Is GDP data a reliable way to measure the health of the economy?

Indian Express, January 4, 2022

The primary yardstick that analysts use to measure the economy’s health is GDP. Economists, technocrats and journalists cite GDP numbers when crafting their narrative about how well the economy is recovering from the pandemic. The Reserve Bank of India and multilateral agencies use GDP statistics to make claims about the future growth path. Yet no one seems to be asking the most important question: How reliable are the Indian GDP data?

The CSO released the current GDP series in 2015, using 2011-12 as its base year. Since then, the new series has been embroiled in controversy. Scholars have pointed to measurement problems, both in the nominal GDP numbers and the real GDP growth rates. Yet none of those problems has been addressed by the CSO, to the best of our knowledge. As a result, the measurement errors still persist.

There are three major reasons why the GDP data, and hence any narrative of economic recovery based on it, are questionable.

First, the growth rate of real GDP is contaminated by the "double deflation problem". Simply put, the CSO calculates real GDP by gathering nominal GDP data in rupees and then deflating this data using various price indices. The nominal data needs to be deflated twice: once for outputs and once for inputs. But the CSO – almost uniquely amongst G20 countries – deflates the nominal data only once. It does not deflate the value of inputs.

To see why this is a problem, consider what happens when the price of imported oil goes down. In that case, input costs will fall and the profits recorded by Indian firms will rise. This increase in profits is merely the result of a fall in input prices, so it needs to be deflated away. After all, GDP is meant to measure the amount of production in the country, which hasn’t changed, at least in the first instance.

But the CSO doesn’t deflate away the increase in profits. Instead, it records a purely nominal increase as a real increase in GDP, thereby overstating growth. Simulations have shown that this effect can be substantial. For further information, see my article here.

Since the cost of inputs is measured by the WPI, a crude measure of the overestimation caused by the absence of "double deflation" is given by the gap between the WPI and the CPI. In the 2014-2017 period, oil prices plunged, causing the WPI to fall sharply relative to the CPI. This meant that real growth was probably overstated.

In the last few months, the exact opposite has been happening. WPI inflation is soaring, reaching 14 percent in November, while CPI inflation has "only" been 5 percent. The rapid increase in the WPI relative to the CPI is imparting an upward bias to the deflator, which increased at the remarkable rate of 8 percent in the second quarter of 2021-22. If this deflator is being overestimated, then real GDP growth rate could be underestimated right now.

A second reason why growth might be underestimated is that the CSO has not updated the sectoral weights. When the CSO calculates GDP, it takes a sample of activity in each sector, then aggregates the figures by using sectoral weights. To make sure that the weights are reasonably accurate, the CSO normally updates them once a decade. It has now been more than 10 years since the weights were changed, and there are no signs of a base year revision. As a result, the sectoral weights are still based on the structure of the economy in 2010-11, when in particular the information technology sector was much smaller. In other words, the fast-growing IT sector is being underweighted, which implies that GDP growth is being underestimated.

But before we jump to conclusions, we need to take into account the third measurement problem – which works in the opposite direction. Measurement of the unorganised sector has always been difficult in India. Once in a while, the CSO undertakes a survey to measure the size of the sector. In the meantime, it simply assumes that the sector has been growing at the same rate as the organised sector. This practice was working well when the two sectors were moving in tandem.

However, starting in 2016 the large unorganised sector has been disproportionately impacted by a series of shocks. First came the demonetisation shock of 2016, which was a severe blow to cash-dependent firms in this sector. Next came the implementation of the Goods and Services Tax (GST) from 2017 onwards, which necessitated a particularly difficult and costly adjustment for unorganised sector enterprises. Then in 2018 came serious problems in the NBFC sector, in turn creating problems for unorganised sector firms, since they were heavily dependent on NBFCs for funding. Finally, the Covid pandemic from 2020 onwards was undoubtedly a much bigger shock for the unorganised sector, compared to the organised sector enterprises.

Despite these severe shocks, the CSO does not seem to have made any adjustments to their methodology for estimating unorganised sector growth. They apparently continue to assume that unorganised sector enterprises have been growing as fast as those in the organised sector. In that case, there would be an upward bias to reported GDP growth.

So, what is the bottom line? Can we say whether the latest GDP numbers overstate or understate growth? The answer is no, because the measurement problems go in different directions. Without more information from the CSO on their methodology we cannot say whether the positive factors outweigh the negative one, or vice versa. But what we can be clear about is that there are serious problems with India’s GDP data. Hence any analysis of recovery or any forecast of future growth of the Indian economy based on this data must be taken with a handful of salt!

The problem with re-basing GDP estimates

The Hindu, November 13, 2019 (with R Nagaraj)

In the next few months, the Central Statistics Office (CSO) proposes to replace the gross domestic product (GDP) series of 2011-12 base year with a new set of National Accounts using 2017-18 as the base-year. According to the chief statistician, this will be done as soon as the new consumer expenditure survey and the Annual Survey of Industries (ASI) results become available. Normally, rebasing is a routine administrative decision of any national statistics office.

