I’ve been working with Michael Kremer, Susan Athey, Chris Snyder and others to design incentives to speed vaccines and other health technologies. AcceleratingHT is our website and now features a detailed set of slides which explain the calculations behind our global plan. The global plan is similar in style to the US plan although on a larger scale. The key idea is that the global economy is losing $350 billion a month so speed pays. One way to speed a vaccine is to invest in capacity for 15-20 vaccine candidates before any candidates are approved, so that the moment a candidate is approved we can begin production (one can store doses in advance of approval). Most of the capacity will be wasted but that is a price worth paying. As Larry Summer says if you will die of starvation if you don’t get a pizza in two hours, order 5 pizzas. Human challenge trials are another way to speed the process.
A global plan is ideal since there are significant benefits to coordination. If each country invests in vaccines independently they will each choose the vaccine candidates most likely to succeed but that means all our eggs are a few baskets. There are over 100 vaccine candidates and they have different scientific and production risks so you want to choose the 15-20 which maximize the probability of success for the portfolio as a whole. To do that efficiently you need countries to agree that ‘I will invest in lots of capacity (more than I need) in candidate X if you invest in lots of capacity (more than you need) for candidate Y’, even knowing that the probability that X succeeds may be less than that of Y.
Vaccine nationalism is making a global plan look unlikely but if each country invests in multiple candidates around the world, as Operation Warp Speed is doing, and if each country guarantees to uphold contracts, we can reach a similar solution.
At AcceleratingHT you can also find our Incentive Design App which computes the optimal vaccine program given user chosen parameters. A big shout out to Juan Camilo Castillo, a newly graduated PhD student from Stanford, who put in a lot of heavy lifting on the app. We have been working on these models under time pressure and I will never forget the late night/early morning zoom calls where Michael Kremer would call out, “I think we need to take into account factor X. What effect would that have?” and Camilo would respond “Give me 5 minutes!” and, as we debated other factors Y and Z, Camilo would hack-away changing parameters and rewriting code till he had an answer. Hire a rising star while you can!