I recently wrote an essay about gun maker liability and the Protection of Lawful Commerce in Arms act. The lines of reasoning applied to the case of gun manufacture and sales are also applicable to the manufacture and sales of pharmaceuticals. In the gun case I brought up two critical points. First, it isn’t appropriate to hold gun manufacturers solely liable for the damages caused by guns when the trigger men are clearly the ones most directly responible for the damage. Second, the gun industry is extremely heavily regulated — to the point that the federal government is implicitly indemnifying gun manufacturers so long as they abide by the rules and regulations. In the end I decided to support the legislation; despite being far from ideal, it seems better than the alternative at this time.

There are some obvious differences between the gun maker liability issue and the issue of massive lawsuits against drug makers (as exemplified by the Merck Vioxx case). The gun maker liability issue centered largely on the question of whether gun makers should be held liable for damages caused by the illegal use of guns by others. Clearly there is nothing analogous to this when we look at the drug maker lawsuit issue. The lack of a potentially complex “chain of causality” (starting with manufactur and ending with a person being shot) could make the drug lawsuit issue easier, but then the complexity of determining causality at all is far more pronounced in the case of drugs than it is in the case of guns.

Most commentaries I have read on-line regarding the Vioxx case are focussed on the problem of determining causality, especially in light of the fact that juries are not really qualified to interpret complex medical evidence. From what I’ve read, the evidence in the Vioxx case was far from conclusive, yet the jury came back with an amazing $250 million dollar settlement. Whether jury competence is a major problem, I can’t say. The size of the award, however, strikes me as outrageous.

Of the approximately $250 million awarded, about $25 million was for actual damages — mental anguish for the wife and that type of thing. This alone is not defensible; the man was 59 years old, worked at Wal-Mart, and his arteries were found to be 70% clogged at the time of death. It’s a topic for another essay, but I have read elsewhere that the average person values his own life at somewhere between one and ten million dollars. Given this man’s advanced age, relatively poor health, and low income, he would undoubtedly come in on the low end of this range.

But the real kicker is the approximately $225 million in punitive damages. What is this for? Apparently the idea is to “send a message” to drug makers, and this it surely does. The message is “think twice before developing new drugs; if anyone is harmed by them, we’ll punish you — big time.”

This is where the economic reasoning applied to the gun liability issue can be applied again. If the system is working properly, a business, whatever it is, receives the correct signals, or incentives. Maybe this guy died because of Vioxx, maybe not. For the sake of argument, let’s assume he did. Does that mean that Vioxx is, on balance, a bad thing and that it should be removed from the market? To decide, we’d have to consider the numerous others who have benefited from the ability of Vioxx to fight potentially debilitating pain. If the benefits outweigh the costs, we want to keep the drug on the market.

It seems that there are at least two possible approaches: one is the free market approach, and the other is the top-down regulation approach. The market approach uses profit and loss signals to give businesses the correct incentives to do the right thing; benefits of products are captured in the form of sales receipts, damages are realized in the form of liability payouts. For this to work, however, the courts need to assess damages realistically.

Placing a dollar value on a human life is controversial, but necessary if any economic calculations are to be made. Once any emotional barriers are broken down, it should be obvious that human lives are not of infinite value as commonly held. Human behavior proves that humans do not place infinte value on their own lives. If we did, we would seldom go outside; the only activities worth doing would be those for which the risk of injury or death was more than counteracted by the expected increase in life span attained. Real life, however, is quite different; people risk their lives all the time in order to simply make their lives more enjoyable. They trade expected life span for both momentary pleasures and for activities that enhance overall quality of life.

If life were of infinite value, nobody would take pain killers.

There has actually been a considerable amount of study in the area of life valuation (try a Google search on “value of a statistical life”). If nothing else, the government sometimes wants a ballpark figure to use when deciding cost benefit trades for safety equipment and devices. For example, should the highway department use a fancy new form of high-traction concrete if it costs a $10,000 more per mile and will reduce fatal accidents by 0.1%? Economic analyses and techniques could also be used as a starting point when determining financial liability in drug cases, and the economic techniques could be used to make a case for a particular valuation in a specific case.

Note that the value of the specific life in question needs to be appropriately discounted if the person was seriously ill or close to death anyway at the time he was harmed by a faulty (or imperfect) drug. For example, suppose a drug is developed for a rare disease that kills its victims within 6 months of the time they are diagnosed. If the drug kills its user with a 50% probability, but saves him if it doesn’t, is it a good drug? All the users have to lose is 6 months worth of life anyway; for it to be a bad drug people would have to value 6 months of life more than half as much as they value the full remainder of their life expectancy, which seems unlikely. Therefore, it’s probably a good drug (assuming something better is not also available).

