Cartesian doubt over GDP growth figures is for the good

Former CEA Arvind Subramanian’s paper questioning conventional understanding of India’s GDP growth figures raises many relevant points; should be debated with more seriousness

In Western philosophy, Rene Descartes is known for his ‘methodological doubt’. His method consisted of the critical examination of every dogma and certitude. Everything that can be doubted must be. At the end of this nihilistic exercise, Descartes reached what he called the ‘indubitable truth’. This ‘indubitable truth’ consisted of the realisation that he thinks, and therefore he must be existing. Consequently, Cogito Ergo Sum (I think therefore I am) has become possibly the most quoted aphorism in the history of Western philosophy.

It is entirely appropriate that the ex-Chief Economic Advisor (CEA) Arvind Subramanian has opened his paper on the Gross Domestic Product (GDP) estimation with a quote on Descartes. For his paper is a sceptical look at the conventional understanding of India’s GDP growth figures.

The broad conclusion of the paper is that the Indian GDP is inflated by about 2.5 percentage points. Subramanian reached this conclusion by examining the correlation of the GDP series with a list of 17 macroeconomic variables. Before the recent base revision, these variables moved together with the GDP growth. After the revision, this stable relationship has broken down, implying that the GDP growth figures are somewhat overstated.

Predictably, his paper has stirred the proverbial hornet’s nest. However, even if his estimates are somewhat tentative, his paper is an important contribution to the discussion on the Indian economy and should be debated with more seriousness. This is so due to the following reasons.

First, GDP growth figures are not intrinsically useful. Their usefulness stems precisely from the fact that they are in a stable relationship with a number of economic variables and can help predict them. If that stable relationship has broken down, there is a need for understanding it better.

An example may clarify. Imagine a firm producing a basic commodity like cement or steel. Buoyed by heady GDP growth figures, the firm projects its future demand optimistically, and expands its production capacity by investing in new greenfield projects. It finances the project by taking loans from banks. Banks too assess the feasibility of the project using the equally panglossian demand growth figures.

The deal is struck and and project commences. Then reality strikes; demand is not growing as fast as expected. The firm is burdened with principal and interest repayment obligations. It asks for debt restructuring and ends up stressing the banks’ balance sheets as well.

Now ask a simple question: Whether the bad investment decision was made due to the slower GDP growth figure or the breaking down of the correlation between GDP growth and the demand for steel? Does it even matter as far as the economic implications are concerned?

To put the matter succinctly, for the optimal investment decisions, both GDP growth figures and their relationship with economic variables like the demand for steel and cement are material. If either is problematic, we need to debate and understand it better.

Second, Subramanian’s paper has an implication for mass consumption. To see this point in the simplest possible setting, imagine two farmers who cultivate rice. At the end of the year, both increase their sales and profits by the same percentage points. But they do it in different ways.

One of them increases the quantity produced; the other shifts to better quality varieties like Basmati. Both will make similar contributions to the GDP growth figures, yet there is a key difference. The farmer who increased the production in his farm would be able to feed more people. His greater production will eliminate hunger and malnutrition. By contrast, the farmer who shifts to better quality variety like Basmati will have no such impact. He would feed exactly as many people as he used to do in the last year. The point is that the volume based indicators are relevant for mass consumption and general welfare in their own right. If they are diverging from the GDP growth, we should be curious.

What the government should do to deal with the Subramanian critique? The simplest answer would be to shoot the messenger. That would be a mistake however.

Rather, the government should take a positive approach and overhaul the statistical system.

In the long run, the GDP estimates should be based on the GST filings. GST is computed on the value added, exactly the variable used in GDP calculations. It is backed by solid documentation. Additionally, it is a high frequency indicator, making estimation of quarterly GDP figures easier.

In the interim, the government should move towards academic standards of disclosure which require the reproducibility of the estimates. Currently, Central Statistics Office (CSO) – the nodal agency responsible for GDP calculations – releases a document called ‘National Accounts Statistics: Sources and Methods’, which contains the GDP computation methodology.

It is a long, tedious document and does not include important parameters, making it impossible for external experts to independently cross check GDP growth figures. Additionally, some datasets used in GDP estimation are not in the public domain. A better way for the CSO would be to release the code and datasets used in the GDP calculation in the public domain, after removing identifying information.

This will go a long way towards establishing the credibility of the GDP growth figures which should be as ‘indubitable’ as Descartes would have wanted them to be.

(Avinash M Tripathi is Associate Research Fellow (Economics) at Takshashila Institution)

(The views expressed above are the author’s own. They do not necessarily reflect the views of DH)