American Enterprise Institute

June 29, 2007

[Edited transcript from audio tapes]

8:45 a.m.	Registration and Breakfast
9:00	Presenters:	H. E. Frech III, University of California at Santa Barbara
		David Hyman, University of Illinois College of Law
		Meredith Kilgore, University of Alabama at Birmingham
	Panelists:	Randall Bovbjerg, Urban Institute
		Jonathan Klick, Florida State University College of Law
	Moderator:	Ted Frank, AEI
12:15 p.m.	Adjournment

Proceedings:

Ted Frank:  We will get this started.  Greetings, everyone. I am Ted Frank, Director of the AEI Liability Project.  Together with Bob Helms, we have brought together this panel on medical malpractice insurance and I am pleased to see such a large crowd here.  I would figure that everybody would be online for an I-phone.  But with so many economists you have probably paid for somebody to wait online for you.  Obviously, over the last 15, 20 years there has been a controversy over the expanding cost of medical malpractice insurance and which has lead to calls for reform of the liability system to, ostensibly, reduce the cost of that liability insurance.  Interestingly, though, in recent years it has been claimed that tort reform has no effect on the cost of liability insurance.  This claim was actually made on the floor in the Senate and the medical malpractice debates last year.  And that leads to the empirical question: Is that claim correct? 

We have a couple of papers here on, precisely, that question and a third paper on the related question of what goes into the cost of medical malpractice insurance, one of a long series of interesting papers coming from the Texas Department of Insurance Dataset.  Without further ado, let us start with the first of these papers from Meredith Kilgore.  This may be the first AEI panel in history of two panelists with ponytails.  Dr. Kilgore is an associate professor at the University of Alabama at Birmingham School of Public Health and associate scholar at the university’s Lister Hills Center for Health Policy.  He has extensive experience in critical care and nursing, and in the conduct of health services research in evaluation of health care technology.  He teaches clinical decision-making and cost-effectiveness analysis at the School of Public Health.  His current research projects include evaluating the effects of tort law and medical malpractice premiums, employer health insurance premiums, and defense of medicine.  He has published articles in a number of leading health journals.  Dr. Kilgore --

Dr. Meredith Kilgore:  Thank you very much.  I would like to start off by recognizing my collaborators on this - Dr. Michael Morrisey, who sends his regrets for not being able to attend and Prof. Jack Nelson from Samford University Cumberland School of Law.  This project was funded by the Robert Wood Johnson [indiscernible] initiative and we are thankful to them for the opportunity to go forward with this. 

The objectives of this study were to estimate the effects of caps on damage awards primarily on malpractice insurance premiums and, secondarily, to estimate effects of other tort reforms; now I will use the term “tort reform” with no prejudice intended, just for convenience.  Some background:  At the time that we were engaged in this, the U.S. seemed to be in a third malpractice crisis.  In the mid-70s there was a crisis of availability.  We saw the rise in premiums but, mainly, withdrawal of firms from the market, and that led to the forming of joint underwriting authorities, physician mutual associations, and changes in the way policies were structured and some state tort reforms, notably, the MICRA Act in California.  In the mid-80s there was a crisis of affordability, which was mainly just spikes in premiums -- led to further state efforts, more caps on damages, tighter statute of limitations, and other attempts to curb the costs of litigation. 

We did a review of the literature, looking at tort reform and malpractice premiums, defensive medicine, and health insurance premiums.  I want to just go over that briefly.  There are some excellent studies; two classic cases - Zuckerman, Bovbjerg and Sloan in 74 to 86 damage caps reduced premiums by between 13.6 and 17 percent, depending on the physician’s specialty.  There were limits placed on discovery and pre-trial screening, and pre-trial screening was also introduced, and had effect on reducing premiums; other reforms did not seem to have any significant effects.  Higher interest rates reduced premiums and this analysis did use state fixed effects; I will get back to that.  There was a Kessler and McClellan study, three years after re-enactment of direct tort reforms, which included caps; they found that the growth of premiums was lower by 8.4 percent.  Direct reforms for them; caps; abolition of punitive damages; collateral source rule reform.  No effect of other reforms in what they call indirect reforms aimed at the process.  And they used a fixed effects model as well. 

On malpractice premiums, on aggregate premiums there are two studies by Viscusi and Born that basically found, again, a reduction in the growth of premiums related to caps on non-economic damages.  And Thorpe in 2004 -- actually he used time-fixed effects and state random effects and found that award caps reduced premiums by 17 percent.  There have been studies reported that showed no effects from premiums.  There was a Gius study in ’98, revisits some earlier work, includes random state level effects and found no effects of reforms.  And Weiss, et al., did a comparison of ‘91 and ‘92 premiums and found premiums increasing more in states with damage caps.

On defensive medicine, probably, the main study is Kessler and McClellan -- looked at 1984 to 94 Medicare data and found, specifically, for heart attack and heart disease patients, that direct malpractice reforms were associated with 5 to 9 percent reduction in expenditures and no effects on patient outcomes that you could see in administrative data; so, essentially, mortality.  Dubay et al., looked at 1992 infant health and prenatal care data, found that higher premiums were associated with increased incidence of prenatal care but no effects on infant health.  Grant McGuiness [phonetic], using Florida data, found that liability exposure lead to about one percentage point increase in risk adjusted C-section rates and Stoddard et al., looking at Pennsylvania physicians, found that nearly everyone reported practicing defensive medicine.

Health insurance premiums-- very limited work on this.  Thurston in 2001 paper, looking at physician-free data concluded that physicians could essentially pass through the cost of malpractice.  And Pauly et al., looking at single specialty grouped data concluded that higher malpractice premiums do not reduce physician net income.  We have not found much on the impact of total health care cost.  We have done some work in this area and we have a working paper that we are shopping around.  Essentially, we have been unable to find an effect of damage caps or other tort reforms on health insurance premiums -- employer sponsored health insurance premiums. 

Now, when we state tort reform information when we first started looking into this we found a great deal of inconsistency.  Just going online and looking for some recent malpractice law, we found glaring discrepancies between what was reported at one site and another -- generally different firms.  Dr. Morrisey had a talk with the CBO and they said they were about 80 percent confident that their data were correct.  They said, typically, every few years they would revisit that.  They would hire an R.A. and they would come back with completely different results from the last R.A. that had looked at it.  So we decided to revisit this. 

Jack Nelson and a team of law students from Samford basically went back and reviewed all the statute law and judicial decisions pertaining to tort reform from ‘75 forward.  And we went up to 2004 and we have updated it to 2005 and they researched this.  And then Dr. Nelson went back and back-checked them; Randy went back and back-checked us and so we are fairly confident.  We are not sure in every case that everyone would code the statutes and judicial decisions exactly the way we did but we are pretty clear that we have a handle on it. 

The issues involved here -- when was the law enacted?  When was it actually in effect and when were court decisions issued that affected the law?  So you could have a law passed and it would be in effect for some years and then a court case would challenge it and it would be declared unconstitutional.  Alabama has a cap on malpractice damages that is unconstitutional, and so if you just looked at the statute you would be in error in saying that was in force.  So once again we had law students; we gave them templates for what to look for.  And then Dr. Nelson, Dr. Morrisey, and myself reviewed executive summaries that were prepared and came up with coding for what was the statute of limitations.  Was there a statute proposed and etcetera?  And where there were discrepancies in coding then we would re-visit the text and reach agreement. 

So the dimensions of malpractice reform that we examined were, first, caps on non-economic damages, principally, pain and suffering; the magnitude of the cap when enacted.  And then we would set the cap in real US$2005.  So we are looking not just at whether there were caps in place but the value of the cap and changes in value over time, which gave us a little more power to see some effects.  We looked at awards -- provisions for awards to be offset by collateral sources.  And there are two ways that this is handled.  Sometimes they can just be considered; the information can be presented.  And in other cases, there is a mandate that awards are reduced by any collateral offsets.  Then there is the length of statutes of limitations; whether there is a statute of repose and the time from occurrence to the final deadline. 

We looked at limits on joint and several liability, requirements for periodic payments, attorney fee limits, restrictions on qualifications for experts, limits on res ipsa, requirements for pre-trial screening, and arbitration provisions.  These are our malpractice premium data span, 1991 to 2004, from the Medical Liability Monitor.  We looked at -- in ’91, we had 22 states with damage caps; 24 had them in ‘98 and 26 in 2004.  There is actually a bit more activity than that would indicate because some states’ laws were struck down; others added.  So there is more than a change of just four states over the time period.  Similarly, we have got various numbers of states with other types of malpractice reform.

For the type of analysis that we are conducting, this table is a bit more informative and this is the number of states with state tort law changes over the period that we are interested in.  We had eight states with changes, either adding or voiding damages caps over the time period; six changed collateral offset rules; nine changed joint several liability; three statutes of limitations; seven statutes of repose; six periodic payments. None of them changed attorney fee limits; five changed expert limits; none changed res ipsa; one pre-trial screening, and three arbitration.  The reason this is worth noting -- that when there are small numbers of states changing laws then the approach that we use is biased in favor of finding nothing. 

