Pay for Delay

I have been following the debate in the pages of Antitrust between two groups of distinguished economists about “pay for delay”, which is when a branded drug maker pays a generic drug maker to delay entry into the market.   The question is how much pay for delay do you need before you can conclude that the payment is anticompetitive and should therefore be treated as a violation of antitrust or competition laws?

The two groups of economists agree to think about the problem this way.

The brand and generic will litigate the question at what date the brand’s patent will permit the generic to enter.  We don’t know what result the court will reach, but we assume that we know enough about how such cases play out to generate an expected date of generic entry under litigation.

Now, the first group (gated link) of economists makes something like the following argument.  If brand litigates it has to spend on litigation.  Suppose that instead of litigating it can settle its dispute with generic by agreeing to permit generic to enter on the expected date of entry under litigation.  This is a nice result because you get in effect the same result that you would get under litigation, but without having to pay the costs of litigation.

So how much will brand be willing to pay generic for this result?  Any amount up to the cost brand would incur going through litigation.    Let’s suppose that brand goes ahead and makes a payment that’s just below the cost it would incur by going through litigation.  The result is efficient because brand and generic have avoided litigation costs.  Consumers of drugs are no better or worse off than if litigation had taken place.  They still in effect get access to the generic at the same time that they would have under litigation.  But brand and generic have both avoided litigation costs, and being able to produce the same amount for less cost is an efficiency improvement.  It’s good for the economy.  Indeed, in this story you end up with a transfer from brand to generic: in paying the generic to settle, brand is in effect allowing generic to keep all of the cost savings associated with avoiding litigation.  So you’ve got more profits for generics, which you might think is a good thing if your goal is to create incentives for firms to compete in drug markets.

The argument of the first group of economists is that payments of this sort from brand to generic up to the amount of litigation cost avoided are efficient and not anticompetitive.  But a payment above litigation cost is anticompetitive.  The reason an excess payment is anticompetitive is that it no longer generates a gain for brand by avoiding litigation costs.   It’s a payment of more than litigation cost avoided, so brand won’t make it unless it gets some compensation in addition to litigation cost avoided in exchange for the payment.  And that additional compensation can only come in the form of increased monopoly profits associated with a delay in generic entry beyond the expected entry date under litigation.  Payments in excess of litigation cost are only profitable for brand if they bring a return in the form of increased monopoly profits.

What I don’t understand about this argument of the first group is why any payment at all, whether above or below litigation costs, does not count as anticompetitive.  What is bothering me is this.  Suppose that payments from brand to generic are banned.  Will this prevent brand from settling the case with generic?  The answer is no!  The reason is that the parties can bargain over two variables: the amount of any payment or the entry date.  If you prevent them from bargaining over the amount of any payment because the payment will be zero, they can still bargain over the entry date.

To avoid litigation costs, brand will be willing to accept an entry date earlier than the expected date under litigation.  How much earlier?  It depends on how much monopoly profit brand loses with each day of monopoly power that it gives up.   If brand loses $1 million a day once generic has entered, and litigation costs would be $50 million, then brand will be willing to allow generic to enter up to 50 days earlier than it effectively would be able to enter under litigation.  Let’s say brand settles for 49 days of advance entry.  Now brand will have lost $49 million relative to what it would have made under a litigated determination of the entry date, but it will have saved $50 million in attorney fees, for a $1 million profit.  And note that every day in advance that brand allows generic to enter means an extra day of profits for generic because generic does not make money until it enters the market; so advance entry is valuable consideration that brand can use to induce generic to settle.

So even if brand can’t pay generic to settle in dollars brand can pay generic to settle in days of early market entry.  What’s more, it’s more efficient for brand to pay generic this way than in dollars.  The reason is that monopoly is inefficient.  It results in prices that are so high that some consumers who can afford the cost of production of the drug and who therefore in an efficient world would be allowed to purchase it are instead priced out of the market.  So total welfare and efficiency increase the earlier generic can enter.   Brand should be indifferent between paying in entry date advancement or paying in dollars; either way it pays to avoid litigation cost.  Generic will prefer the dollar payment because it will not capture all the value foregone by brand in pushing up the entry date.  But that’s just the point.  Generic won’t capture all of that value because consumers will capture some of it too.  And that’s why it’s better to force the parties to bargain over entry dates as opposed to dollars.

Once you understand that entry advancement can be used as coin in settlement negotiations, the use of any amount of dollars in procuring settlement starts to look purely anticompetitive.  The payment of dollar settlements becomes a substitute for the paying of entry advance settlements, so brand ends up making payment in money instead of making payment in monopoly power reduction (which is what entry advancement amounts to).  Every dollar of payment by brand is pay for delay, which is to say, a dollar purchased in exchange for additional monopoly power.  Brand can either avoid litigation costs by giving up monopoly power dollars or by paying monopoly power dollars out to generic.  Why wouldn’t you always want to force brand to pay by giving up monopoly power dollars?

Suppose that brand settles for dollars with generic by agreeing to pay generic $49 million.  The parties agree that entry will be on the expected entry date under litigation.  Now suppose that instead the agreed entry date is advanced by three days.  Brand will no longer be willing to agree to the settlement because now in effect brand will lose three days’ worth of monopoly profits.  If that comes to $3 million, as we suggested above ($1 million/day), then brand’s total payout for settlement will rise to $52 million, which is more than its litigation cost savings of $50 million.   Brand’s willingness to pay $49 million in cash to settle the case is conditional on it being able to retain its monopoly up to the expected entry date under litigation.  The payment it makes is not just consideration for a settlement, it is consideration for a settlement that permits brand to retain its monopoly up to the expected entry date under litigation.  It is payment for monopoly power.  Without the ability to pay dollars, brand would be forced to bargain away its right to postpone entry until the expected entry date under litigation (at least it would if its bargaining power remained unchanged).  The dollars buy it the ability to avoid bargaining away its monopoly power.