Background to the dispute

But these are not normal times in India for users and producers of the national accounts. For the past four years there has been a raging controversy over the current GDP figures on account of questionable methodologies and databases used. According to official data, the annual economic growth rate has sharply decelerated to about 5% in the latest quarter, from over 8% a few years ago. The reality may, however, be far worse. Independent studies using multiple statistical methods to validate the official GDP estimates by the former Chief Economic Adviser, Arvind Subramanian, and Sebastian Morris of the Indian Institute of Management, Ahmedabad, have suggested that the annual GDP growth rates during the last few years may have been overestimated by 0.36 to 2.5 percentage points.

Why is there such distrust in the official GDP figures? To understand the origins of the dispute, one has to go back to early 2015 when the CSO released a new series of GDP with 2011-12 as base-year, replacing the earlier series with the base-year 2004-05. Periodic rebasing of GDP series every seven to 10 years is carried out to account for the changing economic structure and relative prices. Such re-basing usually led to a marginal rise in the absolute GDP size on account of better capturing of domestic production using improved methods and new databases. However, the underlying growth rates seldom change, meaning that the rebasing does not alter the underlying pace of economic expansion.

The 2011-12 base year revision was different, however. The absolute GDP size in the new base year 2011-12 contracted by 2.3% (compared to the old series), and the annual GDP growth rate went up sharply from 4.8% in the old series to 6.2% in 2013-14. Similarly, the manufacturing sector growth rate for 2013-14, swung from (-) 0.7% in the old series to (+) 5.3% in the 2011-12 series. Such large variations in growth rates for the same year may be justified if the material conditions of production warranted. But the higher growth estimates recorded by the new series did not square with related economic indicators such as bank credit growth, industrial capacity utilisation or fixed investment growth. Thus began the questioning of the new GDP series.

Demonetisation as shock

The suspicion of official output estimates became particularly intense after the demonetisation of high valued currency notes in November 2016. By most analyses, the economic shock severely hurt output and employment. For example, the Ministry of Finance’s Report on Income Tax Reforms for Building New India (September 2018; convenor: Arbind Modi), provided data on fixed investment in the private corporate sector based on actual corporate tax returns. It shows that the fixed investment to GDP ratio fell sharply from 7.5% in 2015-16 to 2.8% in 2016-17 (suspected to be on account of demonetisation). However, surprisingly, the ratio in the national accounts went up from 11.7% in 2015-16 to 12% in 2016-17.

Similarly, chief economist of the International Monetary Fund, Gita Gopinath’s academic research paper (co-authored) published by the highly regarded National Bureau of Economic Research in the U.S. in May 2019 showed an adverse effect of demonetisation on growth rate. Yet, the official GDP for the year 2016-17 grew at 8.2%, the highest in a decade.

The root of the problem

The source of the problem, according to many economists, is the underlying methodologies for calculating GDP (in the 2011-12 series) which they claim are deeply flawed, as well as the new dataset used in estimating the private corporate sector’s contribution. Some of the recent, prominent criticisms are as follows.

In a first, the CSO estimated value addition in the private corporate sector using the statutory filing of financial results with the Ministry of Corporate Affairs. The private corporate sector accounts for about a third of GDP, and spans all production sectors, and roughly about half of the private corporate sector output originates in manufacturing. The database of the Ministry of Corporate Affairs has been criticised by many as unreliable; hence it is possible that the private corporate sector output has been overestimated. For example, the Ministry’s database on “active” companies — that is companies claiming to have submitted audited financial results regularly for three years — seems to contain many companies that are actually inactive (not producing output on a regular basis). Last year when the National Sample Survey Office (the government’s premier, independent, data-gathering agency), used the Ministry of Corporate Affairs list of active companies to canvass a sample of companies in the services sector, it found that up to 42% of the sample companies were not traceable, had failed to provide the information for the survey, or had failed to provide audited accounts.

For estimating GDP of the private corporate sector, questionable methods are also used for blowing-up unverified “sample” data into estimates for the unknown and varying universe of “working” or active companies.

State domestic product (SDP) estimation uses many of the same databases and methodologies used in all-India GDP estimation. The methodological changes made in the 2011-12 base-year revision have adversely impacted the quality of SDP estimates on two counts. First, the Ministry of Corporate Affairs data does not have factory identifiers (that is, location of production units, but only has the location of the company head office); it has distorted distribution of the SDP estimates across States. Second, for estimating value-added in the informal or unorganised sector, State-specific labour productivity estimates are unavailable in the 2011-12 series. Hence the method used distorts output estimation.

The CSO has denied the claim that the underlying methodology is flawed and that there are serious problems with the new database being used. The official response throughout the debate has been that the 2011-12 GDP series follows global best practices (meaning, following the latest United Nations System of National Accounts guidelines) and applies better methods using much larger datasets; hence the official estimates are blemish-less. This ignores the fact that India has always followed UN guidelines, and that larger data sets are not necessarily better.