The actual values of some things here -- and this shows the worth of looking at damage caps in constant dollars rather than nominal dollars.  If you look in nominal dollars from 1991 to 2004, there was essentially no change in the average value of caps.  But if you look at it constant dollars, they went down from 606,000 to 438,000.  So that is the effect of controlling for inflation means, the actual value of the damage caps is getting tighter over time.   Statutes of repose went up slightly; statutes of limitations, basically unchanged; that is our data.  We used premium data from Medical Liability Monitor; for some of our calculations we used the Medicare Resource Base Relative Value Scale. 

Our original intent was to use the malpractice component of physician reimbursements to do the entire analysis, but it turns out there are a lot of things that affect physician reimbursements, the least of which is trends in malpractice premiums.  So that data did not turn out to be useful for that.  What we did use it for is to be able to do some projections. Besides changes in premium rates, what would we say about the likely total budget impact of a national cap on damages?

We used the consumer price index to convert everything to constant dollars.  We used financial markets for information on the Dow Jones, NASDAQ, and Treasury bill prices, and we used the AMA survey for distributions of physicians by specialty.  So once again, most caps are set in nominal dollars; a few states index caps.  California current damage -- has a current damage cap enacted in ’75 of US$250,000.  In 1975, that was equivalent to US$870,000 2005 dollars; so major changes over time.  This is US$2004; actually we have updated it since this presentation. 

This is just trends in real malpractice premiums for the highest star OB/GYN premiums, followed by general surgery and the lowest are internal medicine.  So you can see from 2001, particularly for high-risk specialties, premiums were starting to take off; otherwise, fairly flat over the time period.  And this is just a comparison of states; this is the five states with the lowest premiums compared with the five states with highest premiums.  So you can see there is huge state-to-state variation in malpractice premiums. 

The statistical analysis -- we are concerned about characteristics that may not be observable among states that would lead some states to enact tort reforms and others not to.  So there would be confounding factors that could not be controlled for in a typical generalized least-squares model, and so we chose to use a fixed-effects approach.  So in this case, malpractice premiums in a region of a state at a given point in time are estimated as a function of a vector of legal statutes in states at points in time.  Another factor, x-variables in this case, mainly, economic conditions –- variables -- unemployment, and market indices.  A time factor that just controls for each point in time allows premiums to vary over time, and not necessarily in a linear fashion.  A state-level fixed effect which says we are going to estimate average premiums for each state and a random error term.  The effect of this is to say we are going to look at changes over time within states.  So in states that enact or change laws or where there are changes in the magnitude of the damage cap through inflation or by legislation, then we will be able to observe those effects on premiums.

We used here fixed effects to control for secular trends.  All estimates use robust standard errors; these are just a way of adjusting so that you are not underestimating the variance when observations come -- are clustered within geographic regions.  In addition to this, we estimate a naïve regression, state fixed effects, and state year fixed effects.  Basically, this is an effort to show how you can go wrong if you fail to use what we consider a proper model specification and we estimate alternative damage cap measures.  One way is we have an indicator variable that says whether there is a cap or not in the state.  And then the absolute -- actually, the log of the value of the cap is put in place; and in other, we have step function.  Is there a cap lower than 2500?  Is there a cap between 250,000 and 500,000 -- 575 and greater than US$750,000? Or there is no cap at all?

So this gives us six equations for three different specialties, so it leads to a really confusing slide.  So, basically, I do not expect [cross-talking] all of them are very important.  That is why we could not trim this down but what we have done is sort of here we are just going to look at -- the dependent variable here is log of annual internal medicine premiums.  So the way that these things are interpreted is a change in one of these variables translates into a percentage change in premiums.  And here we are just looking at internal medicine.  If you look at the second column here where it has got minuses and pluses, that is the direction we would hypothesize.  We would expect that having a damage cap in place would lead to lower premiums and we would expect that as the magnitude of the cap increases -- in other words, as the cap is less binding then premiums would go up.  Similar [sounds like]  we consider collateral source rules -- changes in collateral source rules with lower premiums as would making them mandatory – mandatory length. The length of statutes of limitations and the length of statutes of repose should increase premiums as there is a longer time frame for people to bring suits, and so on.  So that is what these are. 

In the naïve regression model, a damage cap with lower premiums, but not statistically significant -- and the magnitude of the cap would also lower premiums.  So that has the opposite of the expected side. Now that flips -- oh no, here a damage cap would actually raise premiums.  So in other words, if you just ran your standard vanilla -- ordinary least-squares regression, you would get exactly the opposite of what you would expect.  And that is because there are differences among the states, essentially, and differences across time.  When you control for that – and here I would just take people over to the full effect -- fixed effects model and things line up nicely.  And this is that for internal medicine premiums, the presence of a damage cap lowers premiums by about 19 percent and a US$100,000 increase in the magnitude of the cap raises premiums by 3.9 percent.  So that means that if your damage cap is around US$500,000 then it essentially accomplishes nothing.  And we actually found that caps above that amount, instead of being ineffective, actually tended to be associated with higher premiums; so more premium growth with caps above US$500,000.

This compares effects on different specialties.  This is the full fixed effects model for each specialty, and this is the step function here as opposed to the raw number up there; a cap of less than US$350,000 was associated with a 25-percent decrease in internal medicine premiums, 27 percent for general surgery, and 28 for OB/GYN.  If you look at the bottom here, caps greater than US$750,000 were associated with across-the- board increases in the -- think of this as an increase in the rate of growth of premiums.  In order to look at economic conditions we had to relax the time fixed effects, so we put in a secular time trend.  And so this takes the year instead of putting an indicator for each year, but that allows us to plug-in these market indices values.  And here we found that the only thing that was significantly different was that if market returns on fairly conservative investments were higher, then premiums were lower. 

So conclusions:  You can have a reduction in premiums from a cap on non-economic damages; lower caps, more reductions.  We found reductions associated also with statutes of repose and we found that larger returns on conservative investment result in lower premiums.  So some argument also for a market cycle, and that is where we are.  Policy -- we did a simulation; if we put a national cap in place and there had been no damage caps in any states up to that point, that would be associated with an estimated savings of about US$16.9 billion per year.  However, a lot of states do have caps in place and extending the -- so enacting a national cap and extending that to all states and making it US$250,000 would produce an expected savings of about US$1.4 billion.  That is what I have.

Ted Frank:  Thank you.  Our next speaker is H.E. Frech III, who goes by Ted Frech.  He is a professor of Economics at the University of California, Santa Barbara and an adjunct professor at Sciences Po in Paris.  He has been a visiting professor at Harvard University and the University of Chicago and an economist at the U.S. Department of Health Education and Welfare.  He has published more than a hundred articles and books on health care and regulation.  Mr. Frech has consulted on the economics of health care for private and public organizations and has testified in U.S. federal and state courts, state legislatures, state and federal regulatory bodies, and Congress. 

H.E. Frech III:  Thanks for the introduction.  You notice with our names you have to really [audio glitch] enunciate [sounds like] -- Frech and Frank.  Okay.  The context for my paper is not -- I do not have slides; I do not like PowerPoint.  So instead -- I do not like Microsoft either; that is another issue.  So there is a handout.  So if you do not have it there must be a way to get it. 

So the first page is just the two articles that my talk is based on; it is more based on the first one, which is in -- I think it has been handed out or at least it is available.  A little bit at the end will be based on the second one.  The context for these papers and the work underlying them was a policy dispute, so it is really a policy-oriented analysis.  And the policy dispute was a couple of years ago in California there was a big move to raise the damage caps, to increase them from -- I think it was actually proposed in legislature -- was to double them; raise them from US$250,000 to US$500,000.  Of course, lots of states have these caps; California, I think was the first, but one of the first, anyway. 

As we see, lots of states have them.  They are focused on non-economic damages.  I want to return at the end to the -- a little bit of the economics of why they are focused on that, rather than damages as a whole or some other measure.  And, of course, they are part of the general tort reform movement in the U.S. that has sort of grown up in response to the enormous change in the U.S. tort environment starting in the ‘70s.  In medical malpractice it is probably the most advanced area in terms of tort reform and the most tort reforms. 

Caps are one of them.  There are also some other ones.  So let me just sort of describe - this is getting very basic here - what the types of damages could be and exactly what gets capped, then describe the California law.  First, for the type of damages, there are economic damages - medical expenses, lost earnings, lost property, things like that.  In principle, these are objective and can be measured by market outcomes, market prices or market quantities.  Then there is the non-economic damages that are like pain and suffering, distress, loss of consortium.  These are, in principle, subjective and they are not directly measurable by some market-observable quantity or price.  Then there are punitive damages, which are designed to punish and go beyond the other two. 

The California law, which is called MICRA, is also the name of a friend of mine’s consulting firm.  Anyway, MICRA stands for Medical Injury Compensation Reform Act; it is 1975.  It set a limit of US$250,000 on non-economic damages, and there were three other major reforms which I’m just going to mention, but they are not really followed very much, although there is a little bit controversy on how powerful the other ones are.  One is it allowed evidence of compensation from other sources, so it is not mandatory; it is the one that allows it to be presented [indiscernible] talking about.  It allowed periodic payments, rather than a lump sum; again, it did not require them.  Those two reforms were considered fairly soft.  It also limited contingent fees for lawyers and the limits are pretty stringent. So there is an argument that that had some affect. 

