10 percent.  So you have a very substantial reduction as the recovery goes up.

Interestingly, the Class Action Report data I think show less of a scaling effect, though you can see indeed theirs show fairly level ranges of fee percents throughout the range until you get above $16 million.  Then, they also kick in and show a scaling effect at the very high end.

I think Page 8 we have already looked at.  Page 9, maybe one or two supplementary results worth stating, and one I'll state that's not in any of the models I've handed out.  Maybe the focus here should be on Models 3 and 4, the nonsecurities cases, just because securities have a little bit of a world of their own, and there's PSLRA.  Well, maybe start with the securities.

The post-PSLRA dummy variable in Models 1 and 2 is positive and significant in our data, suggesting that, if anything, fees went up after the products liability of the securities law.  There's so many reforms, it's hard to keep track after the securities law reform went up, although that result is not replicated in the CAR data.  So I'm not sure that we should conclude that it happened.  I think we could conclude we find no evidence that the PSLRA reduced fees.  If anything, they stayed the same or went up.

If you look at the middle columns, Columns 3 and 4, where we're controlling for a bunch of things about the cases.

First of all, notice the extraordinary significance of the gross recovery in explaining the fee amount.  Indeed, in Model 3, we get an adjusted R squared of 91.91, suggesting the model is doing a pretty good job of fitting the data, which would be quite consistent with that scatter plot we saw, where everything was tightly clustered around the line.

A couple of things were satisfying to find, and that is when the cases were read, and we sort of read into the cases or interpreted them as saying this is a high-risk case, we found a positive boost in the fee.  When we found a low-risk case, we found a negative boost or a negative hit on the fee.  I'm not sure why this is going on, particularly in light of worries about the state courts one hears.  But if you look at the last explanatory variable, the federal case dummy in our nonsecurities cases, it's positive and significant.

So that, although state courts are regarded as tort hell these days and class action hell, in fact, from these data, one finds that, holding constant all of the other things we can about the case, federal courts are actually awarding significantly higher fees than state courts in class actions, which I don't actually have a theory for why that is going on, but it is there.  Maybe it is just some spurious result, but it seems worth reporting.

The other result, I guess, again, returning a bit to less to our policy debate about class actions, but to our focus on sort of efficient calculation of fees, and this is in the paper, but not in any of the handouts.  And that is courts and lawyers often have to spend a lot of hours through one method of fee calculation.  That's the loadstar method, where the lawyers have to sort of figure out their hours, and the court scrutinizes it, and does a calculation and knocks them down usually a bit.  It's a pretty elaborate, time-consuming process, which I imagine judges just hate, the sort of judges serving as accountants checking on lawyers and actually looking at time sheets and things like that.

What we found in models in the article that I'm sorry I didn't put it in the handout is if you just forget about the loadstar method and just say what did the client recover, you get better models explaining the level of the fee award to the lawyer, which suggests to us that courts, which is I think a savings for all of us, courts can save some time, lawyers can save some time if judges were a little less reliant on the loadstar method and simply said, "What did you recover for your clients?  Here is the presumptively correct fee," and the loadstar method would be used less often, and yet we'd get a better fit to the pattern of awards that have happened in the past, and I think that could be a substantial savings in many cases.

I guess I am done.

MR. KICK:  Thanks, Professor Eisenberg.

For our first discussant, we're going to hear from Paul Rubin, who is the Samuel Candler Dobbs professor of economics at Emory University.  He also holds an appointment in the law school at Emory.  He's got specializations in law and economics, a new book in illusionary economics and industrial organizations.

Paul?

PROFESSOR RUBIN:  Thank you.  Thank you for inviting me.

My remarks will be somewhat schizophrenic.  Professor Eisenberg and I are both academics, and this is a paper published in a scholarly journal, but we are here in a policy institute, so my comments will be a mixture of academic and policy comments.

Most of my comments are about issues that are either not addressed or questions that are not answered.  As I was reading this paper, I started to feel like a referee more than a commentator because questions kept arising in my mind that I would have expected--in fact, I went back and looked.  It's a new journal, and I'd heard of it, but I wasn't too familiar with it.  As some of you may know, some law reviews are not peer reviewed.  They're student edited.  And so I went back and looked at the website of the journal and found that it is a peer reviewed journal, but, nonetheless, many of the questions that come to mind--I'm always glad to have a new journal.  It's good to have new journals, although, after today, I won't try to publish anything there, but at any rate--

[Laughter.]

PROFESSOR EISENBERG:  We won't let you.

PROFESSOR RUBIN:  That's why I won't try.

[Laughter.]

PROFESSOR RUBIN:  At any rate, many questions arose in my mind, and as I say, they are the kind of questions that I would have expected a referee to ask and would have expected answers to in the paper.  So I don't understand the refereeing process.  But I think another negative of having these questions is, if you read the article, they could have been oversights, but it makes it look like the article is more likely to find something in favor of the current status quo than might have been true.  It might not have been true because, as I say, we don't know the answers to the questions, but it would have been nice to have some information.  So that is the general tenor of my questions, and I will just go through them.

How was the time period chosen--1993?  I talked to my colleagues who are lawyers, and they said there was nothing particular going on in 1993.  There is no mention in the paper.  There may have been a reason why 1993 was chosen as the early date, but again there is no mention in the paper as to why that was chosen.  It gives us a 10-year period, which is a fairly short time.  For example, if these are class actions where settlements were too high, in some sense, it went up in the '80s and became too high in the '90s, and then leveled off, we could still have a policy problem, but a study looking at just a 10-year period might not pick up that problem.

I can quote from another paper in the same issue of the same journal that Professor Eisenberg edits.  Average jury awards tend to be highly variable from year-to-year, making it difficult to distinguish trends over relatively short periods of time.  That's an article that looks at a 40-year time period, as I say, they're both from the same journal.

So I think the time period is too short, but there may have been a reason for it.  Maybe Geoff just didn't want to read more cases.  On the other hand, the CAR data is a much longer period.  So the question arises either why wasn't a longer period chosen or, if there was a reason for the shorter period, why wasn't that addressed?

A second lack in the paper, and to me what seems to be really an important one is, we have all kinds of data on time trends in recoveries, average recoveries meaning median fees, nothing about number of cases.  We don't know if the number of cases has been--we don't have annual data on the number of cases.  Again, we don't know if it's been going up, going down.

And, again, there is no discussion in the paper of why that is not there.  It simply isn't there, at least I couldn't find it.  I had the paper on-line, so I could actually do searches, and I couldn't find any discussion of it, reading it and searching it.

The paper makes judgments about trends in fees and trends in relief, based on its own sample.  And based on that same sample, it seems to me it could have made some inference about or some evidence about the number of cases per year, a very important variable, one that is simply not addressed.

A third point.  In two places, there's quote on Page 28 of the full paper, "Contrary to popular belief, we find no evidence of `something,'" and then there's a later quote, "No real dollar increase in the level of fee awards in a major case over the course of a decade," is not the sort of fact we're accustomed to hearing.  Both of those statements are not sourced.  Lawyers are more famous than economists for giving sources, but in both of these cases, the authors are addressing what they view as the popular or policy perception of problems, but they are not sourcing that to anything, and that's important because there is one source given in Footnote--I've really spent too much time with lawyers when I start to quote footnotes--but Footnote 43, an article refers to, "A sudden rise in jury awards, as well as increased risk of class action."

