Economists vs statisticians — the battle being fought over the soul of India’s GDP data

Economists say there are weaknesses in India’s GDP data. But statisticians claim the accusations are based on flawed understanding, saying while GDP has problems, the economists are looking in the wrong places.

Statistician TCA Anant and economist Arvind Subramanian | YouTube/PTI

New Delhi: Even as several international agencies have upgraded their economic growth estimates for India for the current financial year, a conceptual battle is raging in the country between economists and statisticians over how accurate India’s data actually is. 

Some economists have been pointing out several conceptual ‘flaws’ in how the country is estimating its gross domestic product, and strongly imply that the Modi government is overstating growth numbers. 

Statisticians, on the other hand, point out that economists don’t quite grasp the deeper statistical methods being used and are shining light on the wrong issues. That is, there are problems with the way India calculates its GDP, but they aren’t the ones the economists think they are.

The back-and-forth covers a wide variety of issues, from consumption levels, to how expenditure is measured, to how households are really spending their money, and whether the government is artificially making its GDP data look better than it is. 

In the economists’ corner, we have three noted names: former chief economic advisor Arvind Subramanian, who was in the Ministry of Finance from 2014 to 2018, including when India changed the way it measured GDP in 2015; Ashoka Mody, visiting professor at Princeton University; and Arun Kumar, who retired as professor of economics at the Jawaharlal Nehru University. 

On behalf of the statisticians, we have two former Chief Statisticians of India: Pronab Sen, who is also currently the chairman of the Standing Committee on Statistics, and T.C.A Anant. 

Anant was the chief statistician from 2010 to 2018 and was intimately involved with the changes to our statistical system made during that period. Those changes are still in effect. 

We also have insights from Surjit Bhalla, a highly data-focussed economist and commentator.

ThePrint has also spoken to current officials in the Ministry of Statistics and Programme Implementation to get their views. Unfortunately, they have declined to be officially quoted, saying they are not authorised to speak to the media. 

The core issue

The entire debate stems from the economists — of whom the three cited above are merely representatives — pointing out problems with the GDP data. 

Their ire is directed at three elements. The first and most recent issue is that while the overall GDP growth seems robust, consumption, according to other sources of government data, appears tepid. This, as Subramanian put it in an article on Monday, was “a sign of serious measurement problems”. The second issue, highlighted by Mody and Kumar, is the ‘discrepancy’ between two methods of estimating GDP — the production method and the expenditure method. 

While the former takes into account the total production of goods and services, the latter looks at the money spent on the purchase of these goods and services. 

In an ideal world, they would be equal. That is, for example, the value of what the farmer sells the buyer should be equal to what the latter pays the former. 

Mody argues that it’s not the case, and that the government is taking advantage of this by choosing the method that makes growth look better. 

The third issue, also raised by Subramanian, is to do with how the Indian system accounts for the effect of inflation on GDP growth. 

India presents its GDP numbers in two forms: current prices and constant prices. GDP at current prices — or nominal GDP — gives you the value of output at the prices prevailing at the time. GDP at constant prices — or real GDP as it is called — attempts to provide the value of that output after taking out the often distortionary effect of inflation. 

Take the example of a car manufactured and sold at Rs 5 lakh in 2011, and the same model manufactured and sold for Rs 12 lakh in 2021. Now, Rs 12 lakh is more than twice Rs 5 lakh, so has output increased by that much? Is this an increase in output or has the value increased because of an increase in prices? 

Measuring GDP in current and constant prices gets around this distortion caused by the change in price levels. However, in order to arrive at the real GDP figure, the nominal figure needs to be adjusted downwards using a “deflator”. 

This deflator is a statistical tool derived from our inflation data, and is also the core reason for the economists’ third issue with India’s GDP data. Subramanian argues that we are using the wrong deflator and so our real GDP growth is actually significantly lower than what the government is presenting.  

What’s wrong with consumption? 

In an article published in the Business Standard Monday, Subramanian and his co-author Josh Felman said that, much like Judas betrayed Jesus, the Household Consumption Expenditure Surveys (HCES) released by the government had revealed the weakness in the consumption numbers in the GDP numbers. 

The HCES is released as part of the National Sample Survey (NSS) series of data releases, while the GDP data forms a part of the National Accounts Statistics (NAS).

Further, the authors say that the consumption number measured in NAS, in the form of Private Final Consumption Expenditure, was less robust than the level of consumption arrived at in the NSS. 

The authors acknowledge that comparing the GDP estimates with the HCES does have its issues because “there are and always will be level differences between the two”. 

However, they go on to say that the direction of movement in the two datasets can be compared. That is, are both growing, are both slowing or are they diverging? The authors show that there is indeed a divergence between the two, which, they said, was how “Judas” (the HCES) was betraying “Jesus” (the GDP data). 

“That means there are two possibilities: either the NSS estimates are better than the NIA estimates; or the opposite,” the authors wrote. “Surely, both sets of estimates have their flaws, which means that different analysts will have different opinions. We favour the former view.”

This is where the statisticians raise their objections. They believe that the NSS data is actually inferior to the NAS data, although they do acknowledge some issues with the latter. Their main concern, however, is that comparing the two is a flawed exercise in the first place. 

“What he (Subramanian) has tried to do is infer directional movement for national accounts consumption from the NSS consumption, which completely misses the point,” Anant explained to ThePrint. 

