Lies, damned lies and statistics PDF Print E-mail
Wednesday, 22 September 2010 00:00
AddThis Social Bookmark Button

Tax collections are up, as is the index of industrial production (IIP). But remove the impact of the higher tax rates in the budget, and the growth isn’t so high. The IIP looks as funny, growth was slowing till June and suddenly rose in July, but only thanks to an unusual spike in capital goods… The new inflation index shows lower growth, even the old one showed a slowing, but RBI upped interest rates. The first step to understand all this is to talk to the Chief Statistician of India, since if the data is dodgy, the prescription is certain to be faulty.

Fresh into the job from the Delhi School of Economics, TCA Anant’s behaviour is still professorial. He’s at the IIC, a minute before I am, and calls to say there’s no reservation in my name. It’s in my wife’s name, I say as I meet him at the elevator, I’m not a member. I’m one at Habitat, I add, to let him know I’m not completely lacking in all the city’s symbols of power—Habitat is not as classy, but it is something! I’m not a member here either, I’m also at Habitat, Anant confesses. A good start! I’m also dying to know about whether the fact that the Director General of the CSO who has filed a case against the process of appointing Anant had anything to do with the GDP data fiasco that took place when Anant was in Tokyo some weeks ago.

A fresh lime soda is quickly ordered, as is the grilled fish for him and lamb chops for me (no, no bread, thanks), and Anant disarmingly rubbishes the conspiracy theory. “He’s the senior-most officer of the Indian Statistical Service and is naturally keen that the job be with one of them, that’s all.” But did he deliberately ensure that enough checks weren’t carried out on the data before it was released, which meant that Anant’s predecessor, the dapper Pronab Sen, had to step in to see what had gone wrong?

The lenient professor isn’t having any of this. No, he insists, the previous additional director general who was in charge had retired and, no surprises here, hadn’t left a robust check-list of dos and don’ts—Anant’s insistence that it is possible to have missed the contradictions in the data is unconvincing since even the journalists discovered it. It was, he says recounting the facts, the perfect storm… the day the furore took place, the officers were at a CSO seminar in Greater Noida where the mobile connectivity was very poor. He says, however, that he’s in the process of strengthening the process of documentation for all datasets. He’s also putting together a more updated document on the method of quarterly forecasting and will put it up for discussion so that the usual data warriors can give him feedback.

In case of the NSS, Anant recalls that Saumitra Chaudhuri (of the Planning Commission) asked if he could even fill up the NSS questionnaire—it takes 2 hours and that, Anant says, could be why Surjit Bhalla may be right that the NSS systematically underestimates consumption of the rich (“who’ll fill up the form in your house?”). So Chaudhuri’s been invited to the January seminar on the NSS, and so probably has trenchant critic Bhalla.

How does Anant like his first government job? Second, he corrects me, he was member secretary of the Indian Council for Social Science Research when Andre Béteille was its chairman. The ICSSR’s job was essentially funding research institutions—27 of them—and it gave him a chance to get to know what was happening in different institutions. The CDS in Trivandrum, he says, did some very useful work on local governance. The most interesting, of course, was when the newly set-up minority affairs ministry found that while it was supposed to do work in 90 minority-concentration districts after the Sachar report’s findings came out, it had no base-level data. Anant then got in touch with ICSSR-funded institutions—Aligarh Muslim University, Jamia Millia University, Institute of Human Development and so on—and got them to help put together a questionnaire (an expert committee under Pronab Sen then vetted it) and some even helped collect the data from a survey. They had to, he says, take a lot of short cuts to get reasonable district data, “which was all right since we weren’t really looking at standardised national data”. He got to see, Anant says, the real issues that came up in a largish nationwide survey.

So why don’t you use other datasets to validate your data, I use the opportunity to plough in. CMIE has data on investments, on corporate profits and even on industrial production, there’s the tax data, NCAER has an income survey. Anant points to the flaws in them clinically, just as he does for his own datasets: “given how the IIP is constructed with 1993-94 weights, you realise how tricky it is to rely on it” is one sample, “it takes at least 3-4 years to construct a new set of weights and a sampling strategy, given the way our procedures are… that’s an issue”…

Mahesh Vyas, Anant recalls, was at an NIPFP seminar a few days ago, using data from his Prowess database. “I wanted to know how inflation was taken care of in the quarterly/yearly data (the IIP deals in real data, not monetary ones). Were the units in the sample the same for all quarters/years?” He’s also not too sure how projects are selected in CMIE’s investment database. The tax data, he says, is a good source but he needs to work with the finance ministry to figure out how to discount for the changes in tax rates, the concessions granted…

Pretty impressive, his knowledge of handling data, I comment, for someone who taught competition law at DSchool. Having dealt with several knowledge-proof types before, to use Raj Krishna’s phrase, Anant doesn’t even flinch. All his eight electives in DSchool, he says, related to econometrics—“just to contrast this, all of Ram Guha’s were in sociology”. This was then followed up with a PhD from Cornell in econometrics. But, I stammer… Yes, he says, he taught competition law as part of the industry course in DSchool, and he worked on regulation too since he helped design the course.

The challenges? Getting his two wards, the NSS and the CSO, to talk to each other! Right now, the NSS even collects data on industry for the CSO but the computers it uses to process this use different software from the ones it uses for its NSS data. At a meeting of state government officials, he recalls, it was pretty obvious that they talked only to certain divisions, not to others.

The horrors? “I tend not to notice them,” he says disarmingly as Vijay Joshi of Oxford and former chief economic advisor Shankar Acharya walk up. Joshi asks Anant about whether the services sector growth numbers are for real. Anant reacts to Shankar’s comment about disguised unemployment in the NSS/CSO with good humour. As we leave, a joint secretary thanks Anant for giving her six staffers for data analysis but wants another six. For the first time in two hours, Anant is at a loss for words.


You are here  : Home Miscellaneous Lies, damned lies and statistics