The 2019 Economics Nobel brought the discussion on randomised controlled trials (RCTs) into our drawing rooms. However, RCT is a century- old statistical technique and has extensively been used in millions of clinical trials till date. To evaluate the effectiveness of a treatment or intervention, in RCTs, a group of individuals are treated and compared with another group, called the control group. A ‘control’ is some existing treatment or no treatment at all. Moreover, each individual is allocated to one of the two competing treatments by some random mechanism like tossing a coin.
The theoretical formulation of RCT was given by Sir Ronald A Fisher, one of the greatest statisticians in history, during the 1920s and 30s, especially in the context of agricultural experiments. RCT expanded its domain into drug development from the 1960s, and has now become an integral part of clinical trials. Of course, there are many examples of RCTs in economic and social experiments since the 1960s, although it became popular in development economics since the mid-1990s, especially through the works ofthis year’s Nobel winning trio.
As a statistician, I was wondering whether Professor Prasanta Chandra Mahalanobis, the architect of India’s Second Five Year Plan, did anything similar. Mahalanobis conducted numerous agricultural and economic experiments for policymaking purposes. In fact, according to C R Rao, perhaps the most well-known statistician of the world today, Mahalanobis “was a physicist by training, a statistician by instinct and an economist by conviction”.
So it was quite natural for him to implement a purely statistical technique like RCT in his economic experiments. Mahalanobis started working on a wide variety of economic problems from the 1930s and continued to do so till his death in 1972. Among them are family-budget and indebtedness surveys for industrial and rural workers, studies on consumer preferences, studies on rainfall and floods in Damodar Valley and Hirakud Project areas, studies on development of roads and poverty as an after-effect of the great famine in Bengal in 1943. However, unlike the 2019 Economics Nobel laureates, Mahalanobis’s main focus was macro-economic problems.
Mahalanobis had a very close relationship with Ronald A Fisher, which was possibly initiated through a research paper of the former in 1925. They first met in 1926, during Mahalanobis’s visit to England. A close personal relationship between them was established and it continued till the death of Fisher in 1962. During this period, Fisher visited the Indian Statistical Institute—founded on this day in 1931 by Mahalanobis and nourished by him—eight times.
Fisher’s first visit was in the winter of 1937-38 and the last one took place in 1962 when he delivered the Institute’s first convocation speech. Fisher built RCT’s theoretical foundations in the context of agriculture. His book Design of Experiments (1935) gives an excellent account of that. So it was natural for Mahalanobis to use RCT as well in some of his surveys and experiments, since he was an ideal combination of a statistician and an economist.
During the 1930s, Mahalanobis designed some ICAR-sponsored experiments on agricultural field trials with S S Bose. Experts in survey sampling like T J Rao, former professor at the Indian Statistical Institute, believe they must have used the concepts of what we now call ‘RCT’! One of Mahalanobis’s famous economic experiments was the jute survey of Bengal in the early forties.
He conducted a sample survey at a cost of only `8 lakh; the complete enumeration (CE) by the government cost `82 lakh. The CE yielded a result of 6,304 bales (1 bale=400 lbs), whereas Mahalanobis’s survey resulted in 7,540 bales, which was very close to the customs and trade figure (7,562 bales)! Though this may not be called an RCT in the exact sense, this (or even the concept of independent subsamples) could be viewed as the precursor of RCT, as Prof T J Rao believes. So is the case with different sample cuts, like circular, of varying radii.
Then, in an Economic Survey in the early 1950s, Mahalanobis and S B Sen used techniques similar to RCT to find out whether a ‘week’ or ‘month’ would be a suitable reference period for collecting consumer expenditure data. An article on this was published in the Bulletin of the International Statistical Institute in 1954. Much later, an NSSO expert group under the chairmanship of Nikhilesh Bhattacharya conducted a pilot survey on this and their findings were published in the Economic & Political Weekly in January 2003.
This dealt with the most reliable reference period for collecting data on key items of household consumption, falling within the group ‘food, paan, tobacco and intoxicants’. It was concluded that for many items, shorter reference periods, particularly ‘last week’, may be preferable to the traditional ‘30-day’ reference periods. An indirect contribution of Mahalanobis in this context is through the ‘Mahalanobis distance’, his most well-known theoretical contribution.
Even the present-day development economists use ‘Mahalanobis distance’ in their numerous works concerning RCTs. Understandably, the volume of Mahalanobis’s works on RCTs is very limited. Is it again an implicit influence of Fisher, who was reluctant to apply statistics to social sciences, due to the ‘non-experimental’ nature of the field? But at least Mahalanobis did something similar to RCT in his economic experiments, nearly half a century before Michael Kremer, Abhijit Banerjee and Esther Duflo did so. He was the Plan Man of India indeed.