What Arvind Subramanian, Ashoka Mody don't get about quarterly GDP estimation

In a social media post, finance ministry has challenged the charges Mody and Subramanian have made, saying that the methodology has been used consistently for years, which is how it showed a ​contraction in the economy during the Covid time. Chief economic adviser V. Anantha Nageswaran has also rebutted their criticism in an article in Mint.

Mody’s criticism centres on what statisticians call “discrepancies". It has little merit, as Anantha conclusively argued. Subramanian’s discomfort is with the quality of deflators used by the National Statistical Office (NSO). These, as he also says, are well-known weaknesses in India’s official estimation process that cannot be addressed overnight (the trouble is that there are no proper price indices for capturing inflation in services, the bulk of the economy). However, for Subramanian to suggest that the way out ought to be to track nominal GDP is problematic.

The debate started when Mody wrote that NSO was “covering up the reality of anemic expenditure at a time when many Indians are hurting, and when foreigners are showing only a limited appetite for Indian goods". His contention is that the “covering up" by NSO led it to estimate GDP by estimating “incomes" in the economy, which put the growth rate at 7.8% YoY in April-June. But if the NSO had instead estimated it by estimating “expenditures" in the economy, growth would have been merely 1.4%.

He conceded that neither of the two measures is free of errors and accepts that both are after all “imperfect" computations for estimating GDP. But in the same breadth faulted the NSO treating “income" as the right one and for assuming “that expenditure must be identical to income earned", which, he wrote, is a violation of international best practice.

Based on these arguments he reached the conclusion there is reason for suspecting that the economy grew at the rate 7.8%; that India’s growth story is “fake", as the growth isn’t coming from components of GDP such as consumption or investment but a statistical entry “discrepancies".

Subramanian and his co-author Josh Felman, who has worked earlier with the IMF and Ministry of Finance, Government of India, followed up with an article, not agreeing with Mody’s arguments, but reaching the same conclusion: India’s economy isn’t accelerating, it’s decelerating.

Their story relies on the NSO’s nominal estimates of GDP. Normally, one would just focus on the real numbers, as they show “real" growth, stripping out the illusory effects of inflation, they write. But since, the GDP deflator (that is used for removing the effects of inflation by the NSO), they argue, is calculated through a methodology that has well-known problems, they prefer to rely on the NSO’s nominal estimates.

Their complaint is that whereas the nominal estimates track the real estimates until the first half of FY23, they decline thereafter – by a whopping 14 percentage points over the past three quarters. “This is a narrative of an economy which has decelerated sharply to very modest levels," they conclude.

Anantha’s defence of the NSO on the issue of “discrepancies" is reasonable. He explains that the discrepancy of 2.8% in the quarterly estimate for April-June simply says that the expenditure side estimate has explained only 97.2% of the income side estimate. It does not mean that the 2.8% that has yet to be explained does not exist. It exists and the estimates of the subsequent quarters will show in which of the components of GDP it exists.

In many earlier quarters, as he explains, the discrepancies figure was negative. That’s because the income side approach does not always give a higher growth figure than the expenditure side. At times the reverse happens. But the NSO consistently states the estimates using the income side. Such as even when the GDP was estimated to have contracted about 25% in the first quarter of 2020-21 due to Covid. “Therefore, the insinuation that GDP is exaggerated does not hold ground," wrote Anantha.

This is credible defence of the NSO’s treatment of discrepancies.

Why are reputed economists such as Mody and Subramanian repeatedly stirring the GDP estimation debate? The trouble this time around, at least, is that they are reading too much into quarterly estimates.

GDP estimation is fraught with several inconsistencies, one of which is the quarterly measure of economic activity. What are quarterly GDP estimates? It is often assumed that these are estimates of GDP during three months of a year. Remember the NSO has not even finished estimating the GDP for the year ended 31 March, which is why only a provisional estimate is available as of now for the year FY23. That estimate will be updated and revised several times over the next two years, as information becomes available.

Then, can NSO really estimate GDP for April - June? NSO’s quarterly estimates are based on the limited information that is available in real time. All that the NSO has for this computation is agriculture production estimates and corporate filings. There's not much other information available. An indicator based approach can only give a crude trajectory, especially when the economy is as complex as ours, not exact estimates.

If it takes six revisions and three years to finalize one year’s estimate, before we can know whether the economy truly accelerated or decelerated, quarterly estimates should be assumed to have limited shelf life and utility.

This is why to question the credibility of India’s GDP performance on the basis of a bellwether indicator lacks merit and the whole debate over quarterly estimates can only be called misguided.

This is not to say that the NSO doesn't need to invest in better estimation techniques or that it does not need new, quality data sources. For instance, an economy as large as India, with all its dependence on services, does not have price indices for services. This is a major gap in the estimation exercise.

The Subramanian-Josh duo’s deflators and nominal versus real GDP criticism actually ought to have been a call for services indices to be constructed so that deflators of higher quality become available.

That well-regarded economists are questioning discrepancies, a component that is less than 2% of GDP, when prices indices are not available for the bulk of the GDP shows how far removed the debate is from the ground reality of estimation and its challenges.