William Boyd in his 1998 novel, Armadillo, created an antonym for serendipity. He called it zemblanity. If serendipity implies pleasant discoveries by chance, zemblanity is “the faculty of making unhappy, unlucky and expected discoveries by design.” The novel revolves around these twin poles of serendipity and zemblanity, bringing out our daily life that oscillates between utopian dreams and a dystopian reality.
Last week, I was torn between serendipity and zemblanity. I was asked to review a book for Frontline magazine, Weapons of Math Destruction , by Cathy O’Neil. Dr. O’Neil started the Lede Program in Data Journalism at Columbia. Her earlier book, in collaboration with Rachel Schutt, Doing Data Science , remains one of the finest textbooks in big number-crunching. However, her latest book explains the inherent problems in big data. She establishes the ironic relationship between the high assumptions behind mathematical models and the inequality they breed. The assumptions are that mathematical models would ensure greater fairness, eliminate bias, and judge by universal rules. But in reality, the book explains how these models become toxic by reinforcing stereotypes, by being opaque and incontestable, even when they are wrong. Reading Dr. O’Neil’s book was a moment of serendipity to learn about the tyranny of numbers.
India’s growth story
The zemblanity moment happened when the Central Statistics Office (CSO) retained its January estimate for growth in gross domestic product (GDP) in 2016-17 at 7.1%. If these figures were right, it meant that independent economic forecasters had got their estimates about the potential slowdown due to demonetisation completely wrong. This newspaper’s Editorial, “Resilience reaffirmed” (March 2, 2017), reveals the dilemma in accepting these figures at face value. It read: “The Survey had also made a cautionary assertion that recorded GDP growth would ‘understate’ the overall impact of demonetisation as ‘the most affected parts of the economy — informal and cash based — are either not captured in the national income accounts or, to the extent they are, their measurement is based on formal sector indicators.’ When dealing with statistics, it is safer to keep all the caveats in mind.”
Prime Minister Narendra Modi used these figures at an election rally at Maharajganj in Uttar Pradesh. “On the one hand are those [critics of note ban] who talk of what people at Harvard say, and on the other is a poor man’s son, who, through his hard work, is trying to improve the economy,” he said. This was indeed a powerful political rhetoric that may well resonate with the people. But does it really address the problems relating to big data collection, its analysis, and the models?
Ever since the government decided to change the base year for GDP calculation from 2004-2005 to 2011-2012 under Prime Minister Modi’s regime, there are more questions than answers about almost all our economic data. We are still not sure about the quantum of notional increase as against the real increase in GDP because of this shift. Second, to have a comparative analysis, we need data that are aligned to a set of rules and categories without introducing a new variable. But this was not available even for the latest Budget figures.
The Budget that was presented on February 1, 2017 removed for the first time the distinction between plan and non-plan categories. The government also merged the Railway Budget with the Union Budget. One has to first actively disaggregate the figures sector-wise and department-wise to compare the figures with earlier estimates and arrive at some meaningful comparisons. In this context, there is no conclusive method to understand the real impact of demonetisation on India’s growth story. According to Pronab Sen, former chief statistician of India, the informal sector in India accounts for about 45% of gross domestic product (GDP) and nearly 80% of employment. If this sector is not taken into account, then the metadata not only remains inadequate but also may be seen as a deliberate move to mislead.
While this newspaper has done an excellent job in reporting the official figures and explanations, there is a gap in interpreting and critiquing big data. With policy decisions becoming a product of mathematical models and data, it is worth creating in-house expertise in this crucial area as the next step in public interest journalism.
readerseditor@thehindu.co.in
