Mahalanobis is relevant now

The Covid-19 pandemic continues to ravage the world. An important part of our arsenal in the fight against Covid is accurate data and statistics. In this context, our efforts in the battle, concerning data and surveys to estimate the damages caused by Covid, would have been much stronger had Professor Prasanta Chandra Mahalanobis been alive today and in the lead.

P C Mahalanobis was born on June 29, 1893— yesterday marked his 128th birth anniversary. A self-taught statistician, Mahalanobis went on to establish the Indian Statistical Institute (ISI) and the Indian statistical system. His initial training in physics might have helped him develop an uncanny knack for conducting perfect surveys— The students at ISI would expand his initials as ‘Professor of Counting and Measurement’. Mahalanobis devised an instrument called ‘Coordinatograph’ for crop survey. In the 1930s, he assembled an instrument called ‘Photographic Profiloscope’, which, by statistical analysis, was observed to “be used with confidence as an instrument of precision for comparative studies in anthropometry.” To measure the error in the estimation of survey data, he introduced built-in cross-checks such as Inter Penetrating Network of Subsamples (IPNS) in the survey design. This is now considered as the curtain-raiser for the extremely powerful and useful resampling procedures like ‘Bootstrap’ in statistics. American statistician William Deming had said: “The main feature of IPNS is simplicity in the calculation of the standard error of an estimate.” Mahalanobis, ostensibly inspired by Kautilya’s Arthashastra, had also engaged independent supervisory staff in field operations by the NSS to collect data.

How accurate were his surveys? Take for example the jute survey conducted in Bengal about eight decades ago. Mahalanobis conducted a sample survey and got an estimate of 7,540 bales (1 bale = 400 lbs) which was remarkably close to the Customs and trade figure (7,562 bales)— the government enumeration estimated 6,304 bales. While the enumeration cost the government Rs 82 lakh, Mahalanobis conducted the survey at Rs 8 lakh with one-fiftieth of the manpower and got a near-perfect estimate. He also conducted a large-scale sample survey of Bengal’s famine-ravaged villages during 1944-45 to assess the extent of the famine’s severity. It was found that the economic condition of nearly 40 lakh people had deteriorated during the disaster.

At this crucial time, when the country is reeling under a pandemic, we need trustworthy surveys to estimate the extent of the damage caused by the pandemic and the findings could be extremely helpful for appropriate economic and social planning. Ever since the pandemic broke out, there have been many predictions— most based on standard epidemiological models such as SIR or SEIR or their simple variants along with the available Covid-data— about the trajectories of the number of infected and deaths. Data scientists, statisticians, computer experts and epidemiologists had also predicted the dates of onset of different waves when they would reach their peak and had prescribed the number lockdown-days needed to brace these waves. But, they all failed— It’s a new disease, ever-evolving in nature and there are many unknown factors.

No one knows the dynamics of surveys better than Mahalanobis and I believe that had he been alive, he could have provided more accurate predictions, helpful serosurveys and devised an effective strategy for vaccination and the distribution of vaccines. Five decades after his death, there are many who still believe that none other than Mahalanobis could have been effective in leading surveys and framing data-based policy decisions for human welfare and national development.

(The writer is a professor of statistics at the Indian Statistical Institute, Kolkata)