The second point I want to make it here is that the caps reduced the incentive to litigate the weakest claims; this is, basically, an application of theory.  Consumers are more likely to litigate if the expected pay-off of a litigation is positive, net of cost.  Net expectation is a probability of winning times what you get if you win minus cost.  So what these caps do is reduce the incentive to litigate and it is a stronger effect for what are called lower quality claims, which are also called weak claims, claims with a lower probability of winning.  And that is illustrated in Table 1 in the hand-out. 

What is done here is we compare the expected value of winning based on some assumptions about costs, and we have claims of the same size. And we call the weak claims - our claimants and the strong claimants - different only in the probability of ultimately winning; 20 percent versus 80 percent.  And we look at different values of the caps because that was the -- sort of the main immediate argument in California legislature was the values of the caps.  And what you can see is the expected pay-off goes negative for the weak claims but not for the strong ones. 

So you would expect more effect of caps in discouraging weak claims; you would also expect more effect on big claims.  We do not have a table for that but that is another effect.  Small claims were -- no one is even asking for pain and suffering greater than US$250,000; it would be completely unaffected. That is the easy way to say that the effect is only on bigger claims.

One of the policy arguments brought at this time was that having these caps reduces access to the courts.  And some have argued against the caps because of that, or argued for loosening the caps -- raising them.  This is pretty tough to assess in a kind of general way, but there are some indicators.  Let us look at cases actually filed per capita in California, which is Figure 1.  The crucial dates here for the reform in California are 1975 when it passed the legislature, and then 1985 when it was upheld by the California Supreme Court. So there was some doubt about it for the ten years in there.  This issue of legal challenges -- Meredith brought it up.  In several other states, the legal challenges have succeeded, notably Oregon.  We’ll look at some graphs of what happened in Oregon.

 Well, what you see here is a rapid growth in claims in the early period; then a slight decline, I would say, described between 1975 and 1985.  Nothing very dramatic; it’s not like they have sort of eliminated malpractice from the courts or anything like that.  Another way to look at this is to look at claim frequency for California compared to other states, and that is Table 2.  And this has a different dominator; the dominator is the number of physicians instead of the number of consumers.  Otherwise, it is essentially the same data, or same basic data.  You can see the frequency at the physician level is not affected much by the 1985 decision.  You can also see that it is consistently substantially higher in California even after many years of the tort reforms - 22.8 percent versus 16.2 percent.  So these reforms do not exactly eliminate the issue of malpractice from either physician point of view or consumer point of view.  This is about as close as you can come in data to what people loosely mean by access, so it does not seem to be a huge effect.       

The next point is that the caps lead to smaller per-claim payments. I want to note that more than half of the states have caps but some of them have caps that are described by legal scholars as weak; they have either high limits or they have exceptions to the caps.  California’s is considered a strong cap by legal scholars.  So we have data that shows payment per claim across states and that is on Table 3, and you can see that states with caps have lower payments even though they are mostly wealthy urban states which typically are higher.  So you can see the average of states with the caps, including California, is US$234,000.  The average without caps is US$295,000; this is 2004 data. 

The next thing I want to look at is the effect of California caps on recoveries.  And here we are going rely on a study done by RAND corporation by Pace and colleagues, published in 2004 -- at least, published as a RAND report in 2004.  They looked at a sample of malpractice claims and actually applied the legal rules claim by claim.  They had a sample of 257 plaintiff victories from ’95 through ’97 and they just analyzed them directly and numerically.  And they found the cap was imposed in 45 percent of cases and awards were reduced by 30.43 percent overall, which means if you just look at the change in the other direction removing the cap from California altogether would raise the awards by 43.75 percent.  Now this is a dramatic effect and it is conservative because it ignores the effect of caps in discouraging litigation, which we just talked about earlier.  On logical grounds, it would discourage litigation on weaker claims.

The next point is the effect of the California cap on premiums.  Now, there are different ways to think about this.  One is the way that Meredith and Mike Morissey, Jack Nelson, the third co-author, and that their predecessors have done econometrically and I liked Meredith’s paper a lot.  Another way is to just think about the economics of the insurance industry, given that we know how much the outlays are going to be reduced or we have a pretty good estimate at that from the RAND Corporation study.  If insurance loading percentage had stayed the same, the caps in California should reduce premiums by 43 percent.  We make a series of economic arguments about the malpractice insurance industry about whether that -- we would expect that loading percentage to stay about the same.  And, basically, the answer is yes we would.  The first thing to note is that states with caps have lower premiums, and this is Figure 2. 

So this is an informal way of doing what Meredith did, just comparing a few states; and again they are the relatively wealthy urbanized states.  You see California is substantially less and Texas and Florida are much higher. This is done for several specialties here - internal medicine, general surgery, and obstetrics.  It is the same ones that Meredith studied; they’re kind of classic ones to study.  California is the only state up here that has the caps.  One thing to note here is that the variation among states is really enormous; it is almost breathtaking, even within the category of having caps or not having caps. This strongly supports Meredith’s approach of using state-level fixed effects.  That makes a lot of difference.  There is obviously a lot going on as you go from one state’s legal structure and culture to another’s that we just do not have a way of picking up.  So a state fixed effect, to the extent that moves slowly, will pick that up.  It is really a very good approach. 

Another approach is to look at natural experiments state-by-state and we did that for Oregon, partly because it is next door to California.  And what Oregon did is they put in a cap in 1987 of US$500,000, twice as big as California’s cap.  Then 12 years later, the Oregon Supreme Court ruled it unconstitutional and removed it.  Figure 3 shows what happened to premiums as result of that.  And you can see, whether adjusted for inflation or not, it is a pretty small effect because there has not been all that much inflation since this period.  But you can see that when the cap was enacted, the premiums went down substantially.  And then they took off again when the cap was removed by the Oregon State Supreme Court.  I want to note here a little bit the grounds for removing these caps by the Supreme Courts.  I think Meredith said Alabama has done it, Oregon has done it; there are some other states that have, too.  Wisconsin has?  Okay.  Wow, big states. 

What they argued in Oregon anyway was that it violates the right of a jury trial; that is in the State Constitution.  What they said is that the right of a jury trial within a state constitution has to mean that you cannot limit what the jury can do if it is a common law cause of action, which a professional tort is.   Maybe David could comment on that, but it seems like really a stretch to me to make that ruling.  Anyway, going back to the natural experiment that Oregon did, it is just one state.  So there is a sense in which it is one observation.  But then there is more than one person in the state, so there is another sense which it is not.  I think natural experiments are a way to be pretty sure that you have identified cause and effect.  So it has strengths and weaknesses.  I think it is a good thing to look at. 

Next, I want to make some more economic arguments about whether the cap -- the savings to the insurance companies from the caps will get passed on.  That is going to depend at – there are two levels it could get passed on.  First, when it gets passed on to physicians, then when it gets passed on to consumers.  So it is really a question of how competitive these markets are, how much of it gets passed on.  The idea that some of it gets passed on has to be the case as a matter of economics -- I’ll talk about that in just a second -- but how much of it gets passed on is an issue of how competitive the markets are. 

Many observers - the General Accounting Office is one of them; or General Accountability, I guess they call it now - think it is competitive --the insurance market is competitive to a fault.  And they have written, as lots of people have, that they worry that the insurance market is so competitive, they will bid down the prices so far that they will run the risk of being insolvent.  This is pretty much a concern with insurance markets generally; it just is not specific to malpractice.  One thing to note is that entry is very easy in insurance, generally, and malpractice insurance, in particular. 

Another point is that many of these insurers are mutuals, so they are owned by the physicians.  So any excess profits they might earn as result of awards being reduced would get returned to physicians in the form of dividends.  So an argument to think that it is not going to get passed on would have to be based on the insurance company absorbing the entire benefit and higher administrative cost for the benefit of lower payments.  I think it is a pretty strange argument.  And people do not usually -- when they realize that is the argument, they do not usually make it [indiscernible]. 

Then we also look empirically at accounting profits of malpractice insurers in California and other states, and that is in Table 5.  This is the return on equity for California medical liability insurance companies compared to the one-year T-bond rate.  And you can see that their return on equity is worse than the one year T-bond rate over the entire period, 13 years.  And the T-bond is a risk-less asset.   The long-run competitive equilibrium, the return on equity for any firm, should be above the risk-less rate because they are taking some social risk.  California malpractice insurers have done worse than that, worse than competitive over 13 years.  To argue that they are not passing on the benefits they are getting from the legal reforms, you would have to argue they’re absorbing enormous amounts in excess of administrative cost, which is, I think, not credible. 

Now for the next stage, whether the physicians pass on the savings in premiums, that is a little more difficult, I think, on just conceptual and economic grounds.  Clearly, physician markets are less competitive than insurance markets, although they have been improving over time.   Also, on economic grounds you would expect different answers for a cap on one state versus a cap on the whole country.  At the state level, there is lots of physician migration.  Physicians are –- they are like quicksilver; they migrate all over the place. 

So you would expect migration to eliminate any excess profits in the state as a result of a state reform and to put pressure on prices.  Now there is a question with physicians of whether these premiums really should be treated as a marginal cost or a fixed cost.  If they are a marginal cost, they would get passed on, at least in part, even if they are monopolies -- physicians.  If they are a fixed cost, maybe not.  In the simplest example where you do not have migration, it would not be.  So a question there is how much do these premiums vary by physician volume?  The answer that we can tell by just reading the, basically, insurance business literature, is somewhat, but not perfectly.  So there is a fixed component and then a component that varies with volume for the typical plan. 