Under the market/liability system, of course, nobody is required to actually make this cost/benefit trade off. All we have to do is make drug makers liable for realistic (as close to actual as possible) damages and let them figure it out for themselves. Let’s say that the average person values the next 6 months of his life at 1/50th the value of the full remainder of his expected life span; he gets X enjoyment from the next 6 months, or 50X from the rest of his life. We could compute the percentage likelihood of death from taking the drug at which people, on average, would consider it a wash to take it or not to take it.

The expected enjoyment or utility associated with taking the drug, assuming for now that the drug is provided free of charge, is (1-P)*50X, where P is the probability of the drug killing the user (and thus 1-P is the probability of it saving his life). The utility associated with not taking the drug is X. By setting these equal and solving for P, we find the the break even point is P=1-1/50, or 98% — which is amazingly high. The assumption of a 50:1 ratio is not altogether unreasonable, but even if we lower it to 20:1 the value of P only drops to 95%.

In real life, unfortunately, the drugs can’t be provided for free. Suppose the drug company charges C (dollars) for a treatment with their drug. Whenever a person dies, the drug company is liable for the damages, X. Since a fraction P of the customers will die and be owed X, the average return from each sale is (1-P)*C - P*X. If the probabilities are not well known, the drug could still be marketed as a cure; customers will be willing to pay as much as 49X for a treatment (the benefit of being cured). If you plug 49X into the equation for the average return, you’ll see that, sure enough, P>98% makes marketing the drug unprofitable.

A practical problem here is that people generally aren’t going to have “49X” on hand — the value of something like 25 years of life. This is, of course, a pretty far fetched example though, where a drug will kill you 98% of the time. Even so, there are some options. One is to simply take out a massive loan. The “49X” price, recall, was the break even price where the person is indifferent between death and paying the bill. Hopefully, it won’t be that bad of a deal. Another option is to realize that the reason it’s so expensive is because the price includes (or in the simplified example consists fully of) the expected loss to the drug company of paying the estate if the patient dies. If the drug company could be indemnified, all he’d have to pay is the actual charge for the drug itself.

Thing is, there is a serious moral hazard problem if everyone gets in the habit of telling the drug company that it isn’t liable for damages. If that happens, the incentives to put out a beneficial drug are drastically reduced. A way around this is to allow consumers to sell their tort rights to a third party, probably an insurance company. The user can use the money from the sale to offset the cost of the drug, but the drug maker still maintains its liability (it simply pays someone else instead). Selling the tort rights for most real life situations (in which the probability of death is very low), will provide enough money to cover the bulk of the bill.

In this wild case, however, where the probability of death is very high, another scheme would be required (because the tort rights are only worth P*X and the bill may be as much as 49X). For this case, the user could make a deal with his insurance company like this: “If I live, you pay me the 49X I need to pay the bill. If I die, you get the liability payout.” So, to the insurance company the value of the deal is 0.98X-0.02*49X. That’s a break-even deal, so the user just needs to sweeten it a little so the insurance company can make a profit. The big difference between this approach and the simple sale of tort rights is that the insurance company only pays you if you live. With the simple sale of tort rights, the tort value ends up in your estate even if you die (which is most of the time in this example).

Once the drug’s record is established, and “P” is accurately (and widely) known, things get a little more complicated. A consumer who actually knows the risks might make a different choice. His net expected benefit from taking the drug is now known to be (1-P)*49X - PX - C. He’ll buy the drug if that expression comes out greater than zero. Setting this to zero, solving for C, and plugging into the average return expression for the drug company, we find that P must be less than 86% in order for buyer and seller to come to an agreement. [See comments at the end of the essay]

This 86% [96%] number has a huge assumption in it, and that is that the consumer places no value on the compensation paid to his estate if he dies. This seems like an extreme case. The other extreme is when the person does not apply any discount to money received after death (which is reasonably possible for someone with heirs), in which case he’ll pay 49X regardless of the value of P. Note that the drug company still isn’t interested in that deal unless P<98%.