So the one cite they do give as to what people are worried about, they seem to be worried about increased risk of class action--data that we don't have anything about because, as I just said, we have no evidence in this paper on the number of cases per year.  Increased risk would be number of cases.  Had there been better sourcing just to the popular literature, the things people are worried about, we could have known more carefully whether this paper was addressing those things.

At least the one example that's given, the paper doesn't address half the issues that the particular news--it's an economist.  It's a Financial Times article, but if we're using popular sources as what our hypotheses are, it would be good to know very explicitly what the hypotheses are.

The article says, "The presence of soft relief," meaning coupons and so forth, has no material effect on the fee.  That's an ambiguous statement.  What it really means is that soft relief is treated like hard relief, so that when the dummy variables are included in the equation for coupons, and even in one specification for questionable soft release, that dummy variable is not significant, which is another way of saying hard relief and soft relief, a dollar in cash to class members, a dollar in coupons to class members are treated the same in this paper.  There's no significant difference.  So I might have said that sentence somewhat differently, but the results in the paper are that, again, relief is treated the same way in both categories.

Another important issue is the difference that Professor Eisenberg mentioned as being ambiguous is state versus federal cases, and it's a puzzle, and he said the results may be spurious.  I have a hypothesis as to why they may be spurious, but I don't know either.  But I note that state cases are 11 percent of the fee-shifting cases, and 23 percent of the nonfee-shifting cases.  I think, as everyone I think would agree, in the world of all cases, more are state than federal.  So this is a biased sample, and it's not clear what the bias is, but one suggestion made is that the bias comes from the fact that these are published opinions, and in state cases, they are more likely to be published if they are appealed.

So that raises, in my mind, the possibility that a higher percentage of the state cases are appealed than are the federal cases, and the percentage of appellate cases is  12 and 16 percent, fairly close to the 11 and 23 percent.

Now, I don't know what statisticians would call multicollinearity, whether there's a correlation between the appellate cases and the state cases because, again, no mention is made in the paper of whether there is such a correlation.  It would have been a nice thing to see.  We might expect such a correlation.  If there is, then results become uncertain and flip around, and you can't explain what's going on, which is consistent with Ted saying that it may be a spurious result, but you simply can't tell.  There's no statistics reported that would have enabled you to determine whether there is this correlation.

Another point, there's a safe harbor proposed--not quite a safe harbor, but a table with the range--and it said, well, if attorney fees are within this range, then they are reasonable.  I suspect that the maximum of this range will soon become the mean, that this paper published, and cited in the New York Times, will become a standard document for class action attorneys, first of all, to use in calculating their fees--what does Eisenberg say we can get--and, lo and behold, I think that will be the number.  And then I don't know whether Ted was proposing updating the number.  If we update it, it should continually rise because I suspect that the maximum will always become the mean as time goes on.

And then there's a point made in the paper that there's an efficiency for class actions that legal fees are smaller, so larger classes are more efficient.  That's true if we take the hypothesis that the cases would have been brought anyway.  Of course, one theory behind class actions is that it enables lawyers to bring cases that would not pay, and then you have to go to the question of are these cases socially useful in the first place.   I've worked on some class actions, some of which I thought were socially useful, so I'm not saying that no class action is socially useful.  On the other hand, I don't know if it's in his sample, but the Milli Vanilli case always struck me as one that didn't have a whole lot of social benefit, and I think there are other class action cases that don't.  So it seems to me, at least, you have to address the question of it's efficient to abrogate the cases if the cases are efficient.

In terms of that, I think it was interesting to note that the mean recovery was $100 million in these cases.  One hundred million dollars is a pretty big number.  In the federal system, in the regulatory system, $100 million makes a rule classed as a significant rule if it has an annual impact of $100 million.  Under that circumstance, OMB, first of all, the regulatory agency examines it, and then OMB examines it, and there's cost-benefit analysis, there's requirements to look for cheaper ways to achieve the same goal.   So a good deal of analysis goes into a $100-million item if it's passed as a regulatory rule.  Here, we may have one judge in Madison County deciding on something that's $100 million.  So the fact that the level of these cases is--the median is much lower, I would agree--but the mean is quite a significant number, a number that has some significance in the federal system, and I think we have to think long and hard about the fact that we are doing things in the court system, with impacts that were they done in the regulatory, the Executive Branch, would take a whole lot more analysis and a whole lot more effort than actually goes into them in the Legal Branch, the Court Branch.

Thank you.

MR. KICK:  Thank you, Professor Rubin.

Since we are running a little bit ahead of schedule, I think I'm going to let Professor Eisenberg address those comments, since I sort of expect that Mr. Beisner's comments will be a bit less technical, and before I turn it over to Professor Eisenberg to reply, I have one technical comment on my own.

I wonder what happens if you re-run your aggressions using a frequency weight.  Presumably, the cases aren't uniformly spread throughout the years, and so it would be interesting to see what happens if you use a frequency rate to run your regression, and see if that changes any results particularly.

Thanks.

PROFESSOR RUBIN:  One more comment I meant to make.  I found these points, I'm not an econometrician, had the paper been refereed by an econometrician, I suspect there would have been other things such as Jonathan's point here.  So I think that's a relevant--

MR. KICK:  The idea being you don't have uniform distribution throughout the years, and if one of the particular things that you're interested in is what's happening by year, what happens if your data are clumped in a particular year?  Each of those is getting weighted at the same weight as observations in other years.  And so if they're--

PARTICIPANT:  It wouldn't be clearly right to try and adjust for that.  It might be right, but it might not be right.

MR. KICK:  Well, it would be nice to see for robustness.

PARTICIPANT:  Yes, exactly.

PROFESSOR EISENBERG:  Ten years of data on every case that one could find, if one were a judge, strikes me as much more than one gets in the court specifically stem for most legal decisions.  So I think 10 years was arbitrary in the sense that it's a round number, and it seemed like a reasonable place to start.  I guess I fundamentally disagree with Paul that that's too short a period to see anything.

Tort reform advocacy and advertising is premised on continuously and ever-increasing awards.  If you don't see them over a 10-year period, that advocacy and advertising turns out to be false.  So I think it's right that it would be preferable to have 40 years of awards.

One of the reasons the RAND study that Paul refers to as having more years than we did, it's the only 40-year data set I know of about the legal system, in terms of jury awards, and it's very valuable because it goes 40 years.  Of course, it only covers two jurisdictions--Cook County, Illinois, and San Francisco, California.  So that had we had 40 years on two jurisdictions, Paul would have sat here and said, "Why do you only have two jurisdictions?"

So there's a tradeoff between depth and breadth, and I think we struck an appropriate balance, given the research question we were asking, which was how can we help judges set fees, not always studying everything that's ever been done in class actions.

On the time point, I don't know the answer off the top of my head, but I think a little more careful reading of the tables would have helped Paul.  If you look at Panel B--well, you can't look at it--Panel B, Page 60, we give the mean and median year in the data.  Now, it's possible everything clumps in one spike in one year, but the median year is 1998--

PROFESSOR RUBIN:  You don't give the annual.

PROFESSOR EISENBERG:  Right, no, we don't, but we say--

PROFESSOR RUBIN:  So you can't tell the trend in cases.

PROFESSOR EISENBERG:  Well, we can say about half the cases were after '98, and about half the cases were before, which is, I admit, not year-by-year, but it's hardly the notion that things are all clumping in, say, the last year or two.  And, of course, we could break it down if we had known that it was important, but we were trying to give a 10-year sample and treating it as a lump.

