Determining rewards in a prize fund
by Aidan Hollis
James Love has suggested, in his post yesterday, that rewards should not be simple linear multiples of QALYs. I appreciate his points, but think that there are some good reasons why simple linear multiples make sense.
1. Simpler rules are better. Jamie suggests -- and I am not sure how seriously -- a rule such as
Reward = a + b * ( QALYs ^ k ) ,where k < 1.
I don't like that kind of rule because I don't think that putting exponents into regulations is all that great an idea. People don't understand what they are getting. Every economist knows that many contracts and rules could be (theoretically) improved if only they were more sophisticated. But people continue to use simple contracts and rules in many situations, presumably because they value simplicity and clarity.
2. What I have suggested is that a drug's share of the reward fund should be equal to the share of the QALYs generated by that drug. Note that there are two things which increase QALYs: the number of people using the drug, and the average benefit per user. Under Jamie's suggested rule, the reward does not proportionally increase with the benefit. This would mean that a drug which offers half the benefit of another drug, but to the same number of people, would obtain more than half the reward. (Under a simple linear reward, it would get exactly half.) What rationale would there be for this? Jamie's rule could also lead to the following kind of behavior: create one new medicine with two equivalent forms. Market it as two medicines, and increase the reward.
3. I think, if there is to be any adjustment, it should be for drugs which treat diseases and conditions which are relatively rare. I think an appealing, and straightforward solution to the rare disease problem, is simply to create a supplementary reward which is only available to drugs treating rare diseases and conditions. Similarly, a fund could be created for drugs which treat the diseases of poverty, such as visceral leishmaniasis.
4. Adding flexibility is attractive, but also dangerous. The problem with flexibility is that it offers more parameters for games-playing by firms. The greater the discretion of the rewards authority, the more difficult the set of problems it faces, and the more rent-seeking effort there will be. If the only thing that is rewarded is creating measurable health benefits, then firms will focus on creating measurable health benefits. (And who can argue against that?) Adding flexibility can in principle allow the rewards authority to do better: but in practice it creates a whole new set of incentives for firms, many of which may not be desirable.
However, I would certainly agree that HR417 is right in not specifying a particular reward mechanism; and I think that the question of an appropriate reward mechanism is still very open.
Aidan Hollis
James Love has suggested, in his post yesterday, that rewards should not be simple linear multiples of QALYs. I appreciate his points, but think that there are some good reasons why simple linear multiples make sense.
1. Simpler rules are better. Jamie suggests -- and I am not sure how seriously -- a rule such as
Reward = a + b * ( QALYs ^ k ) ,where k < 1.
I don't like that kind of rule because I don't think that putting exponents into regulations is all that great an idea. People don't understand what they are getting. Every economist knows that many contracts and rules could be (theoretically) improved if only they were more sophisticated. But people continue to use simple contracts and rules in many situations, presumably because they value simplicity and clarity.
2. What I have suggested is that a drug's share of the reward fund should be equal to the share of the QALYs generated by that drug. Note that there are two things which increase QALYs: the number of people using the drug, and the average benefit per user. Under Jamie's suggested rule, the reward does not proportionally increase with the benefit. This would mean that a drug which offers half the benefit of another drug, but to the same number of people, would obtain more than half the reward. (Under a simple linear reward, it would get exactly half.) What rationale would there be for this? Jamie's rule could also lead to the following kind of behavior: create one new medicine with two equivalent forms. Market it as two medicines, and increase the reward.
3. I think, if there is to be any adjustment, it should be for drugs which treat diseases and conditions which are relatively rare. I think an appealing, and straightforward solution to the rare disease problem, is simply to create a supplementary reward which is only available to drugs treating rare diseases and conditions. Similarly, a fund could be created for drugs which treat the diseases of poverty, such as visceral leishmaniasis.
4. Adding flexibility is attractive, but also dangerous. The problem with flexibility is that it offers more parameters for games-playing by firms. The greater the discretion of the rewards authority, the more difficult the set of problems it faces, and the more rent-seeking effort there will be. If the only thing that is rewarded is creating measurable health benefits, then firms will focus on creating measurable health benefits. (And who can argue against that?) Adding flexibility can in principle allow the rewards authority to do better: but in practice it creates a whole new set of incentives for firms, many of which may not be desirable.
However, I would certainly agree that HR417 is right in not specifying a particular reward mechanism; and I think that the question of an appropriate reward mechanism is still very open.
Aidan Hollis
10 Comments:
I would like to correct this statement by Aidan, about my proposal:
----"This would mean that a drug which offers half the benefit of another drug, but to the same number of people, would obtain more than half the reward. (Under a simple linear reward, it would get exactly half.)"-----
I suggested that the reward for QALYs be non-linear, and decreasing. I did not propose decreasing in number of patients, but rather decreasing in QALYs. So, the problem Aidan suggests above would not be an issue.
Ok, Aidan's right... I need more coffee.
Looking again that the "two drug" issue raised by Aidan, in terms of decreasing rewards per QALY, if the reward is for incremental benefits, the second drug would have to show that it offered benefits not available from the first product.