This brings me to the position of the second group (gated link) of economists.    These economists argue that the first group is wrong to claim that payments from brand to generic in excess of brand’s litigation cost are always anticompetitive.  They give examples of two ways in which they think that such payments can be procompetitive.   First, they argue that brand might be risk averse, meaning that it is willing to pay more than just litigation cost in order to avoid the uncertainty of litigation and procure a settlement instead.  If brand is willing to pay more than litigation cost for settlement, reasons this group of economists, then a payment from brand to generic in excess of litigation cost is not necessarily a payment for prolonging monopoly power and therefore not anticompetitive.

The problem with this argument is that as we saw above brand can use entry advance instead of dollars to bargain for settlement.  If brand is willing to pay more than the equivalent of litigation cost in entry advance days because it is risk averse and places a greater value on settlement than merely attorneys fee avoided, then so much the better for the economy and consumers.  Banning dollar payments for settlement will force brand to bargain away even more of its days of monopoly power and even more of the inefficiency associated with monopoly will be avoided.  Letting brand instead pay more dollars to settle means that the economy will not be able to benefit from avoiding all the days of monopoly that brand otherwise would have been forced to bargain away to obtain the settlement.

So this argument is not only not a refutation of the argument of the first group of economists that payments in excess of litigation cost are anticompetitive, it is also not a refutation of the argument that any amount of payment from brand to generic at all is anticompetitive.

The second argument in favor of payments above litigation cost made by the second group of economists is that brand and generic might have different understandings of the expected entry date under litigation.  Assume that brand has it right but generic thinks it would do better in court than it actually would and that therefore it would be able to enter 130 days earlier than the actual accurate expected entry date under litigation.  We’ve said above that brand would only be willing to pay up to 50 days of advance entry in exchange for settlement.  Assume that generic makes $500,000 a day when it enters and would have litigation costs of $30 million.  Based on its bad information, generic won’t accept a settlement for entry advance of less than 70 days (if it accepts a settlement for 70 days, it avoids $30 million in litigation costs but gives up 60 days of business relative to its perceived litigation entry date, so it gives up $30 million in profits, allowing it to just break even — any additional postponement of entry will cause it to make a loss relative to its perceived entry date under litigation).  Since brand won’t offer more than 50 days and generic won’t accept fewer than 70, there will be no deal and litigation costs will not be avoided.

The second group of economists points out however that if brand is allowed to pay generic enough money, then a deal can be struck and litigation costs can be avoided after all.  When generic has bad information about the entry date under litigation, trading entry date advance won’t result in a settlement.  But trading dollars could work.  The reason settlement doesn’t work with entry advance is that generic thinks mistakenly that it is already going to receive through litigation the entry advance days being offered by brand. As a result, entry advance ceases to be a valuable coin for purposes of bargaining over settlement.  But dollars are still valuable.

For simplicity, let’s assume that neither brand nor generic is prepared to bargain over entry advance days.  If brand pays generic an amount that covers generic’s perceived losses for each day by which generic postpones entry beyond its perceived entry date under litigation, then brand will be happy to go along with the postponement.  But will brand be willing to pay?  The answer is that brand will of course be willing to pay an amount up to litigation cost.  But if that is insufficient to cover the cost to generic of the postponement, brand will still be willing to pay if brand is able to fund the extra payments by postponing entry beyond the effective date that brand thinks entry would take place under litigation.  For this to happen, the extra monopoly profits brand is able to get from postponing entry will have to be large enough to pay generic’s perceived losses associated with generic’s extra postponement of entry beyond generic’s perceived entry date under litigation.  This won’t always be possible, but sometimes it will.

The key point here for the second group of economists is that the amount of the payment that brand needs to make to settle the case might be greater than the cost of the litigation.  You might immediately think that such a payment would have to be inefficient precisely because it exceeds litigation cost, but that would be wrong.  The payment involves a division of the gains from increased monopoly.  The gains from monopoly to producers are redistributive — they involve a transfer of wealth from consumers.  So they do not directly measure the loss in total welfare of monopoly.  To see whether the settlement increases or decreases total welfare you have to look at the reduction in total welfare (producer plus consumer surplus) created by postponing entry beyond the effective entry date under litigation and compare it to the gain of settlement arising from total litigation costs avoided by the parties.  For some settlements, the hit to total welfare caused by more monopoly will exceed the benefit of litigation cost avoided, but for others it will not.

For the second group of economists the fact that such settlements are not always efficient is no problem; their goal in the debate is simply to establish that payments in excess of litigation cost are not always welfare reducing and therefore should not be banned per se.

I think this argument does succeed at showing that reverse payments in excess of litigation cost can be efficient relative to a scenario in which no settlement takes place.   But I don’t think it is a reason for courts to permit payments from brand to generic.  The reason is that this solution is nevertheless always going to be less efficient than the situation in which the parties successfully bargain over advance entry days instead of dollars.  If we can find a fix for the information asymmetry problem (assuming that as an empirical matter this is a big enough problem to warrant our concern in the first place) that is cheaper than permitting pay for delay in the sense that it results in less entry postponement, then we should go for that alternative fix.  So for example if we could find a way to eliminate information asymmetry with respect to the parties at low cost, then we could maintain a per se rule against payments from brand to generic.  Fewer data privacy protections for corporations or a rule requiring drug companies to exchange litigation expectation analyses might do the trick.