We also have, I think, which would also I think undermine the assertion that the results would be compromised by accounting for years, and  now I think you do have it in front of you somewhere.  If you look at the regression, for example, in the handout--maybe it's not in there--we have an age variable, which isn't quite the same thing, so maybe that's a bad point.  I'm pretty sure the regressions had, at some point, a year variable, and had it been significant--there it is.  There it is.

PROFESSOR RUBIN:  But not for number of cases.
  PROFESSOR EISENBERG:  No, not for number of cases, but whether it affects the points we're making, which is the size of the fee and the relation between the fee and the award, we did look at the year variable, we did not ignore it, and we found that it didn't affect--

PROFESSOR RUBIN:  That, I understand.

PROFESSOR EISENBERG:  Let me go on to your other points.  So I think the number of places aren't clustering in one year, since the midpoint is around '97 or '98, based on the table.

 Paul is critical of not having a source of the upward--the claim that things are going upward in the tort system over time.  We don't cite to source that.  I confess error there.  If you think things aren't going upward in the tort system over time, fine.  Withdraw tort reform proposals, and let's move on.  I thought we were hitting what is a commonly accepted perception of the legal system is that things are going up, and if you need more newspaper cites, I'm sure there's half a dozen or more.

I think Paul is right.  It would improve the precision of the point if we had cited more sources for this impression of the legal system.  I think they're available.

I think it's a very good point that we need to know the number of class actions.  I don't think it's terribly important for our research, in terms of helping judges set fee awards.  I think, as a policy matter, it would be very important to know how many class actions are going forward, how many class actions are being filed, but it would also be important to know how many are being certified and what happens to them because the raw numbers can be quite misleading.  Anyone can file a class action, and I suspect most of them get filtered out, so that if the system is being biased towards spurious class actions, one would expect judges to take care of that.

The fact that soft relief had no effect, I'm not sure what to make of it.  I think it's because soft relief isn't that common, at least we didn't find it to be that common in these reported opinions, and it's also, you know I think there's some subjective judgment that we probably had to exercise in deciding whether something was soft and whether to call it good soft relief or bad soft relief, and those are not sort of self-defining terms.  And I think the defense bar and the plaintiff's bar would have vastly different views as what counts as beneficial to the clients.

State versus federal.  I think there is more to be done there.  If state courts don't publish their opinions, we're limited, and even state courts are looking to published opinions to get information about what the going rate is, so I think we were serving our research need there.

I think it's a fair point to ask would the cases have been brought anyway, and of course a very important point to ask which class actions are socially useful.  And I think it would be important if the debate about the utility of class actions could focus on the mass of cases and whether they are socially useful or not.  What the debate tends to be is someone trots out an anecdote, Milli Vanilli, gets a chuckle, and then hopes that the audience goes away thinking, "Wow.  This is a really screwed-up area of law."

Whereas, the whole point of this type of research is to avoid the anecdote, avoid the Milli-Vanilli sort of cheap shot and see what's really going on in the system because the really frivolous cases tend to get taken care of.

I guess I've covered most of it.  Sorry. Thank you for the comments.

PARTICIPANT:  One quick point--

[Audio break.]

MR. KICK:  Thank you.  Our second discussant will be Mr. John Beisner, who is the head of O'Melveny & Myers, firmwide, 120 attorney class action practice, and specializes in the defense of class actions, mass tort matters and other complex litigation.

MR. BEISNER:  Thank you, Jonathan.

Well, I wanted to start by complimenting the study.  I think that this effort should be applauded.  I think that, Professor Eisenberg, as you said, there's a lot of speculation out there about fee amounts in class actions, and I think this effort to look at published opinions and get a sense of what's going on out there in the real world is to be applauded.

And I think with respect to the analysis of the federal courts, I think it's an appropriate approach to learning what the fee amounts are out there.  I have some concern, as I think both of you suggested earlier, that this will become a document, and the charts will become a place where courts will go and that these levels of attorney's fees will be locked in.

I have a feeling that the percentages that are listed here are really higher than they should be because they don't reflect the sorts of efficiencies that were being discussed earlier.  These are in the range of what I think you would expect to find in individual settlements, that is, individual claimed settlements, and if there is to be any truth to the concept of efficiencies in litigating class actions, one would logically expect those percentages to be somewhat lower business of the inherent inefficiencies.

And one lawyer litigating a large number of claims simultaneously, you would think that the percentage would be lower that would be awarded in these cases.  But, again, I think that with respect to the federal court claims, it's an appropriate analysis.

I am gratified to see that it certainly suggests that the federal courts have been in sort of a "steady as she goes" mode in recent years and that the percentages are not increasing, even with the influx of securities class actions in federal court.

The problem I guess that I have with this comes with the attempt to compare in the study what is going on in federal courts with what is going on in state courts.  I think a lot of the attention in the New York Times article and elsewhere has been on that piece of the study, and I think that there are some real issues as to whether the study appropriately analyzes that comparison.

Basically, where this has come up is in the context of the pending class action bill that would expand federal jurisdiction over interstate class actions and would allow removal of more class actions in federal court.  There has been a suggestion that has been written, and this study indicates that Congress should reject that concept because the federal courts, as the study suggests, are awarding higher attorney's fees in our state courts.

And one of the motivators by many of the sponsors for the legislation is the concern that state court attorney fee awards in these cases are to high and that they wish to move the cases to federal court where they will be handled more appropriately on the fee subject.  So this study, and the conclusions in it, suggesting that the federal courts are awarding higher fees has been the subject of a lot of discussion.

The concern that I have with the analysis on federal versus state really goes to the data that were analyzed here.  The article abstract at the front of it, if I can get the computer to work here, says that this is a study of two comprehensive class action data sets, covering 1993 to 2002.  As Professor Eisenberg described earlier, there are really two sets here.

The first set is cases that were pulled off of Lexus and Westlaw, using a uniform research approach, a search logic, and there are 362 cases pulled from Lexus and Westlaw, using that research methodology.  And as you noted, there were 630 cases that were published in a compilation in Class Action Reports.  So those are the two data sets that we're looking at here.

The question I guess that I want to raise, though, is comprehensive an appropriate description of either of these databases, when it comes to a federal versus state comparison.  I am not questioning the comprehensive nature of these with respect to the federal cases in isolation, but I am questioning whether they are sufficiently comprehensive to allow a valid analysis with respect to the state court cases.

Part of the reason that I am raising this issue, if you look in the article, the authors' hypothesis is as follows, and that is that fees may be higher in state courts because counsel may be able to file in remote jurisdictions with few judges and significant potential home court advantage.  These attorneys likely select state court jurisdictions that they believe will be generous with fee awards.

I should also note, in fairness, that this same section of the article talks about the state court fees potentially being higher because of the risk of a reverse auction--

[Tape change.]

MR. BEISNER:  --on the same subject.  Some counsel will go to a court that will readily approve a settlement and award fees so as to achieve for the defendants the lowest possible payout to settle the entire litigation, and I think that's part and parcel of the same idea here, that the hypothesis or the suggestion is that there are some state courts that would be more receptive to approving a reverse auction settlement.