The major issue is that NSS consumption is “problematic” not just because of its survey design, but also because it’s failing to adequately measure key expenditure components such as health and education spending by households, he added.  

“The NSS also does specialised estimates for education and health,” Anant further explained. “Separate reports are released for each. Those surveys are designed specifically to capture health and education expenditures, respectively. They invariably reveal that households spend more on health and expenditure than what the HCES shows they do.”

In other words, the HCES was not accurately measuring how much people are spending on health and education, which is a serious issue because, as Anant noted, these two components are forming increasingly large shares of household expenditure, while the share of food is falling.

Anant is not the only one raising objections about the quality of the HCES data. Surjit Bhalla, in an interview with ThePrint, has previously talked about how the 2017-18 edition of the HCES — which the government scrapped — contained “the worst data he had ever seen”.

Pronab Sen, too, says that comparing consumption and production — which is what GDP captures — is fraught. 

“The data on the production side and expenditure side are of varying quality,” he explained. “The data on the expenditure side (of which consumption is a part) is actually quite poor.”

The other issue, he said, was that we are largely concerned with the real GDP numbers and not the nominal ones, which means that the GDP deflator used becomes an important factor.

“You are measuring production largely using the Wholesale Price Index (WPI), but you are measuring consumption using the Consumer Price Index (CPI),” he said. “What we do know is that the WPI is far lower than the CPI. So, what will happen is that you will boost the production estimate because the deflator being used is low. The consumption number is largely okay.”

Put simply, the WPI measures prices at the wholesale level — when goods are sold at wholesale markets in bulk — while the CPI measures the price at the retail level, when it is sold to the consumer. The two indices have different baskets of items that are measured and assign different weights to even the items common between them.

Therefore, the two indices often vary in their estimates of inflation for the same month. Using each of them as deflators and comparing the results would naturally lead to different outcomes. 

So are we getting our deflators wrong?

The first blow in this particular debate was once again thrown by Subramanian and Felman, who, in an article in the Business Standard in September, argued that using the GDP deflator being used had failed the “smell test”.

“The GDP deflator peaked at 11.6 percent in the first half of FY23, then collapsed to just 0.2 percent in the first quarter of FY24,” they wrote. “Does this really correspond to our sense of inflation trends in the economy? Has inflation really been eliminated in recent months?”

In other words, they ask whether the deflator was capturing the true picture of inflation in the country, and whether this was being incorporated in the GDP numbers. They also compared the deflator being used to the level of retail inflation as measured by the CPI and found that retail inflation was significantly higher than deflator in recent quarters.

“The GDP deflator is not hard data,” they added. “It is not measured directly, like the consumer price index (CPI); it is instead a derived number, calculated through a methodology that has well-known problems.”

Further, they argued that the wrong deflator was being used for the wrong sectors. For example, they pointed out that the WPI was used as the deflator for a number of non-traded service sectors, and implied that the CPI would be a better deflator here. 

Here, too, Anant and Sen pointed out some conceptual flaws with Subramanian and Felman’s arguments. 

First, Anant pointed out that most of these non-traded services sectors were not measured by the CPI either, so using that index would run up against the same problem that Subramanian and Felman pointed out — of being inappropriate for the relevant sector.

This was another case of the criticism apparently missing the real issue at hand.

“A far better criticism would have been why we don’t have particular indices to measure price movements in these sectors yet,” Anant told ThePrint. “The Ministry of Commerce has been doing some work on this, but they have not finalised it yet, which is something they should do.” 

Sen, on the other hand, pointed out the conceptual flaw behind Subramanian and Felman saying the deflator was a “derived figure” that could not actually be measured.

“This is not correct,” Sen said. “The WPI is measured directly, where the prices of various goods and services — there is an established basket — are measured from the wholesale markets.”

Sen does point out that both the WPI and CPI are very outdated, based as they are in 2011-12, and need to be updated at least once every five years. But Anant points out that even this updation would likely yield only “marginal” improvements, and that significant changes need to be made in other areas, such as devising price indices for sectors currently unmeasured. 

The discrepancy problem

The third set of attacks on GDP data came from Ashoka Mody, whose article in September — days before Subramanian and Felman’s — accused the Modi government of “beautifying” its GDP numbers by using the production method of calculating GDP instead of the expenditure method. The expenditure method, he said, would have resulted in growth being much slower than portrayed. 

This argument has met with much criticism among statisticians both within and outside the government. Sen, for example, points out that India follows the production approach because our expenditure data is very patchy.

“So, there will be a discrepancy,” he explained. “The level of discrepancy we are seeing is not significant, and can be explained by the fact that the data flows are erratic.”

That is, the discrepancy levels seem magnified when the quarterly GDP data comes out because the data on expenditure in the economy is poor, and not enough of it is available each quarter. 

However, this discrepancy shrinks as more and more data becomes available and is assimilated into the First Revised Estimates of National Income (which comes out a year later) and the Second Revised Estimates (released a year after that). 

The second objection comes from the statistician currently in the Ministry of Statistics who did not wish to be named. 

“The argument that we are choosing the production method because it makes our growth look good is absurd,” he said. “India has always used the production method from the beginning, and we don’t change methods as and when the other one looks better.”

Disclaimer: T.C.A. Anant is related to the author.