Growth story in shreds

A former senior adviser to the government on economic matters delivers another blow to the credibility of India’s growth rate, but the methodology of his study is seriously flawed.

Arvind Subramanian, former Chief Economic Adviser.

IN no branch of knowledge do academics enjoy as much clout as in the discipline of economics. And, economists who are in close proximity to power enjoy not just clout but also impunity from consequences that they unleash on millions of unsuspecting souls. When the former Chief Economic Adviser (CEA) to the Government, Arvind Subramanian, recently published a quasi-academic paper challenging the official estimates of India’s gross domestic product (GDP) and pegging growth rates at significantly lower levels than the official figures, it would have stumped most Indians. How is it that the same person who ought to have had a commanding view of the Indian economy while in office until last year (Arvind Subramanian was CEA from October 2014 to June 2018), but did little to question the estimates of national output and growth, now authors a paper from the safe distance of Harvard?

Subramanian’s paper, “India’s GDP Mis-estimation: Likelihood, Magnitudes, Mechanisms, and Implications”, published as a working paper by the Centre for International Development at Harvard University, claimed that the Indian economy grew at only 4.5 per cent—not 7 per cent as officially estimated—between 2011-12 and 2016-17. To a government that has already been bearing the brunt of criticism of the obfuscation of official statistics on multiple fronts, this was a big blow, coming as it did from the same man who, until the other day, appeared to sit comfortably with those very statistics.

Subramanian gave a boost to his claims by simultaneously writing an article on the same day in an English daily in India, clearly aiming to maximise the impact of what he considered to be explosive claims. He claimed that methodological changes effected in the new series of GDP (base 2011-12) that was introduced in 2015 after the Narendra Modi government assumed office for its first term overstated the value of the national output. To be fair, the changes in the methodology were initiated by the previous government headed by Manmohan Singh well before Prime Minister Narendra Modi assumed office in 2014.

The government reacted quickly. The same evening, on June 11, the Ministry of Statistics and Programme Implementation refuted Subramanian’s claim and asserted that its methodology was not only right but also complied with international norms on national accounts statistics. The Economic Advisory Council (EAC) to the Prime Minister urged an “academic debate” instead of any attempt to “sensationalise” the issue. It also promised a detailed rebuttal to Subramanian in due course. Crucially missing in both defences were any reference to the controversial nature of the data source, whose examination is also a glaring omission in Subramanian’s exercise.

Two important aspects of the change in the base period for estimating GDP and growth initiated in the last revision appear to have ignited the never-ending controversy over the estimates ever since they were announced first in 2015. The first was a fundamental shift in methodology—a changeover that emphasises the value of output instead of the quantity of output as was done earlier. Several economists and statisticians, including the former Chief Statistician Pronab Sen, have justified with some merit the argument that an emphasis on value would capture better the effect that changes in quality, productivity and indeed the introduction of entirely new products would have on the realistic estimation of national output. Intuitively, the introduction of new variants of cars, computers with better capabilities, mobile phones with enhanced features, or steel plants with better automation, which imply in some cases that a similar comparable product did not exist earlier, indicate that a simple estimate based on quantities of output misses a significant aspect of the national output. Logically, the emphasis on value, rather than quantity, would appear to be better at capturing not just improvements in products but also the gains that accrue from enhancements to quality and productivity. Thus, prima facie , one cannot quarrel with this shift in methodology. The government’s claim that this methodology is followed the world over is also true. The problem appears to lie elsewhere; the second major problem—and an apparently intractable one at that—is the source of data that are used to compute the national output from which economic growth rates are derived.

Dubious data

The shift to the new methodology for the 2011-12 series was based on a shift to a complete dependence on the Ministry of Corporate Affairs (MoCA) for data required for computing GDP emanating from the industries and services sectors. This is where the problems began to get worse to a point where the computation of GDP has become suspect to not only economists in India and across the world but to investors and the general public. Unlike the earlier estimates of industrial output that were based on data sourced from the Annual Survey of Industries, the MoCA database (termed MCA-21) is based on what registered entities file with the Ministry. Evidently, these are little more than balance sheets filed by Indian companies; and anyone familiar with the quality of balance sheets filed by even large corporates, let alone thousands of smaller ones, knows how reliable these are.

