Misinterpretation of farmer suicide data

Shamika Ravi has published an article on farm suicides using NCRB data. She makes four important claims.
1. The number of farmer suicides has actually been falling in recent years; fewer such deaths were recorded in 2016 than at any time in the previous 16 years. Nearly twice as many Indian housewives commit suicide as farmers do. 
2. Suicide mortality rates are higher in the relatively wealthy states of Maharashtra and Andhra Pradesh than in poorer Bihar and Uttar Pradesh. 
3. Nearly twice as many Indian housewives commit suicide as farmers do. 
4. India’s farmers are doing far better than many realize. One way to gauge the well-being of rural households is to look at how much they’re buying: they now account for 45% of the fast-moving consumer goods sector in India. This is remarkable given the vast disparity in disposable incomes between urban and rural households, and it implies that improvements in rural infrastructure, connectivity and digitization are translating into higher demand. Over the last three years, rural sales grew significantly faster than urban sales in both volume and value; consumption growth currently stands at a robust 9.7%.
It has been published by Indian Express, Business Standard, The Print, Bloomberg, and Mint. This has been commended by many others using terms like "hard data", "evidence-based", "logic based" and so on.

In reality, this article is exactly opposite to what it is claimed to be. It does a data abuse of serious proportions and logically incoherent for the following reasons.

One of the first sanity checks done in any data analysis before reaching such far-reaching conclusions, of the kind that the article does, is to first understand the data. The article shows no hints of this, even remotely.

NCRB data methodology changed post-2014. Hence, the fall of suicides witnessed in 2016 is not comparable with the data till 2014. This has been pointed out numerous times earlier by many people and in particular P Sainath (herehere, and here). In brief, the issues are 

  1. Lack of training of the who collect the data, who are unaware of the change in the categories in NCRB data.
  2. States started reporting zero suicides from around 2011. For instance, as P Sainath points out, Chattisgarh and West Bengal started reporting zero suicides from 2011. But the three year average of the number of suicides for these states before 2011 was 1,567 and 951. This is not to deny that farmer suicides can decrease. But, it is only absurd to think that a yearly average of 1500 and 1000 came down suddenly to 0 just within a single year. So, in these states with 0 suicides, the number of farmer suicides is less than that of other categories like young, housewives etc. Surely, the farmers must be living in a paradise in comparison to other professions!
  3. Police put "tenant farmer" suicides in "agriculture labourers" field because many tenant farmers do not have pattas. In 2014, farmer suicides fall by 50% accompanied by an increase in agriculture labour suicides.
  4. Increase in "Others" category of suicides. For five states accounting for 70% of suicides, "others" column went up by several hundred percent, along with the decrease in the farmer suicides. It's because states started categorizing farmer suicides in Others.
  5. Women farmers are put under "housewives" category because of patriarchal reasons and also that they don't have their names on pattas. It explains the high number of housewive suicide as compared to "farmers".

For all these reasons, the claims of the decrease in farm suicides in 2016, poor states having zero suicides, and suicides being higher in other categories of people are absurd.

One can debate about the solutions to farmer suicides - loan waiver or input subsidies - but it is criminal to deny even the existence of the problem, that too with intellectual dishonest acts of data dressing.

It is expected of a serious researcher to do basic data sanity checks before making such far-reaching claims. Unfortunately, neither did the researcher nor the numerous publications that published the article and those who sang praises for it bothered not to. More importantly, this is not the first time that Shamika Ravi has resorted to such data dressing. In my memory, this is at least the third such instance, the previous instances being her analysis of Odd-Even policy and demonetisation, which I had pointed out earlier.

No wonder that Dr Arvind Subramaniam once expressed concern over the sycophancy of Indian economists.

No comments:

Post a Comment