From the perspective of those who model, amnesia is a desirable human attribute. How many of us now remember a model given wide publicity towards end-March? We were told: (a) 300 to 500 million Indians would be infected by end-July; (b) 30 to 50 million cases would be severe; (c) India’s mortality rate would exceed 3%; (d) With a lockdown, infections would peak in early May. There is a difference between mortality rate and case mortality rate. Mortality rate is deaths divided by population.
Case mortality rate is deaths divided by number of infections. ICMR has testing guidelines, which are dynamic and evolve. Initially, testing was limited to those with symptoms, those with international travel history, those in contact with confirmed Covid-19 patients and so on. Now others have been added. But the point is there is no universal testing. ICMR has also told us asymptomatic patients are 69% or 80% (two different datasets). Cases are those detected. Therefore, a mortality rate will be far lower than a case mortality rate. A few states have case mortality rates higher than 3%, but not all-India, and certainly not a mortality rate more than 3%. We haven’t reached end-July. But on (c) and (d), this widely publicised model was wrong.
On April 24, findings of another model were presented. The end-March model was private, devised abroad. The April model was indigenous and government-blessed, in the sense that a responsible person within the government presented it and took ownership. We were told: (a) Peak addition of daily cases would be on May 3 and the number would be 1,500; (b) There would be no new cases after May 16. Plain wrong again. There are difficulties in modelling, given uncertainty. All models have assumptions, which may be nullified. The problem is when modellers present findings with certitude, ignoring possible weaknesses in assumptions. If the government has to plan for Covid-19, it needs to know when infections will peak, when they will taper off. Yet, epidemiologists have made the government none the wiser. There are ongoing experiments with bats, pangolins and chimpanzees.
Despite these, we are a long way off from a vaccine. That’s the only thing known with certainty. No one expects medical experts to have perfect information about a possible second phase or third phase. There is nothing wrong with saying, “I don’t know.” What I find odd is those who argued, using muddled models, about a peak in early May are now talking about a peak in July. Not a word about egg on the face.
The worry is because those models were used to argue for lockdown. If the peak is now going to be in July, are we arguing for lockdown extension till end-July? If models were wrong about peak in early May, how does one know they will be right about peak in July? The peak might well be in September. Will we have a lockdown till then? We are continuously told that lockdown has flattened the curve. A lockdown reduces transmission. That’s plain logic. But the moment an expression like “flattening the curve” is used, there is a hypothetical curve, compared to which, the flattening is supposed to happen. No one knows what that hypothetical curve is. To use the right expression, that’s a counterfactual.
These aren’t like laboratory experiments. Sure, one can construct a hypothetical curve, based on assumptions. But as I have explained, assumptions have been proved wrong and been based on inappropriate data imported and implanted from abroad. Nevertheless, everyone seems to blindly assume some flattening has happened. For all one knows, the lockdown may have simply shifted the curve to the right, delayed the inevitable and postponed the peak. The reason I mentioned the mortality figures is to illustrate that this isn’t something to be inordinately scared about. Most people who will get infected will hardly notice it. Yes, some will be serious. But that’s no argument for epidemiologists to hold the country to ransom and stifle the economy, especially when they have been wrong in the past and may continue to be wrong in the future. Indeed, we are now being told we will have to live with the virus for a long time.
Think of it this way. Since everyone has been affected by the virus, everyone knows the concept of R0, the rate of transmission. When R0 drops to around one, we can confidently assert the pandemic is under control. All lockdown type restrictions can cease, since there is no longer any significant transmission. Don’t get fixated on the number 1, it might also be 1.5. The precise number is not the point. By the same logic, before arguing for lockdown, we should have had some sense of what R0 was, 3 or perhaps more, to justify the lockdown. But we simply didn’t know. We didn’t know since we had no idea about the spread of disease in the entire population. It’s impossible to know about an entire population, but a decent sample would have sufficed. There has been no such sample and there are no such data. As I mentioned before, testing isn’t universal. There has been no universal testing for a sample either, quite apart from the fact that testing isn’t infallible. It can give false negatives (obvious), but it can also give false positives (less obvious).
Add to that wide variation in tests across states. Delhi has done more than 7,000 tests per million population, West Bengal less than 800. Perhaps tests should be proportional to confirmed cases. With roughly the same number of cases as West Bengal, Andhra has done around 4,290 tests per million population. The case mortality rate varies between 0% in Chhattisgarh and 9% in West Bengal. Questions can be raised about sanctity of such widely varying data. If data are fine, we have no explanations for variations, unless it is the obvious one of problems with health infrastructure and immediate attention paid to serious cases.
Data has the appellation of being the new oil. As I have outlined, we have no idea about how the virus is behaving in India. We have no Indian data to populate the models. Therefore, data from other countries is used to oil models. Hence, models lack both explanatory and predictive power. They aren’t credible. Basing major policy decisions, with livelihood impacts, on such dubious models is not a good idea. Much better to be less sanguine and say, “I don’t know.” Uncertainty is about that.
Bibek Debroy
Chairman, Economic Advisory Council to the PM. Views are personal
(Tweets @bibekdebroy)