If the company introduced the products at roughly the same time, Aidan's point is well taken. A blue and green verison of the same product would generate greater rewards than a single yellow verison (the only important difference being the colour), so long as the products are not compared against each other.
The government could also declare that the products were essentially the same, and treat them as a single innovation for purposes of the reward.
Whether this or another approach would be desired would depend upon whether or not there were signficant medical benefits from structuring rewards in such as way to not give too much of the budgeted rewards to a handful of products with huge client propulations, at the expense of products serving smaller populations.
If the reward is thought as the primary signal to stimulate soup to nuts investments in R&D, the simple QALY rule might make sense for much of the budget. But if the reward system is thought of something that helps private firms commerialize inventions largely investigated by public sector research, something that more closely coresponded to the risks and costs associated with funding clinical trials might make more sense.
I think we are in fact somewhere between these two points.
Jamie,
The point you make on rewarding firms which have used the results of public research in their innovation is very important. I don't know how one should deal with it.
Maybe one option would be for the research funding agency (not the prize authority) to ask for a share of the prize when their research was used?
I can certainly see the concern that firms might be unjustly enriched, and the rewards available for other firms made unjustly smaller, when a firm took an innovation funded publicly and then used it in its patented drug. But note that such a solution is not a problem of the rewards authority, and I think that is the way it should be.
Aidan,
In terms of government funded R&D, there are different issues. Governments (and other donors) can and do try to obtain intellectual property rights in donor funded inventions. In practice, this is only partly effective, because with the patent system, there a lots of ways to privately appropriate public knowledge goods -- and certainly this is the case in terms of medical research.
But I had in mind a different consideration. If private sector investors are playing the role of funding later stage development, but not the earlier and riskier R&D, the rewards they need to invest in clinical trials are related to the risks and costs of the trials, something a prize fund could take into account. If a firm needs X to invest in human use clincial trials with fairly well known risk parameters, and the reward is 1000*X, you are not necessarily targeting incentives most efficiently.
Jamie,
Musings on your last point...
Suppose that it is merely a matter of spending $50m on some clinical trials, which will either show effectiveness or not. The risk parameters, as you suggest, are well known. The product has a 50% chance of success. Then all that should be required is a reward of $100m or so. But suppose that the product, if developed, is expected to have tremendous therapeutic value, and so will receive billions of dollars of rewards under the reward fund. Then this appears to be an excessive reward.
I can see that there might be an excessive reward sometimes. However, there is research required to get the firm to the point of having a product with well-known risk parameters and values. If this work is done by the government, then the problem is for the government to capture some of the value. If the work has been done by someone else, then they will want to capture some of the value. The reward fund gives the incentives even for that kind of basic research.
This problem, exactly, is faced already in the pharmaceutical industry, where some basic research will produce an insight which can be used to create a commercial product. The reward fund may, like the patent monopoly system, overreward (even from an ex ante perspective) the first firm to take advantage of the basic insight. But this will make firms hungry to keep up with basic insights, or even to produce them themselves. The result may be a more effective R&D industry. So I am not sure that I really share your concern over this.
Something to think about some more.
Aidan,
I don't have a general objection to "excessive" rewards, per se. I think when you ask investors to make risks, part of the deal is that there is going to be an attractive upside for the investor. What society gets out of the deal is the decision making of the investors, who, when facing a loss of their own money, are expected to be more "focused" and disciplined. If investors are more efficient in picking the projects that are likely to succeed, it's worth paying for that.
That said, you don't have an unlimited amount of money. Does a potential reward of $10 billion rather than $5 billion really buy you that much more focus, if the investor is simply deciding whether or not to invest $100 million to clinical trials on a project?
Think of two possible distributions of prizes for successful projects:
Case 1
$10 billion + $5 billion + $1billion = $16 billion.
Case 2
$5 billion + $4 billion + $3 billion + $2.5 billion $1.5 billion = $16 billion
Case 1 would provide 3 prizes of $1 billion or greater. Case 2 would provide 5 prizes of $1.5 billion or greater. Which scenario would be best? Case 2, if you got everything you would have had in Case 1, plus other products. But maybe Case 1, if you needed $10 billion for some fabulous product you would not have gotten in Case 2.
How you look at this depends somewhat on what you think the "pull" funding is really doing.
I tend to think Case 2 (the flatter distribution of prize money) would be better.
Jamie,
I agree with your assessment of the two cases in the previous comment. The issue is really how you can (1) have a simple mechanism (2) that minimizes strategic games-playing and (3) maximizes the value of the drugs developed. I think that you are focused on (3). I am suspicious that (1) and (2) are very important in terms of achieving (3).
One solution that might satisfy both of us might be to set a maximum reward per drug. (For example, no drug is ever allowed to receive a prize of greater than 15 of the prize fund in a given year.)
The simulation that Barry Nalebuff suggested will be useful, I think, in looking at how different reward structures might work. Be great if IMS data is available.
At the end of my previous comment, that should have read "1%" instead of "15".
Post a Comment
Links to this post:
Create a Link
<< Home