Bear in mind that the 86% person would still rationally want to take the drug even at a 98% fatality rate, as long as he didn’t have to pay for it. The cost that he would have to pay, and the fact that he would notice the loss of that money if he lived is the problem. Since he doesn’t care about money his estate gets after he’s dead, he has no reason not to sell his tort rights for approximately P*X and apply that to the cost of the drug. Note that, to make things simpler, I assumed that C dollars for the next six months is as valuable as C dollars for the rest of the person’s life. This will probably not be true in general; C dollars with only six months to spend it is probably worth less. The more the near term money is discounted, the more the person is willing to pay, and the closer we get to 98% again.

I may have lost all my readers by this point, but I wanted to go through the math to show how a free market system with a liability rule can motivate a drug company to do the right thing — that is, so long as the damage awards are correct. As explained, it’s not necessary for the probabilities to be known for the drug company to get the proper feedback; though they would certainly want to have an idea of it, lest they lose their shirts in court later. In no case is the drug company motivated to put damaging drugs on the market, and if a beneficial drug is available, insurance-like arrangements can be used to allow consumers to opt-out of the benefits of liability without changing the incentives to the drug maker.

If the damage awards are excessively high, as in the Vioxx case, drugs that are released will be skewed heavily toward “excessively safe,” meaning that a large number of other beneficial drugs will be kept off the market in order to prevent the “one in a million” from dying of complications. People, especially juries, need to be less emotional and more rational when thinking about damages. I’m sure it’s easy to get caught up in the plight of a single patient like Mr. Ernst (the one who supposedly died from taking Vioxx), but what about the millions of others who can no longer benefit from the drug in question? And while there are certainly emotional barriers to placing dollar values on lost human life, there’s no escaping the fact that that’s what you are doing, as a jury member, when you determine the penalty in a drug case. The dollar value chosen goes right into the calculation that determines whether a possibly beneficial drug should be released.

Like I said before, if human lives are of infinite value, then there is no place in the world for pain killers. If juries insist on delivering damage awards consistent with an (almost) infinitely valued human life, the liability system will do it’s job and drive pain killers off the market.

Having said all this, we could question whether a liability system is even appropriate at this time. Even more so than in the case of gun distribution, the sale and distribution of pharmaceuticals is heavily regulated by the government. When the FDA gives it’s approval to a drug, it is more-or-less explicitly telling people that it considers the drug to be “safe.” If it is going to police the industry in this way, and attempt to give people a sense of security by approving some drugs and disallowing others, doesn’t it only make sense that the FDA should suffer the liability penalty when things go wrong? Where is the FDA while all this Vioxx nonsense is going on?

To summarize, a free market in pharmaceuticals, coupled with a liability rule that has the drug manufacturer pay real damages when things go wrong, will automatically achieve the desired effect of having only beneficial drugs on the market (or at least stay on the market). This will be true so long as damage awards are reasonable. The awards don’t need to be perfect; if they are close, then drugs that are clearly beneficial will be profitable — some at the margin will be on the market when they shouldn’t be, or off the market when they should be. We can tolerate a little inefficiency, but awards such as those in the Vioxx case are clearly off the chart in terms of reasonableness. They send a completely wrong message to drug manufacturers: don’t release a drug unless it has absolutely no side effects whatsoever.

The FDA is a wildcard in all this; it generally appears overly cautious — keeping beneficial drugs off the market — then reverses itself when public opinion turns against it. In the end, a properly functioning liability system leaves the FDA as nothing but a hindrance. If the FDA is going to maintain its role as determiner of drug safety, it should at least step up and take some (or all) of the blame when a drug it approved doesn’t work out as expected.

[Added 10/16/05]: There is an error in the math that arrived at the 86% value above. It’s not actually relevant to the main thesis of the essay, but I thought I should correct it. The value of a cure, to anyone, is 49X-C: 49X for the extension of life minus C, the cost of the treatment. If a person dies in treatment, and he places no value on money added to his estate after his death, he simply loses X, the value of the six months of lost life. The fact that his estate is compensated is of no relevance. All put together then, the value of the drug to this type of customer/patient is (1-P)*(49X-C)-PX. For it to be worth taking, C must be less than (49X-50PX)/(1-P). As a sanity check, set P=0 (drug works every time) and you’ll see that it’s worth 49X to him, as expected.

The drug manufacturer’s equation doesn’t change: for it to be a good deal to them, under the liability rule I prescribed, C must be greater than PX/(1-P). The maximum value of P where both the drug maker and customer are happy is about 96%. This doesn’t seem like much of a difference from the 98% value we had before (the highest P that is of value if the drug is free), but it’s enough difference to drop the price to 25X.