So one would expect, coming from this hypothesis, that the data analysis here, the data collection effort would be focused on those courts which have become, in some circles, to be known as magnet state courts that are tolerant of those sorts of settlements.  And the concern I have is as follows:

If you look at the Westlaw and Lexus research source, one of the problems with Lexus and Westlaw is that neither contains trial court decisions for most jurisdictions.  And so if you look at the research that was done on Westlaw, it has the federal court decisions from all 51 states and the territories included in the research.  As I said earlier, the research is a pretty comprehensive review of what went on in the federal trial courts.

But if you look on the state side, there are only a handful of jurisdictions whose trial court decisions are reported on Westlaw and Lexus at all.  For example, if you look at what Westlaw and Lexus, there are no, absolutely none, none of these decisions from Alabama's state trial courts.  They are not recorded on the system.

So the concern I have is that if you look through the list of what was included in the study, and you simultaneously think about the state court systems that have been identified in other research as being magnets for class actions; in other words, the places that do exactly what the authors said they were concerned about would skew the numbers to suggest higher fees in state courts, those states are excluded from the research.

Alabama, which there's been at least one study I'm aware of is viewed as a magnet state court, is not part of the data.  Arkansas, which I think, at least anecdotally among practitioners, has become increasingly a hotbed for class actions, no data.  California, if you look at the data compilations on settlements that have been questioned by the House and Senate Judiciary Committees, many of those are California, not part of this data.

Florida, identified in an article that I coauthored as a magnet state court, not included.  Illinois, Madison County, perhaps the most notorious for approving settlements that are viewed by many as abusive, benefitting primarily lawyers and not class members, not included.   Louisiana, in that same category.  Oklahoma, which has been increasingly a hotbed, not included.  South Carolina, Texas, which is another jurisdiction that some research that we have done indicated a very high number of class actions filed in that state.  The trial court decisions, that is, decisions approving settlements, not included in the research.  The same is true of West Virginia.

So I really question whether the authors have really looked at any data or any significant amount of data that would address the hypothesis about the magnet jurisdictions that are most likely to approve settlements that have very high attorney's fees in comparison to the real benefit for the class members.

Now, the other set of data that was used in the study, the Class Action Report data, presents similar issue.  The total number of cases in that study, 630.  Only 67 of those were from state courts, around 10 percent.  And you have the same problem here as well.  Those data, if you look through all 67 of the state court settlements included in there, they don't include cases from the magnet court jurisdictions we were talking about.  I think there is one Madison County in there, but if you looked at it, it really wasn't a settlement at all.

And Jefferson County, Texas, another magnet jurisdiction, has approved a large number of settlements in which the class members receive little or nothing and substantial fees went to the attorneys involved.  None of those were included in the study.  And of the 67, I guess another interesting fact, is 10 of these come from the Delaware Chancellery Court, and I think that is significant, at least in terms of federal, state analysis, because these are largely again securities cases that have their own particular baggage coming with them.  And those sorts of cases that were in the Chancellery Court are not going to be the sort that are removable under the Class Action Fairness Act.

But perhaps the most interesting piece of the Class Action Reports' data is, as the article note, those data from the Class Action Reports, when they compiled those data, they excluded any class settlements that were coupon settlements.  In short, the cases in which you have the greatest risk of the lawyers getting all of the money from the settlement and the consumers getting little or nothing weren't included in the study at all.

Now, if you look at the analysis of the Class Action Reports' data, as I said, the total number of state court cases were 67.  If I'm reading the analysis correctly, the analysis was that the mean attorney's fees from those settlements were around 27 percent.

But what I found interesting, in looking at the study, is there is another compilation of state court settlements out there.  Indeed, one piece of it is cited in the article, and those are the House and Senate reports concerning the class action bill that has amassed large numbers of state court settlements that Congress has found to be, that those two committees have found to be abusive.  They are not in here.  You can't find them in the study.  They were excluded from the analysis altogether.

And if you just took 50 of those settlements, and there's many more than that that have been documented in the Congressional reports, and you added them in to the data from the Class Action Reports' materials, the mean attorney's fee at that point moves up to around 58 percent.  In other words, of the recoveries obtained in those settlements, if you add in just 50 of those that Congress has identified in their materials coming out of hearings and other investigative efforts, it means that close to 60 percent of the recovery in the cases went to the lawyers, not to class members.  And there are many more that could be included that would move that number up even higher, I think raising real questions about any conclusions that state courts are managing the attorney's fee issue better than federal courts in terms of the amounts that are being awarded.

Now, again, let me make very clear, I would not advocate an analysis that would view the congressional compilations as being comprehensive either.  That was an effort to amass class action settlements that members of Congress who have been working on this effort viewed as being abusive.  It is not comprehensive either.  But I think that this data set, these two data sets that were used in this study, can hardly be called comprehensive either because they excluded all of the abusive settlements, as far as I can tell virtually all of them that have been identified by Congress through their efforts.  That whole body of data has been ignored in this analysis.

So I guess the question I have is whether the conclusions of the study, with respect to the federal-state relationship, were appropriate.  One conclusion is we predicted that fees as a percentage of the recovery would be higher in state court class actions, and in federal courts this prediction is not confirmed by the evidence.  If anything, the opposite is true.  And the other finding at Page 77, "We find evidence that fees tend to be higher in federal courts than in state court in nonfee-shifting cases."

Again, let me say, going back to the authors' hypothesis to start with, and that is that fees may be higher in state courts because counsel may be able to file in remote jurisdictions with few judges and significant potential home court advantage.  These attorneys likely select state court jurisdictions that they believe will be generous with fee awards.

My respectful submission on this is that hypothesis has not been tested by the study because there was no effort made, in the end, to actually look at those jurisdictions and get a handle on what was going on with settlements in those states.

MR. KICK:  Thank you, Mr. Beisner.

We'll let Professor Eisenberg address those points before I open it up to questions from the floor.

But tacking onto Mr. Beisner's point, this isn't exactly a criticism of the Eisenberg and Miller study, but more maybe a criticism of interpreting as being there are no trends.  Perhaps a more powerful test, which would clearly mean more data, but a more powerful test would look for safe specific trends because presumably there are some states that aren't offenders, and they'll actually be doing things to cut back on the attorneys' take in class action settlements.  And those really probably aren't what people are thinking of, those states aren't what people are thinking of when they complain about any tort crisis or lawyers' problems.  What they're thinking about are certain offenders, and so it may be more probative to look for state-specific trends, and particularly at the states that are thought to be these magnet courts.

Professor Eisenberg?

PROFESSOR EISENBERG:  Yes, I'd love to see the other systematic research showing magnet jurisdiction because I am not aware of that study.  I am aware of anecdotes gathered by Congress.  I don't think it goes much beyond that.

We've been through this tort hell and magnet jurisdiction debate before, and I think the first one Mr. Beisner mentioned was Alabama.  Alabama was reputed to be tort hell for punitive damages for years.  It was the poster child.  It was BMW v. Gore, and because of that, when they did one of the most comprehensive studies of punitive damages, the RAND Institute of Civil Justice went out of their way to include Alabama.  And what they found was, there was very little extraordinary about punitive awards in Alabama.  When one didn't cherry-pick anecdotes, but when one looked at the mass of cases in the punitive world, Alabama was quite normal.