Need for a review

The proposed change over to a new base-year of 2017-18, is, in principle, a welcome decision. However, considering the methodological disputes and data related questions relating to the current national accounts series, as illustrated above, what would the rebasing potentially accomplish? Doubts will persist so long as the underlying methodological apparatus remains the same; feeding it with up-to-date data is unlikely to improve its quality.

In view of the problems with the current series, a chorus of academic and public voices has proposed setting up an independent commission of national and international experts to review the GDP methodology. The ideal time to do this would be now so that solutions could be found and incorporated into the new GDP series. Conversely, if a new rebased series is introduced without any changes it will only entrench the existing methodological problems, and ensuring that the debate will continue for the next half decade. And as the debate continues, so will the loss of credibility.

Communicating uncertainties in GDP data

Communicating uncertainties in GDP data, Ideas for India, February 19, 2018 (with Amey Sapre).

Studying revisions in Indian GDP data

Studying revisions in Indian GDP data Ideas for India, November 28, 2017 (with Amey Sapre).

GDP conundrum: Is India booming?

GDP conundrum: Is India booming? Ideas for India, November 16, 2016.

The great Indian GDP measurement controversy

The great Indian GDP measurement controversy The Leap Blog, September 10, 2016.

Real GDP is growing at 5%, not 7.1%

Mint, March 14, 2016

Are our gross domestic product (GDP) numbers credible? Many commentators have expressed their doubts. But no one has yet identified problems with the Central Statistical Organisation’s (CSO) methodology. This is because they have been looking in the wrong place. The problem is not, as many have suspected, in the nominal numbers. It lies in the system for constructing the deflators. This methodology is flawed, yielding exaggerated estimates of the speed at which the economy is growing. As a result, policy decisions such as the Union budget are taken on the basis of a “broken speedometer”.

No one is talking about the deflators, but they are important. That is because the real numbers are derived by taking nominal data on the economy and deflating them by price indices. So, if inflation is understated, then real growth is going to be overstated. And this is what has been happening.

In the latest data released by the CSO we focus on the gross value added (GVA) numbers. This measure is conceptually similar to the old GDP at factor prices. In nominal terms, GVA increased by 7.9% in the third quarter (October-December) of the fiscal year 2015-16, below its usual level of 10-15%. This increase translated into a 7.1% real growth, because the deflator reportedly increased by only 0.7%.

Could India’s inflation be so low? In effect, the CSO is saying that despite India’s booming economy, producer inflation is lower than that of the recession-wracked economies of the West, or even that of Japan, which has been wrestling with deflation since the 1990s (see adjoining chart.) This is not plausible.

How could the CSO have come to such a conclusion? The answer is that in a number of sectors, it has proxied the deflator by using the wholesale price index (WPI). This approach is problematic. For one thing, the WPI and the GVA deflator can move in different directions. To see this, we need to look at value added from the income side. Firms’ value added is paid out to the factors of production, namely labour, land and capital. The biggest share of this income goes to labour, which means that wages form the bulk of the deflator. We do not have good data on wages, but we know that they have been going up. Yet, the WPI has been going down—by 1.5% in the third quarter.

Next, we turn to value added from the production side. When commodity prices decline, the commodity-heavy WPI will decline, as we have just seen. But the GVA deflator will need to increase. This may sound strange, but the reasoning is straightforward. Plunging commodity prices boost firms’ profits, increasing their value added, which is what the nominal GVA measures. But since ‘real’ GVA is measured at constant prices, this increase in nominal GVA needs to be deflated away. Hence, the deflator needs to rise.

In other words, by relying heavily on the WPI, the CSO has been led astray from the true deflator. As a result, the growth estimates have also strayed from reality. Over the past year, the dynamos of the economy—investment and exports—have remained mired in a slump. Yet, according to the CSO, the economy, excluding agriculture and government, expanded by 9.6% in the third quarter. This growth was powered by two sectors—trade and finance—that reportedly grew by around 10%, and one—manufacturing—that grew by 12.6%. In other words, the CSO is telling us that good times are back. But this is not consistent with what we see all around us.

If the WPI is such a poor proxy for the deflator, what numbers should be used in its place? There is no single answer. For the service sectors, one could look at the consumer price index (CPI) for services such as health and education. For manufacturing, one could take the CPI for clothing/footwear and miscellaneous household goods. All of these indices are running at 5-5.75%.

So, let us be conservative and use 5% as the deflator for finance and manufacturing. In that case, financial real GVA would show a growth rate of 2.6% as opposed to the reported 9.9%.

For the manufacturing sector, only 70% of the estimated GVA is derived by deflating the nominal numbers from the corporate sector, while the remaining 30% is based on the Index of Industrial Production. Accordingly, one needs to take 70% of the nominal GVA, deflate it by 5%, then add back the remainder. In that case, real growth of manufacturing GVA comes to 7.7% as opposed to the reported 12.6%.

With these adjustments for finance and manufacturing, keeping the numbers for the other sectors unchanged, we get a real GDP growth rate for the third quarter of 5%. In other words, the economy is struggling, not racing ahead. Now, that seems consistent with what we are seeing.