And then the degree of competition matters; under perfect competition, any change in marginal cost would get passed on; that is rational and profit-maximizing to do that.  If competition is imperfect, any change in marginal cost would only get passed on partly but it would get passed on in part.  For example, if you had a monopoly with linear demand curves, which is a simple standard benchmark, half of the reduction in marginal cost would get passed on by a pure monopoly.  So when these proponents of this change in California were arguing that physicians would just keep all the money - all the money; they literally were saying that - it is ridiculous on a priori grounds, on conceptual grounds. 

At the national level, you do not get much migration.  The physician licensure in the U.S. and the amount of immigration that is allowed is politically set, not very dependent on malpractice premiums or physician incomes or anything else.  So at the national level you would not get migration.  So the passing on would only be based on whether it changed marginal cost and how much.  So you definitely expect less passing on at the national level.  Now, the studies -- the older study that we had found at the time we did this work was by Danzon, Pauly, and Kington in 1990, using earlier data.   And they found that about 16 percent of premium savings were passed on by physicians.  Now, there are two more recent studies that Meredith notes. Basically, as I understand it say that they are 100 percent passed on; so the profits are not absorbed by physicians.  So I would say we do not really -- we do not have a terrific handle on that, passing it on the second part. 

Another point to make here is that there have been loose arguments in favor of caps-- that without caps physicians are all going to retire or something; we are not going to have any physicians.  And if you look at aggregate levels, it does not look like this.  It looks like that is pretty crazy, pretty extreme.  But there are some problems by specialty areas and particular geographic areas that you miss when you look at aggregates, even like state level aggregates; particularly, low-income areas in states with no caps and particularly in OB, there is a real problem. 

And to illustrate that - and this is in Table 7 - this has malpractice premiums by specialty for selected counties.  Los Angeles County, which is high for California, but California has caps so they really are effective.  It is much lower than the other ones here, which are states with no caps.  And I want to quickly focus on Dade County, Florida and, predictably, compared to LA, Florida has no caps; California has caps.  Look at OB/GYN premiums in Los Angeles; it is US$66,000.  In Dade County, it is US$344,000.  You look at what Medicaid allows for normal birth, vaginal birth, it is basically US$4800 in LA, US$4500 in Dade County, Florida.  What that means is every year you have to deliver 14 babies in LA just to pay for your malpractice premiums before you start paying for anything else; in Dade County, Florida, it is 61. 

So these kind of numbers create a real problem in attracting competent physicians particularly competent physicians who are willing to spend a decent amount of time particularly in this service which is a very cognitive service; it is mostly counseling and monitoring.  I would really be surprised if Medicaid’s mothers were getting very good service in Dade County; they are probably not great in LA either but I bet it is worse in Dade County. 

The last issue I want to talk about is from the other paper that you do not have, and this was an argument made by the proponents of raising the cap.  Actually, it is not quite the last argument; it is the last statistical one.  Was the California good experience a result of the caps and other regulation, or regulating premiums directly, which was done as a result of a referendum in California?  And the dates make it possible to confound this a little bit; in 1975 the cap was passed; in 1985, it was upheld by the Supreme Court; in 1989, a Comprehensive Rate Regulation and Insurance in California was passed by the voters. 

So at least what the cause of stability from the ‘80s on, you could possibly confound it.  And there is an organization that I had never heard of before this, called the Foundation for Taxpayer and Consumer Rights that argues that it was all regulation; it was not tort reform.  And this is a consumer advocate group, and I went on their website the other day and I think it is more of an advocate for regulators than consumers but, anyway, that is their argument.  And this was seriously made in the California Legislature, this argument. 

It looks to me, though, like it is mostly the caps.  For one thing, the regulation does not apply to most malpractice insurers in California because it exempts groups that are organized in certain ways, legally, which most of the California ones are.  Another reason it probably does not matter much for California, before we get to the numbers, is that most of the insurance in California is written by mutual firms where the rate hardly even matters because it all evened out by dividends at the end of year. 

And then third point, the statistical one, is that California medical malpractice rates have grown slower than other forms of regulated insurance in California.  And that is Figure 1 from a different paper, which is at the end.  You get the lowest one here is malpractice which substantially dipped and it did not re-cross what it was in 1988 until 2001; whereas the other kinds of insurance that were also regulated really and much more detailed by the California Department of Insurance went way up. 

Now I want to close with a little bit of conceptual analysis of why we cap non-economic damages.  Why do we pick on those kinds of damages?  Well, to think about this we have to step back a little bit from the current debate and say, “Why do we even have a tort system?”  It is basically two reasons:  One is to provide efficient insurance; and the second is, in this context, to deter malpractice or, generally, to deter negligence and carelessness. 

Well, an economic analysis of insurance involving death or personal injury, which is slightly technical - and I can give you references after if you like - points out that it is a bad idea in the sense that consumers would not choose to have lots of insurance for bad states of the world.  It is easiest to see for when they are dead; they would not want to have huge payments when they are dead; it does not make sense from an insurance point of view.  Pain and suffering suffers from this but not nearly 100 percent replacement.  That is the point.  So there is a real argument for not having so much insurance for this stuff in general, and especially for pain and suffering. 

Then there is also in this economic literature specifically about this in the conceptual level.  There is also what is called judicial economy.  Remember at the beginning I said that economic damages are tied to something objective - a market price, a market output; non-economic damages are not.  So they are much harder for courts to get a handle on, juries and judges to make sense of it.  The other point is that in medical malpractice it does not seem that medical malpractice deters very well; it does not seem that it matches very well with the actual malpractice, the payments and decisions. 

So in medical malpractice area in particular, the caps do not seem to be working -- the caps do reduce damages, payments, and do reduce premiums.  And the case for having high non-economic damages for deterrence seems much weaker than in other areas of tort.  So at least for this area, I think the caps on non-economic damages make good economic sense.  And I would point out in many countries in general, you cannot get non-economic damages or they are sharply limited.  The U.S. is wildly an outlier in this respect; the caps get us back a little bit towards the international norm for rich countries.  I’ll end on that.

Ted Frank:  Thank you.  Our next speaker, taking a different tack on these issues, is David Hyman, who is a professor of Medicine at the University of Illinois.  He is the Galowich-Huizenga Faculty Scholar and Director of the Epstein Program in Health, Law and Policy.  His research and writing focuses on the regulation, financing of health care.  While serving as special counsel to the Federal Trade Commission, Professor Hyman was the principal author and project leader for the first joint report ever issued by the Federal Trade Commission and Department of Justice, Improving Health Care: A Dose of Competition.  He has been a visiting professor at the University of Texas and George Washington University Schools of Law and a professor at the University of Maryland School of Law. Professor Hyman--

 David Hyman:  Thank you, Ted.  You can see the title up there.  Let me start by mentioning the paper has three co-authors: Bernie Black, Charlie Silver, and Bill Sage.  But I’m the designated punching bag for today, which seems to be my role here at AEI.  And as Ted has already said, this is completely different. 

I’m not going to talk about caps; I’m going talk about defense costs involved in medical malpractice and other personal injury litigation, focusing on the State of Texas.  So why study Texas -- and I will not be able to show you that if this does not work -- there it goes.  Well, it is a big state; it spends a lot on health care; it has long been perceived as a problem jurisdiction from the civil litigation perspective, and you can see I have put some quotes up there.  The Lone-Star State has been a playground for pillaging trial lawyers from Forbes; multiple counties are tort hell-holes according to the American Tort Reform Association.  The AMA identified it in its first cohort of crisis states for medical malpractice in 2002.  So if you are interested in looking where the fire is, Texas is, I think, going to be on the shortlist of those states. 

The other reason is Sutton’s Law; that is where the data is.  It is one of the really only two states that have a pretty comprehensive closed claims database.  It is publicly available and covers a fairly lengthy time period; the other is Florida.  Each of the two databases has comparative strengths and weaknesses.  I would be happy to talk about those during the Q & A if people want to.  Texas is basically -- data requirements are as if you payout more, if you are an insurer, and they have a very broad definition of what counts as an insurer, and you pay out more than US$10,000 and that is nominal, not real, you are required to file a report with the Texas Department of Insurance that includes, depending on the dollar amount, either a fair amount or a lot of information about the claim, including all payments from all sources and particularly from our perspective today, defense costs.  If you pay out more than US$25,000 you have to file a very detailed report.  We use the more-than-US$25,000 paid claims as our threshold because those are the claims that we can be quite certain involve medical malpractice.  And if you want to do comparisons over time, it is helpful to make sure you are using a consistent data set.  And we have done a number of papers on this, including Stability not Crisis, which we presented here I think two years ago. 

So why study defense cost?   Well, there are a couple of reasons.  One is it is not up here; it is consistent with our scorched-earth policy of doing everything we can with the Texas closed claims data set to understand what is going on in the civil litigation and claiming system.   And we actually have a caps paper as well but today we are talking about defense cost.  The other reasons for studying it are, more seriously, in our earlier paper, this was the only part of the system that showed really significant change.  The earlier paper looked at ’88 through 2002 and observed that pay-outs, claiming rates, and so on were more or less stable or even declining, depending upon what you controlled for.  But defense costs were going up and they were going up at a pretty good clip as you will see in a minute, and it seemed sensible to try and see what was going on and why it might be going on. 