So I come in, with this history, quite skeptical of claims of tort hell, magnet jurisdictions and the like, at least until it's based on comprehensive research.  If the policy debate is this study doesn't do comprehensive research and should be ignored, the concession that Congress hasn't done comprehensive research, and therefore we don't really know what's going on, it seems to me is very hard to ask Congress to spend time changing the law, when we haven't yet figured out what's going on.  We have a few anecdotes from a few places of bad settlements, and that doesn't make a case for taking up the valuable time of the Congress with reforming the law.  We need more data.  Perhaps that's right.  We should do those studies.

Magnet jurisdictions are a funny thing.  And, again, I am not sure they are proven.  I am not sure how you would prove it.  Consider the following.  Someone goes to Madison County, Illinois, and wins a big case.  There are lots of other potentially winnable big cases out there, but they see a winnable big case, and they say, "Well, we might as well bring it in Madison County, Illinois.  They did well."

So, if those cases flock to Madison County, Illinois, I guess it's fair to call it a magnet jurisdiction, but it could also be, hey, we saw a lot of cases that would have been filed elsewhere are flocking there for the same reason that defendants move cases to federal court, because they think they might have a slight edge, but it doesn't mean necessarily that Madison County is generating lawsuits.  It may become a magnet, and a magnet doesn't mean a net increase, it means an agglomeration in a place, maybe for good reason, and if it's abusively filed there, there are probably ways to deal with that.

One of the striking things about the magnet jurisdiction hypothesis is I think I missed--I didn't get them all, but you listed Alabama, Arkansas, California, Florida, Louisiana, Oklahoma, Texas, West Virginia.  It reminds of the sort of cross-examination question of an expert.  How many other experts have to be wrong for you to be right?  How many states have to be tort hell magnet jurisdictions before we say, well, gee, maybe they're not all out of line.  Maybe there is no tort hell.

If one needs to list 10, and these are not necessarily small states, if 10 major states are all tort hell, maybe they're not the outliers.  Maybe that's the norm.   How many states have to be getting it wrong for the class action bill to be getting it right?  It seems to me, again, that burden hasn't been established in the limited research, noncomprehensive research, done in support of the class action bill.

I think it's very important to cut back on abusive settlements, but the argument strikes me--I'm not a class action lawyer--the argument strikes me as very one-sided.  Suppose we take the worst case, coupon settlement, no cash to the clients, the lawyers walk away with 60 percent of the value, magnet jurisdiction tort hell.  What would have happened without a class action mechanism?

Until you address the social utility of class actions, one has to say, okay, small settlement, coupons, is the defendant deterred from engaging in cheating behavior?  If the answer is, yes, then the 40 percent that the clients are getting is 40-percent more than zero, which is what they would have gotten without the class action mechanism, and maybe the transaction costs are high, but where should the loss lie?  With the cheating company who relies on cheating a lot of people out of a little so that no one can agglomerate cases to bring an efficient lawsuit or should the loss lie with the consumers who are defrauded?

And if the lawyers, and if you have to convince lawyers to take 60 percent to get it, fine.  I'd rather have the cheating company pay than have the consumers bear the whole loss.  And if it takes giving lawyers some piece of it, that's fine.  Part of the system is to deter cheating, and, you know, read your phone bill.  See how many things in it are fully disclosed at the time you signed your cell phone contract, see how many fees are added that you probably don't even know about. And there's massive cheating going on in little pieces, enough so that no individual lawyer could bring the action.  And the transaction costs of small, little cheating are going to be high.

So it's not all clear that it's socially inefficient to have that added deterrent, even if it means lawyers are getting 95 percent of the take.  I mean, that's extreme, and I don't think it happens very often.  I don't think these valueless settlements are very common.

So, when Congress goes back and does its proper homework and studies the tort hell in detail and sees not just the extreme cases in those places, maybe they'll find out that in these places 95 percent of the class actions are fully valid, socially useful things, and the 5 percent are the outliers and not a reason for federal involvement in a traditional state area.

What's to me sort of shocking about the tort reform movement in this area is they have the resources.  Yes, there's rotten cases.  They know about them.  They know about the massive cases.  Let's do the comprehensive study so that policy can be made not on the basis of headline-grabbing anecdote with the worst-possible case from Madison County, Illinois.  Let's do it on the basis of full study of the good cases, and the bad ones, without filtering.  We don't have enough state cases in here because we couldn't get at them efficiently.   But I suspect the industries funding this bill or backing this bill have the resources to do the job right, and why don't they do it?   Why just throw out anecdotes and outliers and not the massive cases in each tort hell.

Let's take one tort hell: Madison County, Illinois.  We know that class action filings have increased there.  We also know, at least according to the St. Louis Post Dispatch, that certifications are not very high there; that, in fact, the judges in Madison County, Illinois, seem to filter out the massive class actions and not certify them.  What happens to those class actions in Madison County, Illinois, other than raw filing numbers, seems to me a bit of information that we should have some detail on before we pronounce it tort hell.

I guess just statistically right, if you have clusters of cancer, you at least go to the trouble of studying whether the cluster place is statistically significantly different from the rest of the country.  There are going to be clusters of high awards, and they're going to have an effect in attracting other awards, but it's not at all clear that's pathological because some place has to get the big case, and they are going to be a natural magnet after that, and people are going to struggle to have jurisdiction there, but I'm not sure that's evidence of pathology, other than evidence of sort of things tend to cluster in life.  It's not just, you don't expect a uniform distribution of high awards across the United States.

MR. KICK:  Thank you, Professor Eisenberg.

Now, we're going to spend the rest of the time taking questions from the floor.  But just as a comment on that last set of responses, I think it is true that we really do need to expend more effort setting the microfoundations of what we call the tort hell by doing things like Professor Hellen's [ph] study on whether elected judges behave differently than appointed judges and things like that.  They will give us a better idea of what proper policy responses are, if any.

And even farther down the road, if we do get into the point where we have got good data, and we need to worry about self-selection problems of people sort of shopping for the right forum, that kind of institutional work will be necessary to be able to solve those statistical problems.  So, hopefully, we can, in the next generation of this research on tort reform or tort issues, hopefully, we can get to the bottom of some of those issues.

So now we'll open up the questions to the floor.  People wait for the microphone, state your name and affiliation.  Does anybody have any questions?

MS. BACH:  Hi.  Pat Bach [ph], AFTC [ph].

Mr. Beisner, did you get to the 50-some percent by adding in the coupon [audio break]?

MR. BEISNER:  Yes.

MS. BACH:  So basically you valued them at zero and then the differential was--

MR. BEISNER:  Yes.

MS. BACH:  So you would expect a distortion then.

MR. BEISNER:  That's why I'm saying--I'm not saying that is an appropriate analysis for that.  I think the approach in this study was to value them, to the extent there any in there, and it's hard to see that there were many there at the value attributed by the court, which may or may not be the real value to the class members.

MS. BACH:  So the difficult thing is really, I mean, those cases are, in some measure, sui generis.

MR. BEISNER:  Right.

MS. BACH:  We have a valuation problem at base with those cases.

MR. BEISNER:  Right.

MS. BACH:  So if you take it out--all right.

The other question is, to only comment I had was I thought Geoff Miller had done precisely some of the work you're talking about in the Tulane series on disproportionate filings in Madison County and elsewhere.  He found that there was, I think--and I don't have the study in front of me [audio break]--he found that there wasn't as extreme a dislocation in the filing rates and in the [audio break] recoveries.  Anyway, I commend that series to you.  It's on the Gulf States, a Tulane [audio break], Law Review articles, I believe.

