(Photo: Mint)
(Photo: Mint)

Opinion | A test of smell-test variables in use to debunk GDP data

Not all the numbers that critics have examined bear a close enough correlation with GDP growth

In 2015, India revised its method of computing its gross domestic product (GDP) by adopting 2011-12 as the new base year, instead of the previously used 2004-05. Since then, India’s GDP numbers have been under the scanner. Why? Though both the series have a high correlation, growth rates under the new series are a percentage point or so lower than the old GDP series, perhaps reflecting wider coverage of smaller firms in the new series. This has upset leaders of the United Progressive Alliance (UPA), as growth under the new series appears to be stronger under the National Democratic Alliance (NDA) government.

However, this political stance is hardly logical. If the new series has a downward bias in GDP growth, it has probably pulled down growth rates under the National Democratic Alliance (NDA) regime as well compared to the older series. One can understand the political war over GDP numbers, but why are (non-political) economists growing sceptical over the new GDP numbers despite the new series making use of the much wider ministry of corporate affairs data on corporate activity, especially for the services sector? The main reason for the scepticism is the so-called failure of the new GDP series to pass a “smell test"—the GDP data is not corroborated by other economic indicators such as bank credit, automobile and car sales, and electricity consumption.

But I step back to conduct a smell-test of these so-called high-frequency smell-test variables. Let’s begin with bank credit growth. The typical argument is that GDP growth averaged roughly 7.5% from 2005 to 2014, when non-food bank credit was growing at 20% a year. How come India is still growing at 7% though bank credit has been growing only at 9% since 2014? Such an argument assumes that India’s growth is highly credit dependent, almost one for one, but this is far from the truth.

First, the correlation between credit growth and subsequent GDP growth is very low–around 0.22 on a scale from 0 to 1, with 1 being perfect correlation. Second, bank credit is so much more volatile than GDP that there has to be something more than GDP numbers that determine bank credit. Third, and more importantly, there are multiple episodes within the test period of 2005 to 2014 when bank credit and GDP diverged. Between the fourth quarter of 2008 and the corresponding quarter of 2009, bank credit growth tumbled from 26% year-on-year to 11%, but GDP growth rose from around 3% in the first quarter of 2009 to 11% in the same quarter of 2010. This growth is under the old series.

More fundamentally, personal consumption is an important aspect of the Indian growth story, but to a large extent is financed by savings rather than loans, contrary to what is true in the US. A recent Reserve Bank of India study (RBI Mint Street Memo) finds that vehicle sales are driven by fuel prices, not availability of credit.

There has also been a structural shift in the way Indian corporates finance themselves. In the RBI Mint Street Memo, I provided clear evidence that more than 50% of new corporate funding is coming from non-bank sources—either equity, non-banking financial companies (NBFCs) or foreign debt. Hence, one cannot expect a very strong link between bank lending and corporate investment. The bottom line is that bank credit is not a great variable to smell the level of GDP growth.

However, there are some better smell-test variables than bank credit, such as car production. It is true that car production and GDP under the old series had a very high correlation of 0.68. However, the old and new GDP series are also highly correlated—around 0.80 (computed using data for 2004-05 to 2014-15 for which the data of both series is available). Then, it is only to be expected that the new GDP series must have also exhibited very high correlation with car production during that period. Therefore, any smell-test argument for the new GDP series based on correlation is bound to fail because the old and the new GDP series are highly correlated.

Hence, if the correlation between car production and the new GDP series has come down after 2014-15, one is forced believe that it’s because of the structural change in the GDP series. However, the fact is that government spending has been the main driver of economic growth and capital investment over the last two years and personal consumption spending have taken a bit of a back seat. Then isn’t it natural to expect the GDP and car production linkage to be weak?

Lastly, I turn to confidence measures. Both the Business Expectation Index (BEI) and the Consumer Confidence Index of RBI, which generally tend to mirror the state of the economy, are negatively correlated (-0.38 and -0.23, respectively), even with the older GDP series that has a base year of 2004-05. This is a startling finding and needs a separate investigation. The correlation structure has not changed at all under the 2011-12 base year GDP series, even for the recent period. It is negative and of around the same magnitude. Again, it’s tough to see the new GDP series failing the smell-test.

In general, in future, as the underlying sample used to compute GDP expands over time to include a larger section of the unorganized sector, one can expect to discover newer drivers of Indian growth.

It is hard to understand the economic logic that made 108 economists question the new GDP data.

Apoorva Javadekar is an assistant professor of finance at the Indian School of Business

Close