Now the other reason for studying it is the defense costs for medical malpractice claim is the sort of thing that is attracted occasional complaints.  In the debate over tort reform it is reasonably common -- it is absolutely common to hear complaints about pay-outs and skyrocketing jury verdicts; it is reasonably common to hear complaints about defense costs.  But that is typically in the context of spending on claims where there is no ultimate pay-out.  So this is part of the waste of the system is we are spending lots of money defending claims that have no merit. 

So the sort of larger context is to view this as part of the transaction cost of running the tort system.  And the historical focus has been not -- when you even talk about that, not on the defense side but on the plaintiff’s side, the fairly significant contingency recoveries that plaintiffs’ lawyers insist on before taking these cases.  And that is part of the larger debate about the compensation efficiency of the tort system.  Let me just mention – and it is referenced repeatedly in our paper – Hersch and Viscusi also have a paper on defense cost as well, using the same data set.  They do not focus in nearly as much on medical malpractice as we do, but it is worth looking at that paper. 

So let me now spend a couple of minutes going over our key findings and then show you some of the back-up figures and data.  So as I said before, inflation-adjusted defense cost or real defense costs are going up at a pretty good clip - 4.6 percent per year over the time period 1988 through 2004.  That is pretty significant increase year by year by year.  After you control for some case-specific factors, it is actually going up by a higher amount - 5 percent per year.  So by the end of our sample period, total defense costs are about 18 percent of pay-outs.  And the annual defense costs are running about US$150 million; that is in 2007 dollars.  But all the other figures I’m going to show you will be in 1988 dollars so you just multiply by 1.7 to get the current dollars.  But US$150 million per year just on the defense cost side, just for paid claims, is real money.  And that has to be funded out of premiums and investment income, just like the rest of the cost of the tort system. 

Additional points -- the rate of increase is pretty much the same for both legal fees and out-of-pocket cost, so defense costs include not just lawyer’s time but experts and filing fees and photocopying of medical records and so on.  And we see the same about-4.6 percent per year increase in both legal fees and in other expenses incurred by the defense.  And as so you can see in the last bullet, we rule out the most obvious explanations for those defense costs going up at that steady clip.  And those are increases in hourly rate, increases in pay-outs, taking more procedural steps to resolve the case; that is the resolution stage case duration.  You might think longer -- the longer a case is open, the more it is going to cost because idle hands are the devil’s workshop in law firms and then do not make partner either.  And then finally the exposure that is the amount of the policy limits.  And we basically can rule out all of those things as significant factors in explaining the quite-consistent increase over time in defense costs for medical malpractice cases.  

A couple more findings on med mal –- insurers’ defense costs go up with a bunch of things that it makes sense that they would go up with.  There is sort of -- appears to be consistent with a straightforward model of rational behavior by insurers.  They spend more in cases where they pay out more, and they spend more where there are higher policy limits and as the risk of exposure is higher, if there are more defendants they spend more.  If the stage at which the case is resolved and the length of time the case is open is longer, they spend more on all those things too.  So they are not just throwing -- I’m not suggesting and the paper is not suggesting this is just open up the checkbook and writing checks in bigger and bigger dollar amounts; it all conforms to a pretty sensible model of insurer behavior. 

Interestingly, this is reserving practices which most people’s eyes glaze over when you talk about insurance to begin with, and then when you get to reserves everybody is snoring.  So I will try and hold down the amount of talk about reserves, but it turns out the initial reserves, both for the indemnity side of it and the defense cost side of it, are not particularly good predictors of the eventual cost and what that is telling you is this is really a hard thing to do.  It is very hard right at the outset of a lawsuit to figure out how much it is going to cost to resolve the lawsuit, let alone the relative share of that that is going to be paid over to the plaintiff and the share that will be paid over to the lawyers and fees and so on; but it is, in fact, striking how poor predictors they are.  But even when you sort of abstract away from individual cases and look across the broader reserving practices, you do see something interesting, which is that although defense costs are going up at a good clip, medical malpractice insurers do not increase their reserves to reflect that reality.  So over time, there is a decline in the ratio of their reserves to the defense costs that they are in fact incurring and paying for. 

All right, now let me switch ears and talk about the non-med mal cases.  Non-med mal cases are less expensive and they are increasing at a lesser clip; it is still significant - 2.2 percent per year - but it is nowhere near what we are seeing in med mal.  The same factors that predict the increases in defense costs in med mal predict them in non-med mal cases.  But interestingly, insurers in these other lines of insurance increase their expense reserves to reflect the much more modest increases in their defense costs.  And initial reserves are better -- certainly not great, but better predictors of pay-outs and ultimate defense costs in this line, in these other lines of business even though they are not in med mal.  So it is a mistake to view insurance as a sort of homogenous pool.  There appear to be differences going on in different types of lines of the market. 

All right, so let me show you a couple of figures, a little bit of color.  This shows defense costs per claim and defense cost as a percentage of pay-out per claim is the red line and the scale is on the left side.  And what you see is in 1988 dollars; we are talking about US$20,000 per claim.  And by the end of the time period, we are talking in to the mid-to-high US$40,000 per claim to resolve these cases.  And in terms of a percentage of the pay-outs, right-side scale it starts around 8 percent and up at the end of the time period, it is close to 18 percent.  So effectively over the period ’88 through 2004 a doubling in the defense cost both per claim and as a percentage of pay-out. 

And then next step is to say, “Okay, what is going on with the reserving practices?”  And I got three lines on this; let me take you through each of them.  The top line is the indemnity reserves as a percentage of actual indemnity pay-outs.  The middle line is expense reserves as a percentage of defense costs, so each of them relative to their – what you are reserving for.  And then the last, the bottom line, the green line is simply the ratio of expense reserves to indemnity reserves.  We are comparing reserves to reserves here.  So what do you see?  Well, you see pay-out; the top-line is reasonably stable.  There is decline toward the end of the time period but then it sort of perks back up at the end.  And it is pretty much there is no statistically significant trend; it is about 35 percent throughout. 

What about expense reserves?  Well, it starts up in about the same general portion of the distribution, about 35 percent average for ’88 through ‘92.  But then you see the lines sort of steadily -- the purple line now in the middle steadily drops off over the time period; by the end of the time period, it is 22 percent.  So this again is the ratio of expense reserves to actual defense costs.  What does that tell you it says reserving practices are not keeping up with the rise on defense cost; that is why it is dropping as a percentage of defense cost.  But then you look at the final line on the bottom; you see it is completely flat; the ratio of expense reserves to indemnity reserves is completely stable.

What does that tell you?  As the paper says there appears to be a rule -- or at least our explanation is there is a rule of thumb which is for every dollar of indemnity reserve, you set aside a certain amount of expense reserve.  Not on a per case basis but across cases.  And the rule of thumb appears to work reasonably well at the start.  But it starts to run into difficulty toward the end when your defense costs have doubled, your indemnity costs are stable, and your reserves for one are working just fine but your reserves for the other derived from a rule of thumb are not.

Now you can see the same point by normalizing defense costs and initial reserves.   So what we did is we took an average of the first three years and set that to 100 and then we plotted out across the entire time period what is going on to initial expense reserves and total defense costs.  And what you see --total defense costs are the black upper line; reserves are the purple lower line.  In medical malpractice, things work reasonably well over the top period ’88 through ’96, or ’95; but after that, it is like falling off a cliff.  The divergence between the two gets increasingly broad.  And if you look at exactly the same sort of formula of normalized expense costs and reserves for non-med mal, you see a rather different pattern.  There is some delay in the mid ‘90s but over the time period toward the end, reserving practices in non-med mal lines of insurance appear to track actual defense expenditures during that time period.  So once again, we are seeing some differences in what is going on in different lines of insurance.

Okay, now let me switch gears from reserves and just talk about dollars.  How much does it cost to resolve cases?  Well, and that ends up varying along things that make sense.  One of them is the type of insurance that you have.  Another is how far down the pipeline the case goes; the longer and the more procedural steps the case involves, the more it costs.  And so the stack bars are from left to right lines of insurance.  Medical malpractice is the one on the far right.  It is the most expensive type of case in regard to its defense costs.  And then stacked within each bar from bottom to top is how far down the procedural pipeline the case goes.  No suit filed is the little tiny sliver on the bottom; suit filed but resolved without trial is the next bar up.  Full or partial trial because we have a bunch of cases that are resolved during trial is the next, and then the cute little blue thing on top is you go all the way through the trial and the appeal is filed. 

So what is this telling you?  A couple of things:  First of all, different lines of insurance have different defense costs but more importantly the more steps you go through the more expensive it is.  This is not particularly surprising but the variation within and across lines of insurance is somewhat noteworthy.

All right a couple more points and then I’ll wrap up.  So far I focused on defense costs as if they are a unitary entity; they are obviously not.  There are legal fees and other -- but legal fees can come in different varieties; in particular people who work in this area may know that there are what are called outside counsel and inside or staff counsels.  Some insurers have experimented with hiring their own lawyers on a permanent basis rather than engaging in a series of transactions with outside counsel.  This is the sort of classic, cozy and make-or-buy decision with regard to legal representation. 

There are some interesting professional responsibility issues, interesting by professional responsibility standards issues; but we are not going talk about those.  What we are going to talk about just for a minute or two is to what extent are people in Texas using inside versus outside counsel, and how is that changing over time and across lines of insurance.