PROFESSOR EISENBERG:  I think the case for filing in Madison County may be--Geoff wrote a little while ago, I think in Tulane.  I think I was in that issue.  I can't remember now, but I think the case for increased filings in Madison County is pretty strong.  I don't think that's--

MS. BACH:  [Off microphone.]  [Inaudible.]

PROFESSOR EISENBERG:  But I think looking at filings in the legal system is not unimportant, but to accuse courts of sort of misbehaving almost without looking at what happens to the filing strikes me as questionable, and I think I don't remember what Geoff found.  I honestly don't [inaudible] if he were here.

But what the St. Louis Post Dispatch reported was certifications were much lower than filings, and one can't just stop that.

MR. BEISNER:  Let me just agree with that.  I don't recall, and I talked to the St. Louis Post Dispatch a fair bit about this, any article in which they had looked at certification for counties.  In fact, several months ago, there was a hearing before the Illinois State Senate Judiciary Committee, at which all of the primary class action counsel who litigate in Madison County was present.  And one of the members of the committee asked that assembled group, who I can assure you accounts for about 90 percent of the class actions that get filed in Madison County, if they could recount any case in which the court had denied a class when it was moved for?  And there was a long conference, and a lot of conferring about that.  They finally came up with one case in which there had been a denial.

I litigate a lot in Madison County, and I can tell you that the class certification is guaranteed if a case is filed there.  There is a tendency of that court not to move very quickly.  And so I have seen some statistics out there saying, well, of the 100 cases filed in a given year, only 20 of them certified.  Yeah, but how many have been denied?  You're not going to find many.  And so I just think that's a red herring because that is not consistent with what's really going on in Madison County.

PROFESSOR EISENBERG:  Let me refer to the St. Louis Post Dispatch article, which I happen to have.

PARTICIPANT:  [Off microphone.]

PROFESSOR EISENBERG:  January 11, 2004.

Again, it sounds to me like the necessary study just hasn't been done, but certainly they would disagree with your characterization of automatic certification.

A tally done by it looks like a local law firm that files them, puts the number of class action certifications as low as 23 between 1999 and 2002, about 12 percent of the 189 class action cases filed in those three years.  In the same time, according to the study, judges decided against jurisdiction for 47 cases.  I haven't been to Madison.  You probably know more than I do.

MR. BEISNER:  I've seen that, and it's really a study of how many in which they have granted certification.  The fact is they're slow, and they haven't gotten to that many.

PROFESSOR EISENBERG:  But they denied it in 47, and you said there were none.

MR. BEISNER:  That's just not accurate.  It's not accurate.

PROFESSOR EISENBERG:  You've done that study?

MR. BEISNER:  I've been in so many of those cases, I can tell you there are not 47 that have been denied there, and why would those counsel, when specifically asked to name cases before a hearing, not be able to come up with them?

PROFESSOR EISENBERG:  I don't know.  It sounds to me like the study should be done, but it can't be based on one lawyer's experience.

MR. LOCKE:  Paul Locke, University of Maryland.

Let me ask a question just for information.  Coupon settlements are taken as being of essentially zero value.  I take it this is where you have a case, well, let's say against an automobile company, and you're allowed to buy a car for $3,000 off or something from that manufacturer.  Has there been any discussion of allowing those to be sold to make them worth something?

PARTICIPANT:  There has been.  Let me start off with an initial point.  I don't mean to paint with too broad a brush on coupon settlements.  There are some instances where they do convey value.  If you have a defective product, and if you get a coupon to replace it, that is obviously something of value.  Those are pretty rare in the collection of these.  Most are not transferrable, and many of them, for example, the one that gets a lot of attention is the Blockbuster Video settlement where coupons were handed out for discounts on future rentals of videos.  Do they have value?  I guess at some level, but the defendant admitted in that case that they give out coupons all the time anyway.  There's no additional cost of that.  Would making those transferrable help?  I'm not sure it increases the value because you can probably get the coupon in the store as well.

MR. LOCKE:  No, I meant the automobile one.

PARTICIPANT:  The automobile one, typically, those are not transferrable.  There are some that have had limited transferability over time.  The CK truck settlement negotiated in Louisiana by GM did have some transferability to it, but that's not typical in these, and that's jut not what's negotiated or demanded by the courts.

MR. LOCKE:  But it seems to me, I mean, if the lawyers represent the client, to some degree, that would be an issue that they would raise as a way of increasing the value of that unless, to put it very crudely in terms of the way that it is often presented, unless the lawyers really are pirates, and they're not interested at all in their clients.

PARTICIPANT:  That's the scenario that's out there.

MR. KICK:  Any other questions?

MR. MULHOLLAND:  Joe Mulholland [ph], AFTC.

I was curious whether on this business of trends, which I suppose is very hard to get a hold onto, especially if the state courts are underrepresented, and I as I understand it, I think there's been a shift of more cases to the state courts.

But I was wondering about insurance premiums.  Is that possible?  I assume you can do that, say, in medical malpractice.  You can get a trend in insurance premiums over time to give you some idea on the movement, but is it possible in these class actions?  Would there be any type of litigation insurance or anything that could be reflected of trends?

PARTICIPANT:  I'm not quite sure I understand, that there are changes in insurance rates as a result of litigation?

MR. MULHOLLAND:  Yes, right.  I mean, presumably, the insurance rates will go up when there's a great risk of litigation.

PARTICIPANT:  Would you be looking for like Madison County specific increases or Illinois?

MR. MULHOLLAND:  No, I'm not sure.  I'm just thinking generally.  I assume--

MR. BEISNER:  Are you thinking of class action insurance?

[Simultaneous conversation.]

MR. BEISNER:  I don't know about class action insurance.  There have been impacts, and I want to be careful about this because I'm not an expert on this, but, for example, you had the litigation in St. Clair County, Illinois, next door to Madison, with respect to insurance on replacement costs for vehicles.  Do you use replacement parts that are supplied by the original manufacturer or can you use parts that are manufactured by others, which is typically significantly cheaper?

Nationwide classification in St. Clair County against State Farm and a verdict in favor of the class, in that case, finding those sorts of insurance practices where you've used the nonoriginal manufacturer parts to be fraudulent.  Well, as a result, a lot of states that were encouraging the use of those nonoriginal manufacturer parts to have the rates down, lost the ability the encourage insurance companies to use that, not just State Farm, but all of them.  And my sense is, as a result, the rates have gone up.  So the end result of the litigation in that case, which is on appeal yet, I should add, to the Illinois Supreme Court, but it's affirmed, the end result is that, in State Farm's case, the insured end up with nothing in that multibillion verdict because they are the shareholders of State Farm, by definition, since it's a mutual insurance company.   The attorneys have indicated they will make a claim for $500 million and the insured generally will probably pay higher rates.  Whether that has come to pass yet or not fully across the board I couldn't tell you, but given that the option of going to those parts is now evaporated for most insurance companies, my suspicion is, in most states, rates have gone up.

MR. BROWN:  John, just to confirm what you are saying.  I do work for State Farm--Darren Brown from State Farm--and you are absolutely right.  Once that case was, although it is on appeal, we did suspend our use of non-OEM parts nationwide, including in jurisdictions that regulations practically require us to consider OEM, non-OEM parts in our repair estimates.  So it's having federalism aspects as well.  You get small courts, through the litigation, actually affecting regulations of a duly elected body in other states.

