Getting real on Covid statistics: Non-traditional data can and must be tracked

Covid statistics are primarily about the number of fresh infections, those undergoing treatment, the number of those who have recovered and the death toll. It is time to add data on things uppermost in the minds of hassled citizens.
Vaccine availability in each locality each day for each eligible group, for starters. How many variants in each region?

Of course, to publish that data, we need to carry out genetic sequencing at scale. Sero survey data, as millions of people are vaccinated, efficacy data on the vaccines must be updated. Fighting a pandemic is not just about medicines and hospitals. Data is a powerful weapon that must be marshalled.


There are lies, damned lies and Covid statistics. The official Covid death toll is 3.3 million globally. This is widely assumed to be an underestimate. Unless a person has tested positive, his death would not be put down to Covid, depressing the death count in low-testing regimes. Then there are collateral deaths.

Someone who would have rushed to a hospital on experiencing heaviness in the chest might now decide to wait it out, rather than catch Covid at the hospital, only for a cardiac arrest to cut short the wait. These are factors beyond the politician’s distaste for large casualty figures on his watch.

The Economist has built a statistical model that estimates the actual Covid toll, measured as excess deaths above the projection of the past trend, to be some 10.2 million. The model yields a figure of one million for India, more than three times the official count.

The model takes into account the positivity rate of tests — if every second person tested turns out positive for the virus, the chances are that a high proportion of the untested population is also positive — sero survey results, the effects of Covid-control measures, captured in mobility data, age (the lower the mean age, the lower the death rate) and such other factors.

India should capture phone data to supplement traditional data, as well: distance travelled by phone wielders, and data traffic shifts by volume and regions.