So let me first start with the medical malpractice.  What you see is the upper purple line is spending on outside counsel.  It does not bounce around very much.  It is a very high percentage of total defense costs; it is attributable to outside counsel.  The line in the middle is other expenses so that is experts and photocopying and things like that.  That is in the sort of low 20’s throughout the entire time period.  And the little line, the green line at the bottom, is spending on the inside counsel at its peak, it is just under 10 percent and at the very end of the time period it drops off to zero entirely. 

We were reliably informed that only one medical malpractice insurer in Texas uses inside counsel.  They did not do it prior to the 1988 time period.  They started doing it; they kept doing it through the early 2000 period and then they dumped it; that is why it goes to zero.  In our last two years there is no spending on inside counsel for medical malpractice cases.  So for what it is worth one of -- we do not have individually specific insurer or patient or defendant identifiers in the Texas data set so we cannot track this on an on-going basis; but it appears that insurer –- medical malpractice insurers in Texas do not think inside counsel is the solution to whatever problems they have with rising defense costs even though they are going up at 4.6 percent per year.

Now you see a somewhat different pattern if you look at non-med mal cases.  And first of all, you see more use of inside counsel.  I’m not showing you that data.  What this is showing you is the probability of using inside counsel as a function of the amount of money at stake because there is more use of inside counsel in non-medical malpractice cases.  You can ask this question, which is not a particularly meaningful question they asked on the med mal site because it is only one company that is doing it in med mal to begin with.  Then what you see -- this sort of line that starts up at 15 percent and drops off to the right is that bigger cases tend to go to outside counsel; smaller cases tend to be handled by inside counsel.  And there is a sort of fixed percentage of cases that are handled -- there are -- at least that have billings for both inside and outside counsel.  But once again, the thing that is omitted from here is we have got 15 percent and about 4 percent; that leaves a very big number that is attributable to the use of outside counsel.  It is the dominant way in which defense expenses are put on across all lines of personal injury in both medical malpractice and non-medical malpractice cases in the state of Texas during our time period.

Okay, so a couple of implications and then I’ll stop.  First important implication is although it does not attract much attention, this is the sort of red-headed stepchild that you do not want to talk about or pay attention to.  And then it is not the glamorous indemnity payout dramatic increases in jury verdicts that people tend to want to talk about.  But there is this sort of steady rumbling boil of increasing defense costs; 4 percent per year in real terms over fifteen, eighteen-year period starts to add up to real money.

 But there appears to be some state variation in this; we do not know for sure because we do not have good data from most other states.  Nonetheless, there are some states where people are not complaining about this at all and other states where they complain about it a little bit.  So thinking about state by state variation and the causes of the same are worth doing.

 Point two: so far, I focused on defense costs and pay claims because –- remember Sutton’s law; that is where we have the data.  There is a whole bunch of zero-payout claims and small dollar payout claims that are associated with defense costs as well.  It is -- what data there is - and there isn’t very good data on it – suggests that defense cost in those types of claims accounts for forty percent of total defense spending; and we go through that in the discussion of the paper to the extent that there is any data available on that; and there is some.

 You add that together with the defense costs that we have quantified and paid, you come you up with the conclusion that total defense costs across the entire portfolio of claims, both paid and unpaid, are thirty to thirty-three percent of observed payouts.  So, defense spending is, interestingly enough, comparable to the spending of plaintiffs on their legal representation as well.  You add those two together and then ask the per-case-efficiency question how much does it cost to move a dollar from the defense side to the pocket of the plaintiff, net of both defense spending, net of spending on plaintiff’s representation and net of insurance overheads, something we have not talked about so far but it is real.  And you get a startling figure.  Best-case scenario:  It costs a dollar to move a dollar; it emphasizes the striking inefficiency of the tort system in transferring resources.  That is the best-case scenario and the worst-case scenario is a buck and a quarter -- a buck twenty, to move a dollar.

The paper also makes the point that we do not know what the optimal level of spending is on defending these cases.  We may be spending more and getting more; we maybe spending more and getting the same or getting less.  We do see a declining measure –- a declining performance on one measure of efficiency; that is spending per case.  But we see a modest improvement in another measure of efficiency; that is, the amount of time it takes to resolve claims is actually dropping over time. The nexus between that and defense spending is a different question; but nonetheless, we do see that. 

And since one of the common complaints about medical malpractice is how long it takes to resolve these claims –- the sort of glimmer of good news is it is not taking as long as it has.  But if you view that as really good news -– well, it is not really good news; it is just a sort of modest improvement from a very low baseline.

 A couple of more implications and then I’ll end.  Inside versus outside counsel, this I have already talked about.  Basically, at least in Texas, nobody seems to think that is the solution to increases in defense spending.  We do see more of it in other lines of insurance.  Again, emphasizing that you have to look across lines of insurance and across states to have a better understanding of what is going on.

Then we make a more speculative point in the paper, which is the failure to reserve adequately and let me stop here and footnote the important point.  Adequacy of reserves is an actuarial issue; it is beyond the scope of this paper.  We do see changes in reserving practices across time and across lines of insurance that suggest that not as much attention is being paid to whether your rule of thumb is accurate as could be -- and that is consistent with the insurance cycle explanation of premium spikes.  It certainly does not establish it but it is consistent with an insurance cycle explanation as opposed to the other explanations that we often debate on.

 The last point I want to end with is a point that we made -instability in our crisis.  Now with two more years of data, we basically can make it again.  The defense cost data allows us to make it in a different way, which is that changes in premiums just do not tell you very much about what is going on in the litigation system.  You need to look inside the litigation system to figure that out.  The reason to do that is the reforms you craft are heavily influenced by the problems you diagnose and you cannot diagnose litigation system problems very efficiently or effectively by focusing simply on premium data.  Thank you very much.

Ted Frank:  Thank you. We move now to our commentors. Leading off will be Jonathan Klick who is an adjunct scholar here at AEI and the Jeffrey A. Stoops Professor of Law and a Courtesy Professor of Economics at Florida State University in Tallahassee.  His research focuses on statistical analyses of the effects of legal changes on individual behavior. He has published academic articles in The Journal of Law and Economics, The Journal of Economic Perspectives, and The Journal of Legal Studies, as well as in numerous medical journals and law reviews.  He is the co-author, along with AEI resident scholar Dr. Sally Satel, of The Health Disparities Myth: Diagnosing the Treatment Gap.  Professor Klick --

Jonathan Klick:  Thanks, Ted.  As Ted said, although I’m in a law school, I’m actually an applied econometrics guy.  So Ted said, “Come in and give a big technical econometrics talk,” and I said “Yes, I do not think people are so interested in that.”  So I’ll try to confine myself to general policy implications, the biggest one being what implications does this research have for policy?  The ultimate answer is not much.  Right?  What we really want to know is what is the effect of litigation on health.  That is what we care about.  If you look at who is up on this panel, maybe only Hyman cares about what is the effect on doctors’ salaries or something like that.  Or maybe some economists care about how do insurance markets function.  But as a policy issue, what we care about is what is the link between what is going on in the court system and ultimate health.

What are the stages in that link?  Well, you got the effect of litigation exposure on insurance markets; that is what we are talking about here.  A couple of other papers, some of my own included, then look at what is going on in the insurance markets, how does that determine where doctors choose to practice and what decisions they have to make.

And then the last link is what effect do those things have on health?  Ideally, in a reduced form, we want to look at just the beginning stage and the end stage –- liability and health.  And oddly enough if you look at the literature, that is the smallest share of the literature.  There are relatively few papers that even worry about this but particularly that do a good job at identifying what is actually happening here.

Tom Stratmann and I have a paper in Journal of Legal Studies looking at the effects of tort reforms on infant mortality, since lots of people have said what is really important is what is going on with OBs.  If that is the case, you should see something in infant mortality and we got some fairly interesting results, but they are fairly noisy and so you cannot draw strong implications from that.

Dan Kessler and Mark McClellan have a very famous paper from a number of years ago in the Quarterly Journal of Economics where they looked at what is the effect of liability exposure on defensive medicine and then what is the effect on ultimate health outcomes.  And they have a striking result; they say “There is a lot of defensive medicine and we do not see any improvement or decline in health as an effect of tort reform changes; we just see higher expenditures when you have more tort exposure.”  It is an interesting result and gets sort of trumpeted around by lots of policy folks.  Problem is they do some fairly questionable things on how they code some things.  And also, other folks who have tried to replicate or extend the results basically cannot find that result.  So famous result but again we do not have a strong hold on what they actually showed.  You go even further back and some of Randy’s colleagues at the Urban Institute looked at the effect of med mal exposure on how frequently and how early people get prenatal care.

Again, some interesting results suggesting that if you have lower liability exposure in an area, people tend to get prenatal care more frequently and earlier in the process.  But results, by and large, are not significant except for some particularly marginal communities which maybe make sense.  Maybe med mal exposure has an effect on access but really only for those communities where -- that are on the margin; if price goes up a little bit they are not going to get any care.  Again, if you look through the literature, there is not much more than that and, really, that is the ultimate question that we are interested in.  I think in these kinds of conferences that we have had over the last couple of years on this topic that is a call that needs to be heard among researchers.