And, yes, you are right.  I have not seen anything recently.  But because we have suspended our use of non-OEM parts when we repair a car, it is adding to the bottom-line costs of doing business, and that does have, over time, the effect of increasing insurance rates, not the sole effect, of course, but obviously when we're paying more, when we could pay less for a fender, it is going to have an impact.

PARTICIPANT:  What was the basis for the liability in that case?

MR. BROWN:  Basis was it was allegedly a violation of the Illinois Consumer Fraud statute, that customers were not--and I'm a little sketchy on the bottom-line legal basis myself, but the inference was, by specifying a non-OEM report, that customers were not given what was promised to them, which is return to full value or complete indemnity.

PARTICIPANT:  So I assume State Farm can resume its old practice if it discloses what it's doing--that there would be no liability if they actually tell their customers what they're doing, and they can do it without risk of liability.

MR. BROWN:  Well, and I shouldn't go any further because that is a matter on appeal, and I am not an attorney, so I'm not speaking from a legal basis.

PARTICIPANT:  I only pretend to be one, in fact,

MR. BROWN:  I only play one on TV.

[Laughter.]

MR. BROWN:  But those are issues that were brought up, and those are issues that are--

MR. BEISNER:  One of the problems is that there was disclosure of that, and the court, in that case, concluded that it was fraudulent, nonetheless.  So even though you had the attorneys general of numerous states coming into that court and saying, "Please do not find this fraudulent," because they decided for the whole country, they made a determination under Illinois law.  And so even though Massachusetts came in, Hawaii came in, and so on and said, "No, we like this practice.  We want this.  It keeps rates down," the courts said, heck with you, and found it to be fraudulent.  Now, it's on appeal.

PROFESSOR EISENBERG:  There would be no disclosure State Farm could make to support this practice or did the court find the disclosure perhaps a little remote or in small print or somewhere where an average consumer might not see it?  There's no conceivable disclosure that could justify this?

MR. BEISNER:  I think the conclusion is why would you take the risk?  We'll just charge the consumer.  Who cares?  States took a shot at reducing it.  Some court finds it's inadequate.  We'll just sock it to the consumer.  Who cares?  That'll take care of it.  St. Clair County doesn't like that.  So the nation is going to have to pay more for insurance.

Why would you take the risk?  How do you know if the next--you know, you made a disclosure the first time.

PROFESSOR EISENBERG:  There are disclosures and there are disclosures, right?  I've seen cell phone contracts which disclose in unreadable print that you will arbitrate in Northern Alaska.

MR. BEISNER:  This is the problem with the state courts being involved in the system.  These were disclosures approved by state insurance commissions that are supposed to be the primary regulators on this, and the courts override them.  So to heck with it.

PARTICIPANT:  I actually know what that disclosure looked like, and it was very prominent, and it specified that these will be aftermarket parts.  And the ruling from the court was that no matter what the disclosure says is that aftermarket parts are inherently inferior, regardless of whether they can prove that in any given case.  So there is no disclosure that you can make that remedies that defect.

PROFESSOR EISENBERG:  Surely, there's a term in the contract which the full disclosure was sort of put up against, right?  Like we will restore the car to full value.  And so you could say to your client,  "We will not restore the car to full value.  We'll give you these parts."   It's hard to believe that this can't be addressed, in which case it will be reversed on appeal.

PARTICIPANT:  No, but this is the problem when you've got a state court that is trying to make these nationwide determinations and overrule what insurance commissioners and the decisions of other states on this issue.

PROFESSOR EISENBERG:  Let's talk about state courts making nationwide determinations.  In fact, state courts, from the beginning, were the guardians of civil liberty in the United States.  Every state court of general jurisdiction in America has powers that strike down federal laws as unconstitutional, and those rulings will stay unless the U.S. Supreme Court, in the rare cases it can hear, does that.

So the notion that state courts would improperly render rulings of national significance is just contrary to the system we started in 1789.  They were the primary guardians of civil liberties.  They were the primary places in which federal--federal--constitutional rights were to be safeguarded.

So that the fact that their ruling is of national significance may be a policy problem.  I don't deny it.  But the fact that they can speak on matters of national law is one of the foundations of our system.

PARTICIPANT:  The problem is that these are not supposed to be matters of national law.  They are matters of nationwide importance.  The problem is that, as in this case, the determination was made that they would simply make the policy determination on this issue that had been made by the State of Illinois and apply them and preempt the determinations that had been made by 49 other states.  And that's what's happening with great frequency.  And that is not, we're not talking here about the constitutional rights.  We're talking about policy choices that have been made by individual states that are being overridden by the states of other jurisdictions by imposing their own policy [inaudible] on other states.

PROFESSOR EISENBERG:  And that may be an area where the Supreme Court has a proper role to play in reining it in.

PARTICIPANT:  It has.  It's repeatedly said that, but they ignore it.

PROFESSOR EISENBERG:  Well, I thought the case was pending, right?

PARTICIPANT:  The case is pending, but let me tell you what happens, and I think this is what you're missing in the analysis of Madison County.

What happens in these courts is you have no chance of winning in those jurisdictions.

PROFESSOR EISENBERG:  What's the win rate for plaintiffs?

PARTICIPANT:  What?

PROFESSOR EISENBERG:  What is the win rate for plaintiffs for trials in those jurisdictions.

PARTICIPANT:  Class actions don't go to trial.  You note that in your own article.  They get settled.

PROFESSOR EISENBERG:  Well, maybe one should litigate one of these frivolous things.

PARTICIPANT:  State Farm did.  How much did you lose?

PARTICIPANT:  [Off microphone.] [Inaudible.]

PARTICIPANT:  Thank you.

PROFESSOR EISENBERG:  I don't have any sense of the actual win rate of plaintiffs in these cases

PARTICIPANT:  Because where they are filed, what happens is you realize there is no chance of winning the case, so you settle.

PROFESSOR EISENBERG:  And how do you know there's no chance of winning--

PARTICIPANT:  Because you're not going--

PROFESSOR EISENBERG:  You had one case.

PARTICIPANT:  --to win in these jurisdictions.

PROFESSOR EISENBERG:  Well, is this delivered from God or is it delivered on the basis of one case or is it delivered on the basis of 100 cases?

PARTICIPANT:  [Off microphone.] [Inaudible.]

PARTICIPANT:  Yeah, you've never heard of the Magic Jurisdiction?

PROFESSOR EISENBERG:  Of course, I've heard of it, but I've never seen a study proving it.  Have you?

PARTICIPANT:  Yeah.

PROFESSOR EISENBERG:  I'd like to see the study.

PARTICIPANT:  That in these jurisdictions, the cases simply settle because they make the judgment that you cannot win there.

PROFESSOR EISENBERG:  And why do they make that judgment?

PARTICIPANT:  Because on the occasions when people go to trial, you get these huge verdicts.

PROFESSOR EISENBERG:  That's one.

PARTICIPANT:  What?

PROFESSOR EISENBERG:  That's one.

PARTICIPANT:  There's also the certifications.  If they're all certified, that would be a sign that--

PROFESSOR EISENBERG:  But they're clearly not all certified.

PARTICIPANT:  Well, there's a debate over that.

PARTICIPANT:  Well, now, wait a minute.  I mean, for example, I guess I resist this notion that there is no data out there.  There's been a lot written on the magnet jurisdictions, none of which is acknowledged in your article.

