There is risk and there is uncertainty. Since the days of Frank Knight, economists have differentiated between the two. Risk has a known probability distribution. For uncertainty, the probability distribution is unknown. COVID-19 makes us confront uncertainty, not risk. In either event, agents maximise expected payoffs. For risk, there is a given probability distribution that can be used by everyone. For uncertainty, there is a subjective probability distribution, which can, and does, vary from individual to individual.
How do I devise this subjective probability distribution? Through information and experience I already possess. There are various rationality assumptions used by economists. They are often violated. Otherwise, behavioural economics wouldn’t have taken off. Typically, given a situation, when your decision doesn’t agree with mine, I say you are irrational. However, with uncertainty, the problem may not be with rationality assumptions, but with differences in subjective probability distributions. Because of COVID-19, there is a certain risk of getting infected. Let’s call this the infection rate — total infections divided by the total population. Do I know what this infection rate is, for India, or for any other country for that matter? I don’t. I am not being pedantic. To the best of my knowledge, no country has done universal testing.
No country has done universal testing for a proper random sample either. The ICMR has told us more than 75 per cent of Indian patients will be asymptomatic. Who do we test? Those who show symptoms, those who have been in contact with confirmed patients and those who suffer from severe respiratory diseases. Most countries do something similar. In other words, when I work out an infection rate based on those tested, there is a sampling bias. This isn’t a proper infection rate. To the best of my understanding, the only country where we have had something like a random sample is Iceland. There, the infection rate was 0.8 per cent. There are similar caveats about the death rate. If I mechanically divide number of deaths by the number of confirmed cases for India, I will get a death rate just over 3 per cent. The global figure is a little less than 7 per cent. But neither of these is a death rate for the total population, since only those with severe symptoms are included in infection numbers. Three per cent or seven per cent are over-estimates. In a controlled environment like Diamond Princess, death rate as a ratio of total passengers, and not those infected, was less than 0.4 per cent. The true infection rate and true death rate are not alarming numbers.
What does this have to do with differential subjective probability distributions? There are slices in India’s population pyramid with rural/urban and other spatial differences too. Consider two extreme types. Type A, who are globalised in information access and morbidity. Life expectancy is 80 plus and there are lifestyle diseases like diabetes and high blood pressure. This co-morbidity increases possible death rates and thanks to globalised access to information, certainly increases perceptions about death rates, making them out to be higher than they are. Some of them have fixed incomes, regardless of what happens to lockdown. Therefore, if you think in terms of maximising expected payoffs with a subjective distribution, high probability is attached to loss of life and low probability to loss of livelihood. I have simplified, but you get the general idea. Contrast this with Type B, someone whose life expectancy is 60, without a fixed income stream and whose health concerns are tuberculosis and water-borne diseases, not COVID-19. Nor is access to information that globalised. High subjective probability will be attached to loss of livelihood and low probability to death from COVID. Both the types reflect subjective probabilities. Neither is “irrational”. There is tension between the two. Type A would like the lockdown to continue indefinitely, until long tail of the infection curve tapers off, perhaps beyond September. Type B would like lockdown to be eased soon, with necessary restrictions in hotspots. There is indeed tension between lives and livelihood. Even if health outcomes and information access are like Type A, but income is contingent on growth, preferences might mirror Type B.
Public policy needs to balance such differential individual preferences. This used to be the aggregation issue of the once fashionable, and somewhat esoteric, social/collective choice theory. Doing injustice to that entire literature and reducing it to column-type language, if preferences are heterogeneous, one set of individuals imposes its choice on the rest. Type A disproportionately influences policy. This determination of aggregate preferences is a dynamic process. Therefore, sooner or later, Type B contests this and as the lockdown is prolonged and livelihood costs mount, discontent surfaces, as it has across a range of countries. There were also welfare economics notions that pre-dated social choice theory, such as compensation principles of Kaldor, Hicks and Scitovsky. The point can be made using the two stereotypes. Specifically, Type A need to compensate Type B for their losses. To state it starkly, livelihood losses suffered by Type B need to be compensated by government through redistributive measures and this has to be financed by higher taxes imposed on Type A. The right question for the Type A is not whether they want the lockdown to continue, but whether they are willing to pay a COVID-tax to support lockdown extension.
This is meant to be a caricature, but it illustrates the public policy dilemma. Note that without a revival in growth, tax-paying capacity of Type B is limited and with job losses, some Type As become Type Bs. The choice is starker.
This article first appeared in the print edition on April 25 under the title “The public policy dilemma”. The writer is chairman, Economic Advisory Council to the PM. Views are personal.
No comments:
Post a Comment