Arbitrary use of data from multiple sources can lead to poor analysis
By R Ramakumar & Ashish Kamra
We write this note in response to the article titled ‘Can over-reliance on MSP harm agricultural states?’ by Shweta Saini and Siraj Hussain (FE, September 24; bit.ly/338h2Du). In that article, the authors compare Punjab and Bihar with respect to the farm Acts passed by Parliament. They compare the ‘income per household’ and the ‘income per hectare’ across these states, and argue that farmers in Bihar generate more income per hectare than farmers in Punjab. They list two reasons for this phenomenon: (1) the absence of APMC Act in Bihar; and (2) the absence of a strong MSP-based procurement system in Bihar. As a consequence, they argue, Bihar has had a more diversified cropping pattern than Punjab, with one-third of the value of output coming from fruits and vegetables.
We do not wish to join issue with the authors on their opinion on policy matters. But we do wish to point to many fundamental errors in their methodology and estimates. These estimates, in fact, form the foundation of their article and argument.
The finding that Punjab’s income per hectare is lower than Bihar’s, which is counterintuitive to begin with, should have alerted the authors to have a closer look at their numbers. According to the authors, who use data from the NABARD’s All India Rural Financial Inclusion Survey 2016-17 (NAFIS), incomes per hectare in Kerala, Bihar and Punjab were Rs 34,910, Rs 4,236 and Rs 3,448, respectively. However, the average monthly incomes per household in Kerala, Bihar and Punjab were Rs 6,284, Rs 1,652 and Rs 12,481, respectively. This, according to the authors, was because of the higher share of fruits and vegetables in the cropping pattern of Bihar, and of spices and plantation crops in the cropping pattern of Kerala. The farm Acts, they continue, would trigger such a diversification of cropping pattern in Punjab also.
We used data from two separate sources to cross-check these numbers. To begin with, we used purely macro-aggregates of value of output (VOO) in 2016-17 published from the Central Statistics Office (CSO). We considered VOO from all crops for all the three states, as well as VOO from fruits and vegetables in Bihar and Punjab, and from spices and plantation crops in Kerala. We then divided these figures for VOO at current prices with the net sown area (NSA) in each state to obtain figures for VOO per hectare. We find the following:
1. Indeed, the share of VOO from fruits and vegetables in VOO for all crops was higher in Bihar at 31% than in Punjab at 12%. In Kerala, the share of spices and plantation crops in the total VOO was 23%.
2. However, these did not translate to higher VOO per hectare in Bihar. VOO per hectare was Rs 1.84 lakh in Punjab, Rs 1.74 lakh in Kerala and Rs 1.27 lakh in Bihar.
In other words, despite a higher share of fruits and vegetables in the total VOO, Bihar did not record a correspondingly higher VOO per hectare than in Punjab. Kerala’s numbers were owing to the large-scale cultivation of spices and plantation crops, which cannot be replicated in other states due to its agro-ecological specificities.
We then looked at a very different data source: The 70th round of the Situation Assessment Survey (SAS) conducted by the National Sample Survey Office (NSSO) in 2012-13. We used unit-level data from this round to assess both income per hectare and income per agricultural household. In 2012-13, the average monthly income from cultivation per agricultural household was Rs 1,715 in Bihar, Rs 3,531 in Kerala and Rs 10,862 in Punjab. But the value of output per hectare was Rs 35,825 in Bihar, Rs 88,785 in Kerala and Rs 78,652 in Punjab—i.e. the value of output per hectare in Punjab was more than double of that in Bihar.
We would like to point out one more issue in the article by Saini and Hussain. The authors have used data on average monthly income from cultivation from the NAFIS. From there, they calculate income per hectare by dividing incomes by the average landholding size, i.e. 3.62 ha for Punjab, 0.39 ha for Bihar and 0.18 ha for Kerala. But these figures for average landholding size are not from the NAFIS. In the NAFIS, the average land possessed is 0.5 ha in Bihar, 1.1 ha in Kerala and 1.4 ha in Punjab. It appears to us that they have taken the average possessed landholding size from the Agricultural Census. If Saini and Hussain had used land possessed from the same source (the NAFIS) itself, they would have seen that income per hectare was Rs 3,304 in Bihar, Rs 5,713 in Kerala and Rs 8,915 in Punjab. In our view, the arbitrary use of data from multiple sources leaves their analysis poor. If they had used data on both income and land from the NAFIS, they would not have arrived at their erroneous conclusions.
As we mentioned earlier, we do not write this to quarrel over whether diversification of the cropping pattern in Punjab is desirable? However, we believe that such a conclusion is not borne out of the way the authors have argued it out. Bihar’s income per hectare is not higher than Punjab’s income per hectare. If we rephrase this following Saini and Hussain’s analysis, despite diversification of cropping pattern, Bihar’s income per hectare was lower than in Punjab.
(Ramakumar is NABARD Chair Professor, Tata Institute of Social Sciences, Mumbai. Kamra is an independent researcher and data analyst)