PROFESSOR EISENBERG:  Well, I said that I think the most serious study I saw of tort hell was the RAND study of Alabama, and it turned out it wasn't tort hell for punitive damages.  So I'd like to see the studies.  It would be good.  Send them to me.  I'd be delighted to see them.  I have no reason not to cite them.

PARTICIPANT:  Okay.  And I would strongly recommend looking at the reports that the congressional committees have developed because they've spent a lot of time looking at these issues, and they're more than anecdotal.  For example, in Alabama several years ago, there was a study of a single tort there--

PROFESSOR EISENBERG:  George Priest?

PARTICIPANT:  What?

PROFESSOR EISENBERG:  No, go ahead.

PARTICIPANT:  No, it's not the George Priest study.  --where they went in and looked at one judge's class certification pattern.  And this one judge had certified as many classes over a two-year period as had been certified by all 900 federal judges over a single period.

Now, you can say:  All right.  That's an anecdote, but that one judge is having a huge impact on the national legal system.

PROFESSOR EISENBERG:  And I guess one has to know which certifications were appropriate and which were inappropriate, right?  I mean, the fact that he certified a lot of them may be that the meritorious class actions flock to him or it may be that the frivolous ones flock to him, but the fact that he certifies them--

PARTICIPANT:  Don't you find anything unusual that--

PROFESSOR EISENBERG:  Yes, of course, it is.

PARTICIPANT:  --that there is a judge that was sitting in Greene County, Alabama--

PROFESSOR EISENBERG:  Of course, I do.

PARTICIPANT:  --that you'd have to--you know, you don't stumble over that.  There's not a big airport there for all of the lawyers to fly into.  You have to spend a lot of time finding that particular county, and the same way is true with Madison County, Illinois.

PROFESSOR EISENBERG:  But why aren't the hard-core data in these tort hell jurisdictions in the studies?

PARTICIPANT:  There are studies out there.  I'd be happy to send you--

PROFESSOR EISENBERG:  Send them to me, sure.

PARTICIPANT:  Because they are certainly not reflected here, and the congressional reports--

PROFESSOR EISENBERG:  You know, congressional reports can get in what gets in.

MR. KICK:  Let's move to the last question from Eric [inaudible].

PARTICIPANT:  I'll put my plug in, which is maybe part of the problem here, and this is the pointy-headed economist speaking, but that the research design here of looking at the cases is subject to terrible selection.  I mean, once we discover, let's suppose you're right about this particular judge, no one in their right mind goes before them, and they settle quickly, and we don't ever see the data, which maybe says that we need to find some other way of looking at the merits of these cases.  That is, what we need is some other secondary market, something along those lines, that we could look at.

PROFESSOR EISENBERG:  I think it's a really serious problem and one of the reasons why perhaps the studies [inaudible], with one important exception, I think.  We rarely have an independent evaluation of the merits of these cases, right?   The important exceptions I think are lots of medical malpractice studies, where you get doctors in after the fact to look at the facts, and people seem to put credence in those.  And the story they tell is of course vastly different than the tort hell story.  The doctors who come in after the fact and look at the study, you don't find studies saying plaintiffs are wildly winning cases they shouldn't have won.

If you look at Neil Vidmar's book on North Carolina, his study of medical practice cases--

PARTICIPANT:  [Off microphone.] [Inaudible.]

PROFESSOR EISENBERG:  --the New York study showed, I don't think the New York study actually followed many of them through litigation because the claiming rates of injured parties were so low.  What the New York study showed was massive incidence of medical negligence by hospitals that led to no vindication whatsoever.

PARTICIPANT:  It was actually iatrogenic injury, but that's different than negligence, right?

PROFESSOR EISENBERG:  No, no.  They had a class of cases in which they thought there was medical mistreatment, and most of those don't filed.

PARTICIPANT:  It's a discussion for later, but [inaudible] about massive.

PARTICIPANT:  Well, low rate of [inaudible].  Massive may not be right.

MR. KICK:  As parting comments, I'd like to pick up on something that Paul raised, this idea that now the Eisenberg study will be cited, and the upper bounds will now become the regular fee for attorneys.

I wonder, given all of the necessary biases that exist in the data, and self-selection biases, even if we had perfect data.  And furthermore this idea that we don't know what constitutes a good case ex post.  So biases in the interpretation.

I wonder--and I'm going to ask each of the panelists this, to give a short comment on this--I wonder do these sort of necessarily faulty studies, good studies in the sense it's the best we had, but necessarily faulty in that they can't answer the questions that people want answered, do they do more harm than good?

Does ATRA jump on the Hellen and Tamrock [ph] study to say, look, we need to ram tort reform through the federal government or do the trial lawyers jump on the Eisenberg study and say, "Hey, what tort hell?"  Are we bettered by these kinds of studies, given that they necessarily can be contentious, Paul?

MR.          :  Well, in the short run, we may not be, but in the long run I think we have to be better off if we know more about what's going on.  As an academic, I have to think that.  But I think even faulty studies in the long run are going to get information that will be useful.

PROFESSOR EISENBERG:  I would note that we seem to have, I don't think intentionally, touched a nerve, perhaps because the class action bill is pending, because the Times, because of that, thought it was worth putting my ugly face in the New York Times, front page of the Business Section, because that never would have happened without the class action bill pending, but that certainly wasn't why the study was written.  Geoff was writing it I think in the hope of maybe people need to call him in as an expert witness.

I don't really know.  I honestly hadn't thought of the point--at the risk of being modest, who would have thought that this study would become the floor for attorney awards?  That's more than you can imagine an academic's impact being.  I certainly would expect plaintiffs' lawyers to not ask for less than at least one standard deviation above the mean in our study.  I would also expect, and maybe this is an agency problem, whoever is resisting that award to not ask for more then, to not support more than one standard deviation below what the courts have approved in the mass of cases.   And it may be the net result of that is not the upper bound becomes the floor, it may be the middle, in an adversary system, maybe the middle becomes where you wind up.  When the judge hears one side saying one standard deviation above, and the other one is one standard deviation below, the judge says, "Well, I'll take the middle.  That's the safe point," assuming you get reasonable reputation of the case for the lower fee, and that may depend on whether a defendant views it as coming out of its pocket or not, and I'm not sure about that.

MR. BEISNER:  Well, I think that dealing with the state court situation is frustrating because I think your comments are correct.  The data that exists for the federal courts certainly are not there with respect to the state courts.  And I think that any of these studies need to be heavily qualified to make clear what they do and they do not fairly indicate.  That's my concern here.

And the reality is you just have to go after the state court research I think differently and creatively because the basic body of data that you want to analyze admittedly does not exist there.  I've heard this argument before--well, go off and do that.  And even though the fire alarm bells are ringing about what's going on in these courts, we should just ignore it until there is some study there.

I have spent a lot of time digging down, doing research, going through each class action filing in some of these state courts.  It is very time consuming.  It is very expensive, and you are not going to get a comprehensive look at what's going on in every state court.  But I think that there is a way to get a snapshot of what is going on in the state court system.  I think that the congressional committees have done that in their reports, realizing that you have to read some of what is there with qualifications.

Is additional research advised?  Sure.  But I don't think people ought to wait around until the building burns down before taking action on it because it would be years before that research will be completed, even if there is somebody there to fund it, and I don't see that entity being there at this point.

MR. KICK:  I'd like to thank all of the panelists for an interesting afternoon and thank you all for coming.

[Applause.]

[Whereupon, the proceedings were adjourned.]