Stepping back and talking about the papers that were actually presented here today, I got some general methodological challenges in this area.  First of all, it is important to note that insurance markets, at least in a number of states, are highly regulated markets. And so what does that mean?  Well, it means in some states, I need to submit my rate changes to a regulatory agency and they need to approve them before they go into effect.  In other states, I can submit my rates and if the regulatory agency does not say “There is a problem here,” well, then, I can put them into effect; and in other states, it is maybe a less –-a more liberal standard than that.

The problem that this introduces is at least in some states you are going to have a significant lag between changes in liability exposure and actual changes in prices.  It is not exactly clear how you can control for this sufficiently.  It is interesting that you might as well say, “Why cannot we just control for different states according to their regulatory structure that exists in those states?”  Well, for about the last two years, I have been trying to put together this data set.  So go through every state for the last twenty years and say, “What were your pricing regulations?”  Well, anybody who has done this kind of thing realizes it is hard to go through the books and figure out what is on the statutes. 

So after about a year, I my RAs had that all coded up, and then I said “All right, start calling people at this agency to figure out if what is on the books is actually what is going on.”  If you do that, it takes a while until you find the guy who has been there for twenty years and actually knows what is going on. And it turns out in at least half of these cases, what is on the books does not really match what they actually do.  So it is not at all clear how you handle this problem and yet it is a crucial problem in any econometric study on the effects of liability exposure on ultimate prices.

There is another problem that I do not think anyone has gotten their skills around in solving and the problem is this:  If I am an insurer and I’m sitting in Texas or Florida or someplace else, and there is some sort of what we might call a liability innovation in another state, am I going to just ignore that and say “Whew, boy, I’m glad that did not happen in my state?”  Maybe, maybe not, but I might be taking that information and using that in my pricing decisions.  Then you might say “Can’t we just solve that by including generalized year effects?”  No, you cannot.  Why is that?  The problem is you got some states that already have tort reforms on the books, some states that do not and what this basically is going to do is this is going to create a specification problem.  That is a problem that I do not think anyone has handled in any of these papers.

Now, specifically on these papers -– I’ll start with Hyman’s paper since he was the last one to talk, and this is actually a problem that goes across their repertoire, so to speak. Statistical identification is all about counter-factuals.  You need to have a comparison group to figure out what would have happened in the absence of a given change.  This is always a problem for single state studies.  It may well be the case that payouts are flat but that does not mean nothing has been happening.  Maybe if you were able to compare to another state, payouts would have been doing something else, and so the very fact that they were flat in Texas actually implies that something was changing relative to a baseline.

How do you solve that?  As David said “You work where the data are.”  This kind of data would exist –- the Florida data is not even on these margins as good as the Texas data.  It is not a problem that you can probably solve, although I would suggest that maybe you do start looking at –- combined data set between Florida and Texas since you do see policies changing at different times between these two states.  It is not ideal but one comparison group is better than no comparison groups.

Another thing that is important is in their data set they seemingly have lots and lots of observations.  Problem is if you know anything about the way Texas is structured, there is lots of concentration in terms of who has coverage from whom.  So at different times in Texas before St. Paul exited the market, there were as many as a third of the doctors in the state getting insurance from the same company.  I think the mutual now has a similarly high concentration.  What does that mean?  Well, for most of the stuff that David and his co-authors are looking at, what that means is you do not really have thousands of observations.  You got relatively few observations.  Why is that?  It is not as though the mutual on one day says, “Oh, we are going to do this.”  And on the next case says, “Oh, we are going to do something else.” 

Probably most of their baseline decisions have already been made.  So that is one data point that is not a thousand data points. If you control for that, which -- you do not have specific identifiers -- you cannot control for it.  The question then is if you could control for that, do all these results just go to mush?  It is possible.

The other thing that is related to that is maybe when the point where they look at is:  Is it attorney’s fees changing that is driving anything?  Maybe mean attorney fees are not the right place to look.  Maybe you actually want to look at attorneys’ fees at the 90th quartile or something like that.  It may well be the case that, yes, the average attorneys’ fees have not changed but the attorneys who are operating in this particular segment of the market are changing like crazy.  We do not know and that would be an important thing to look at.  Why is that related to the previous comment?  You do not need lots of changes.  It does not necessarily mean that everyone in this industry has switched which attorneys they are using or what they are paying attorneys.  You get one big insurer doing it that can drive everything.

And one last point that applies to this paper and all the papers is everybody keeps deflating just by CPI basically.  That is not what you want to be deflating for most of this stuff.  What you want to be deflating for is a health-specific CPI.  We know that health prices have changed quite a bit differently than regular prices and have some sort of waiting between the health CPI and the region-specific wage changes.  That is what is going to drive med mal judgments, some combination of what are the medical costs and what are the wages that were missing.  So all this work on trying to deflate and get things right is sort of missing the boat on that score.

How about the Ted Frech paper?  Just a general kind of claim that puzzled me for a while –- yes, I’m on board; I believe that insurance markets are pretty competitive.  I do not believe that there can be a long-run equilibrium profits for these guys.  But then it does raise the question:  Why do these firms and why does this industry lobby for changes if at the end of the day, everything is going to be computed away anyway?  That is a puzzlement; it is a puzzlement to be sure.  I don’t know that it has a particular good answer but at least, it is something we should worry about.

And then lastly on the Kilqore paper –- something that arises in that paper that is striking in comparison to all of the other papers –- they get a collateral source effect.  Everybody else gets a collateral source effect, too; they get an effect that you pass a mandatory collateral source offset, insurance rates go up.  I find you pass a collateral source reform mandatory offsets and if the mortality goes up -- Paul Rubin and Joanna Shepherd find you pass collateral source reform, accidental deaths go up.  What is going on here?  Collateral source is not one of these reforms that people worry about and think about, but the fact that it is showing up more robustly than almost anything else suggests one of two things: One, theoretically there is something interesting going on here that we should probably try to get our heads around; or two, this is a specification check and the fact that something goofy is showing up in everybody’s regressions must mean we are all off the boat here.  For my own personal interests, I sort of hope it is the former but it could very well be the latter.

I guess, one last thing to note here –- you could -- and this is my last –- for Ted’s benefit, I’ll end on a technical note.  You say in the year fixed effects equations, you cannot control separately for market returns but actually you can.  And the way that you do this is you go through the insurance regulations; one of the things that states regulate is what can you put your money in?  And so if you find some states that say “Hey, look, it is bond funds or nothing,” or you find other states that are more liberal on that front, you could actually tease out both year effect independently from a market returns effects.  Again, the punch line to take away is we have not gotten anywhere close to the first order question: What is the effect of liability on health?   We do not know.

Randall Bovbjerg:  You are supposed to turn this on, right?

Ted Frank:  Right.

Randall Bovbjerg:  Well, this morning, I made a PowerPoint; I do not like Bill Gates, either, but I went with PowerPoint because I figured you needed to know how to pronounce my name.  I sit here as the fifth presenter and so I recognize that I must be representing Friday; specifically casual Friday.  I know one thing I am not representing is one of these the --

Male Voice:  Sometimes it takes a minute for you to realize that he wants to talk.

Randall Bovbjerg:  Well, I’m trying to –

Male Voice:  [Indiscernible].  The left button.

Randall Bovbjerg:  Yeah, I know.  I’m trying to -- is there a laser in here?  No laser.  What I’m trying to say is I represent Friday.  I certainly do not represent the Urban Institute.  And so I need to say that and we’ll go from there.  Press the left button -- there we go.  This is the wrong one.  This is the wrong one.

Male Voice:  I was going to say, “I saw this [cross-talking].

Randall Bovbjerg:  They loaded the wrong one or I loaded the wrong one.  Well, you know mistakes happen and that is why –- go ahead and turn it off.  That is why we have malpractice lawyers and, happily, I’m holding only a rhetorical scalpel this morning and not a real scalpel.  So my liability is probably limited by the natural state of things and I do not need a cap to protect me.

Let me say, I have four things that I want you to take away.  When you do a commentary, you have to get to something catty [sounds like], which I had in that earlier presentation.  And so I guess, there would be a lot of nitpicking about things but the big picture, I think, is the following:  Tort reform does work.  It works as intended.  That is to say, it shifts money from one side, the plaintiff side, to the defense side.  And therefore, because Ted is quite right that the markets are competitive in the long run or regulated - you can have it either way –- in the long run, that reduces premiums and there are lots of different studies that show premium effects of different sizes.  The Kilgore, et al. paper is the latest of many.  It is amazing that this should still be debated.  If you had held a meeting like this, attempting to get commentaries on medical malpractice 25 years ago there would have been one panelist.  Her name was Patricia Danzon and she basically created this genre of analyzing claims and thinking carefully, as Ted tried to do in much shorter scope, about what the theory and data show about medical malpractice.  Make sense of it.  Economic analysis, medical practice -- she summarized it in a book in 1985, just in time for the second big national crisis.

And here we are at the end of the third and we are still debating the same old crap.  And the positions on both sides have not changed.  If you look at the positions for and against tort reform, you can tell that tort reform makes a difference. You do not need a statistical analysis to know that these two sides care deeply.  And if tort reform does not affect –- premiums, does not affect the legal system, does not do anything, why is there such lobbying for and against it?  It really boggles the mind that we are still arguing about that. 