To compound matters, the National Sample Survey Office (NSSO) recently revealed that a significant proportion of registered units simply did not exist and had to be weeded out. One Bengaluru-based small businessman told Frontline that all one needs is an address and an electricity bill confirming that address in order to register as a unit. “An electricity bill with a claimed address can be purchased for a price in India, and once you have this, you are up and running as a business unit even though you may be producing zero value,” he said. He pointed out that the mushrooming of such “bogus entities” had quickened after the introduction of the goods and services tax (GST). Although such entities are aimed at creating a channel for fake and bogus tax invoices for claiming fraudulent tax credits, the fact that these units are registered as active entities also falsifies the true extent of economic activity and hence, the value of output, that is, GDP. From here, it is only a small step to mismeasure growth rates. Clearly, the MCA-21 database seeded with bogus entities in large numbers not only misses crucial activity but also overstates activity when none is happening.

It appears that the question of growth rates—a politically sensitive issue to a polity whose basic question in the Modi years is whether he has delivered better growth when compared with his predecessor—becomes even more intractable when the MCA-21 database is used as the source of the raw data. In order to do this, one needs the new data to be extended backwards on similar and comparable lines. But a fundamental change in methodology that has been effected in the new series, with an emphasis on the value of output, requires similar data from the earlier period to also contain comparable data. And, it appears that the MCA-21 database simply does not have this with any degree of robustness. It appears that questionable value assumptions made for data from the past have distorted the data. This implies that the understatement of value of output from the past has resulted in an overstatement of growth. This explains the illusion of growth in the Modi years based on a statistical artefact.

To make matters worse, the MoCA has not been transparent about the manner in which it has “adjusted” the data it has gathered. Economists of long-standing reputation dealing with Indian GDP statistics—such as Professor R. Nagaraj—have complained that they do not have access to unit-level data in order to check for inaccuracies or errors in the data gathered by the MoCA. Above all, the MoCA’s own standing as a statistical office is highly suspect, primarily because it was never configured to perform this important statistical function. It was designed primarily to oversee the compliance of registered Indian companies, a task in which it has obviously failed, given the rising tide of corporate malfeasance and fraud.

When the new series was released in 2015, most professional economists were incredulous. Subramanian himself made a few noises about having to check the compatibility of the new estimates with other observations about the Indian economy that suggested a far more lacklustre performance. Even the then Reserve Bank of India (RBI) Governor, Raghuram Rajan, appeared unconvinced. Fundamentally, sceptics were unconvinced that the Indian economy was clipping along at a nice pace of more than 7 per cent growth rate when most other economic variables suggested otherwise. Crucially, industrial investment had slowed down since 2013, credit offtake by industries had slowed down, agricultural growth remained anaemic, and employment had slowed down significantly. How can these be happening in a fast-paced economy—the fastest in the world, they asked.

Expert somersault

Subramanian’s latest paper indicates that it is an attempt to salvage some credibility, which was evidently injured by his association with the Modi regime. The crucial part of his latest work—one wonders why he could not have achieved this from within the walls of government when he was its chief adviser—rests on his claim that actually growth is about 2.5 percentage points lower than what the GDP estimates suggest. This is a staggering claim, which explains why Subramanian attracted so much media attention immediately. First, it blows a hole through the Modi regime’s claim that the Indian growth story is not only intact but is a world-conquering one. Second, it has immediate policy ramifications; if indeed, the growth rate is not what we have been told it has been, the government needs to intervene quickly in terms of providing support.

But Subramanian’s methodology suffers from serious flaws, which in days to come may well draw flak from professional economists. Significantly, he has not addressed the well-known problems associated with the reworked data for the earlier years, which have been computed from the MCA-21 database, or the serious problems associated with the database itself. Instead, he has basically taken a set of 17 indicators, which he claims are “strongly correlated with GDP growth for the period 2001-2017”. This is a motley group of variables, ranging from sales of commercial vehicles, airline passenger traffic to the Index of Industrial Production and credit offtake.

Using regression analysis, he makes the claim that the overestimation of growth arises from the construct of the new GDP series. This is flawed for several reasons, which probably explains why Subramanian claims that his paper is still a “work in progress” and why the paper remains only a working paper and not published in a peer-reviewed journal. First, the analysis based on the variables he has assembled only indicates that they moved differently from the national output; it does not preclude the possibility that growth may have emanated from other areas that he may have missed, which means that his construct is methodologically flawed.

More importantly, the estimate of national output, that is GDP, requires the statistician to count everything, not just parts of a whole, which is what Subramanian has done. Even more critically, by failing to critically examine the nature and quality of the MCA-21 database, he has missed the elephant in the room! That database appears to be the fatal flaw in the Indian GDP computing system, one that guarantees a prolonged controversy.

There is much irony in Subramanian’s tryst with policy-crafting. For four years he presided as a key adviser, urging fiscal prudence, rectitude and restraint, even as the data, in his own hindsight, would have suggested the opposite course. Only an economist with access to the levers of policy can get away with a somersault of that kind without batting an eyelid.