How to make sense of India’s official growth estimates

The gross domestic product (GDP) of any country is widely considered to be an important barometer of its economic health. Given that it is used as a universal denominator to standardize levels of public debt (using the debt-GDP ratio) or stock market capitalization (market cap-GDP ratio) across time and space, movements in GDP and its subcomponents attract a great deal of attention. Unfortunately, the process through which GDP is compiled does not get enough attention. Hence, most analyses of the GDP numbers do not take into account their limitations.

This is a curious historical anomaly, since the pioneers of national accounting—Colin Clark, Simon Kuznets, J.M. Keynes, and V.K.R.V. Rao—were all intimately aware of the errors involved in estimating national income estimates. In his 1941 book National Income and Its Composition, Kuznets derived error estimates for each sector of the American economy. His analysis suggested that the estimates of American national income were subject to an error margin of +/- 10%. Rao, alongwith P.C. Mahalanobis and D.R. Gadgil, arrived at similar error margins for India in the final report of the National Income Committee in 1954.

The examples set by these pioneers were lost on their followers. Modern-day national accountants do not provide any error estimates while publishing national accounts figures. This leads many who are unfamiliar with national accounting methods to believe that such errors simply don’t exist.

The consequences of such errors can be staggering, and can be illustrated with a simple example. Consider an economy with high quality databases, Perfect Land, where the reported GDP is estimated with an error margin of +/-1% each year. Suppose the reported size of its economy was 1,000 in 2021 and 1060 in 2022. This means that the actual range for the size of the economy was [990, 1010] in 2021 and [1049, 1071] in 2022. Perfect Land’s reported GDP growth in 2022 was 6% [(1060-1000)/1000] but its actual growth rate could have been anywhere between 8% [(1071-990)/1000] and 4% [(1049-1010)/1010].

Now consider another economy, Imperfect Land, where the databases used for national accounting have several holes. Suppose it reported the same GDP figures for 2021 and 2022 as Perfect Land, but those came with error margins of +/-5%. It can be shown that a reported growth rate of 6% for Imperfect Land would mean actual growth anywhere between 17% and -4%. Year-on-year growth rates in this case almost lose meaning. Any growth comparison between Perfect Land and Imperfect Land would also turn out to be misleading.

India’s national accounts system perhaps lies somewhere between Perfect Land and Imperfect Land. Given the weaknesses in India’s statistical system, annual data on many sectors is unavailable. India’s national accountants use outdated data, heroic assumptions and rough proxies to fill these gaps in the national accounting database. Hence, the first step to make sense of our growth rates is to put an error band around GDP figures and GDP growth rates. You can apply your own judgement on how big those error bands should be, depending on whether you think India is closer to Perfect Land or Imperfect Land.

Second, even if the final GDP estimates of a country are as accurate as those of Perfect Land, the provisional GDP estimates may still have wide error margins and resemble those of Imperfect Land. This is true for all countries, but more so for developing countries such as India, where reliable high-frequency indicators are unavailable for many sectors. So it is wise to factor in wider error bands for provisional or advance or quarterly estimates of GDP. These estimates are calculated using a quick-and-dirty method that Indian national accountants are not very proud of. But it is these ‘kuccha’ estimates that are used most by economists and policymakers.

Third, the error bands for sectoral components of GDP are likely to be much higher than for the aggregate. This also reflects in revision patterns. A 2017 analysis of GDP revisions by the economists Amey Sapre and Rajeswari Sengupta showed that the extent of revisions was much more for some sectors such as mining and trade than for agriculture. Since some sectoral revisions tended to cancel out others, the extent of revisions in the aggregate GDP figure were less than the average sectoral revision.

Fourth, look at a broader set of indicators to cross-check what the national accounts statistics are telling you. If alternate indicators for a particular sector paint a similar picture as the national accounts data, there is good reason to believe it. If there is a mismatch, you should treat the national account estimates with caution. Finally, ignore the circus around growth forecasts. When an army of national accountants is unable to provide a final estimate of GDP in the year gone by, there is little reason to believe that any economist can forecast next year’s growth rate till the first decimal.

The author is a Chennai-based journalist. His Twitter handle is pramit_b