Second, tort itself works less well than intended.  Now it is not a compensation or, as Ted called it, an insurance system.  If it were, it would be a truly crappy system because it is so slow.  It covers so few cases, which people have not mentioned.  You could talk about efficiency, how much it cost to delivery the compensation that you deliver, but the main thing is that the rate of claims is usually low compared with every study that looks at errors and medical insurance.  How efficient is that?  Those are all the false negatives.

We focus on litigation because that is where the numbers are to a certain extent that there are numbers, but the bigger picture is different.  Actually, that is both points two and three, it is a very expensive system and that it works less well than intended.  And I guess the fourth would be that there are alternatives that are different from this 30 years’ war-debate that we have had.  We will probably be in a truce for a while now because premiums have stabilized but it will come back. 

The slides that I had were to show sort of Insurance 101.  The actuaries who develop insurance rates, of course, are famous for being the people who tell them how to drive the car and the insurance company by looking in the rearview mirror.  And what do they look at?  They look at the claims rate; how many claims?  They look at the severity, they call it -- frequency of the severity, the amount per claim; and you multiply those together and, by God, you get money.  And then you have expenses that pile on top of that, which is what was looked at by the people in Texas, Hyman, et al.  And then on top of that, you have other loadings for return on capital, risk, premium, and on top of that you have a market adjustment.  That is not the actuaries; that is actually people looking sideways at the other cars.  You know, can we sell this product at this price or can we maybe get a little more?  And so, those are the points that one examines.  So you get the Texas paper looking at the costs.  You get a little bit in the Frech paper looking at the claims rate and the severity of claims.  You get to look at the premiums, which are the result of this, and it is all imperfect for lots of different reasons.  And I will point some of them out as we go along.

Finally, when you get past this sort of actuarial view, you have – well, what is it that determines how fast those other cars are going?  And that is the insurance cycle and the degree of competition and supply and demand.  Capital moves into and out of this industry quite readily and we have a cycle.  It is quite recognized; to a certain extent it goes across lines.  It appears to be quite a lot worse for malpractice than other lines, which is a fruitful area that ought to be investigated more.  I think a lot of what people are after when they see tort reform is predictability.  And one of the things that people do not talk about California is the drastically shorter time from beginning to end of cases.  That makes it a lot easier to set your premiums and to drive the car by looking in the rearview mirror.

So then when you look at a cycle like that you say, “Okay, we have this period, even if it is 10 years.”  Like the Kilgore paper - all of the 1990’s; or you look at Oregon and you see things happen.  And as we just heard, you have to look also at what is happening elsewhere.  And so you look at Oregon and you see premiums going down or premiums, in general, were flatter going down in a lot of the country during the ‘90s.  They have been overpriced in the ‘80s and there was a lot of competition.  So, it really does take, as we have just heard, comparison and some institutional knowledge to do these things right.

Let me make a few comments about each paper.  Let’s start with the most detailed and specific one, the Texas paper.  The -- I guess it is the Hyman, et al. paper, despite his being second.  I was here a couple of years ago and I think it was Black presenting then but they decided to send the first string this time.  Someone who thinks fast on his feet and can deal with catty remarks from commentators.  One thing I learned from working with Mark Pauly years ago is that once you get into something, almost anything, it gets to be interesting. 

And that is what has happened here.  They got this dataset; they called it the Sutton problem.  I have always thought of this as the lamppost problem; the drunk who lost his keys and is looking under the lamppost because that is where the light is.  And it is very interesting to get into this, to wring every last drop out of this dishrag that are the datasets and that is what this is.  And it is really quite interesting to me.  If you look at the title of this paper, compare it with the other one, Stability, Not Crisis.  That is a substantive thing and it has policy implications.  Now, look at the title of this paper, Defense Costs in Medical Malpractice and Other Personal Injury Cases.  Aha, but there is a subtitle: Evidence from Texas.  And there are some bits of stories in here, but there is not yet a real killer story. 

I think one of the very interesting things as you look at this and you say, “Okay, these costs seem really high and they have changed somewhat in Texas.”  And I was trying to compare it with what I learned in the ‘80s and early ‘90s, talking with the insurance companies that these costs were something similar, both on the plaintiff’s and the defense side.  And it was something like a third.  And I got to table whatever it is for the figure where it was 18 percent even after rising the costs of lawyers.  And I thought, “My god, this is cheap in Texas.”  And then I got to the end and it was back up to 35 percent.  What is the quick answer on that?

Male Voice:  The quick answer is – these are – the figure is pain claims which is more than 25,000 so you have got all the defense costs in zero to 25,000 that are –-

Randall Bovbjerg:  So, that was not captured, okay.  So what I thought was going to be a story is not a story.  And there has been –- there was talk in the ‘80s and early ‘90s that insurance companies were getting much more on top of litigation costs.  And it was not so much having in-house counsel as opposed to outside counsel.  It was this brave new world of litigation management.  You call it management in order to get the sense that you can manage it.  That is kind of what I saw in that chart that showed a steady bar going across the bottom of – those are the in-house counsel looking over the shoulder of the out-house counsel. 

But I wonder why that chart is done in terms of the reserves, and particularly initial reserves.  Because insurers will tell you they got to put a number on it when the case comes in.  And that gets adjusted periodically, and over time the number gets closer to the final thing, except, of course, in the case of trials where nobody really knows what is going to happen; tremendously predictable system.  They are still negotiating over the verdict while the jury is out or they do not even tell the court that they have made a secret high-low agreement because they do not trust the court to come out with a reasonable arrangement.  So they are going to put bounds on it, but should the law put bounds on it?  No. 

Going beyond the Texas article, the Kilgore piece as I said is one of the latest to come out.  It is careful.  They really put a lot of work as the Texas people did.  You do not see the sweat that went in to cleaning the data and being able to count these things.  And in the Kilgore, when you do not see the sweat that went into the legal database or any of the other things, I think that I got more credit than I deserved when I just heard that I had checked the database while I was a reviewer and I spot-checked a couple of things that looked wrong and they fixed them.  But they did put an awful lot of effort into that and I think that, in general, the results fit the rest of the literature.  And they are right.  There is a significant or a substantial effect on premiums.

It is very hard to analyze premiums.  The lamppost happens to be located on top of medical liability monitor, which is a survey and it is a good thing that they do it.  It is a survey largely of the mutual PIAA style companies, not of all companies.  They do not get the same people reporting every time they do the survey.  You get different numbers of companies reporting; in about 17 or 18 states there are two rating areas.  So, in California, for example, there are sort of the big cities and the other southern and northern.  In Florida, it is Dade and Broward versus other -- in New York, it is New York City versus other and so forth.  But Iowa or Wisconsin or D.C. is all going to be one rating area. 

But if you got three different prices from three insurers and some of them are from one place and some are –- you have a real problem in deciding what the rate is in California or New York.  You have that even in Iowa or in D.C. where you might have different people in different years. There is just a lot of imprecision in the data.  Furthermore, what you are observing is the posted price.  Ted mentioned the dividends.  You do not know about dividends.  Worse than that, you do not know about underwriting.  Remember the other old story?  “This is an outrage here at Giant. You charge me 79 cents a pound for bananas and the sign down at Safeway says “39 cents a pound.”  The manager says, “Well, why do you not go up to Safeway?”  I say, “I cannot.  They have not got any bananas.” 

And that is what happens to some share of doctors during a tight, hard market in insurance.  So there is a posted price but I cannot get it because I make mistakes or I have something else that they do not like in my portfolio, my history.  You cannot control for all of this stuff in the lag -- the regulatory lag was just mentioned; there is some kind of pricing on it.  And the relationship to the cycle is very, very hard to figure out.  You try lagging this effect, you have the problem of whether the legal expectations are set only by in that state or generally; you have the problem of whether the laws expect to be overturned.  All that stuff makes it very hard.  I think there is one huge problem with this article.  It comes up in the abstract.  Happily, there is an abstract.  And that is the overall effect is said to be US$ 16.9 billion of tort reform.  Is that an annual effect?

Male Voice:  [Inaudible].

Randall Bovbjerg:  Would be the annual effect.  Well, the estimates and if you look at Ted Frech’s paper at Note 22, there is mention of this sort of a study.  And anyway, this output from The Lighthouse [phonetic], the benefits consulting and actuarial firm which estimates -- and if you go to the most recent year in 2005, estimates just shy of US$30 billion for everything -  physician, hospital, dentist, the whole nine yards.  So that the physician impact could US$29.4 billion just seems -- excuse me, could be US$16.9 billion just seems implausible on its face.  And given that you say that there is a savings of 20 percent, my simple mind wants to multiply 16.9 times five and say that is physician premiums and that is clearly wrong. 

It was also curious to me -- and maybe there is an econometric explanation, but if the effect per US$100,000 of moving the cap upward or downward -- and this is a very creative thing; not many studies have tried to do that. But it is hard to do.  If that effect is on the order of 3 percent for every US$100,000 how does it drop off to negative at US$500,000?  So, we know 250 works, so to speak, for California; change of 3 percent going up a 100 but you get past US$150,000 more and it does not work at all.  That seems curious to me.  So there is probably something screwy going on in there.

The last article is the most ambitious.  Ted was never one to shy away from a topic and it really contains a whole lot of different stuff.  I mean it has a whole range of everything that was in the Danzon book.  It