FREE

Event update: India Economy - Substantial GDP mis-estimation; Urgent policy response needed

Arvind Subramanian’s latest on overstatement of India’s GDP indicates the likelihood of India’s policy automobile being navigated with a faulty or even a broken speedometer, thus leading to overestimation of the speed of growth and eventually, the current size of the Indian economy (by 9-21%). While the paper bases the main cause of overestimation to deflator-related issues (which led to the overestimation of value-added component), our own research further indicates the possibility of over-estimation in nominal GDP too (implying key fiscal ratios are understated by similar proportions). The paper concludes that India’s real GDP growth from FY12-17 should have been ~4.5% versus the reported average of 7%. The author has substantiated his point by comparing 17 hard data-based indicators (credit, 2-wheeler sales, CV sales, etc) with real GDP growth pre-2011 and post 2011. Prior to 2011, these indicators not only posted robust growth, but also had fairly positive correlation with GDP growth. However, post 2011, 11 of these 17 indicators were found to have negative co-relation to GDP, which seems counter intuitive, in our view. As controversies surround over-estimation of India’s GDP, this paper adds to prevailing concerns. The report also gives a perspective on weak corporate performance, banking sector ailments, rising unemployment, lack of investments and weak exports, despite robust GDP growth over last few years. Based on the growth estimates in the report, we believe the current policy support will be insufficient to revive growth, indicating an urgent need for both monetary as well as fiscal policy response.

Extent of overestimation: As per Arvind Subramanian’s paper, the reported GDP growth of average 7% between FY12 and FY17 has been overstated. Actual growth may have been around 4.5%. The report pegs the growth at 3.5-5.5%, with 95% confidence interval. Consequently, the level of GDP might have been overstated by about 9-21% over the mentioned five years. The paper intentionally leaves out estimates for FY18, which it believes were further marred by demonetisation and GST. If FY18 were to have been included in the analysis, the extent of over-estimation would have been even higher, in our view.

Key causes of overestimation: The paper lists three key causes of over-estimation of GDP, two of which are related to issues in usage of deflator, which tends to over-estimate the value created component of GDP. The three key causes of over-estimation are discussed below:

• Moving from volume to value-based estimates in manufacturing: As GDP calculation moved from volume to value-added basis during the transition phase of the base year (from 2004-05) to 2011-12, value estimated had to be deflated by prices to get the real magnitude. While input values should have been deflated by input prices and output values by output prices, the same was not done. Unfortunately, input values were deflated by output prices. Huge fluctuations in oil prices post 2011 impacted input costs dramatically. Hence, when input values were deflated by output costs (which was higher in value compared to input costs series), it underestimated the input series. Consequently, value-added component (difference between output value and input value) was overstated. This brings the author to three key testable propositions:

o  Proposition 1 - Formal manufacturing value-added significantly affected: According to the author, unless there is technical efficiency in the economy, volume growth and real gross value added (GVA) growth should not diverge much. Consequently, significant divergence in correlation between GVA manufacturing with IIP as well as manufacturing exports before and after 2011 is a point of concern. Both IIP and manufacturing exports exhibit a positive correlation with formal manufacturing value-added growth, which turns negative post 2011 (Exhibit 1,2). These results seem counter-intuitive, especially given the lack of significant technological changes experienced by the economy during this period.

o  Proposition 2 - Formal manufacturing value-added growth over-estimated: Prior to 2011, the divergence between real formal manufacturing sector GVA growth and IIP manufacturing growth was small and varied around 0.5 percentage points on both sides (going back to 2001). However, post-2011, the divergence is almost entirely one way; real GVA growth consistently exceeds IIP growth by about 5.9% on average (Exhibit 6), indicating the extent of overestimation of GVA.

o  Proposition 3 – Overestimation of formal manufacturing, led by output–input price wedge: While WPI is a proxy for input prices, CPI is a proxy for output prices. Difference between the two has gone up significantly post 2011 (WPI being significantly lower than CPI). Consequently, usage of output prices for deflating input series, has led to underestimation in real magnitude input. This, in turn, led to over-estimation of the value added component. This is further elaborated by an increased correlation between GVA-IIP difference and CPI-WPI difference post 2011, which indicates price-led overestimation of output.

• Deflating services by manufacturing deflators: While consumer services value has been deflated by relevant CPI services index, producer services were deflated by WPI manufacturing index, which is a cause for worry. Producer services roughly account for 20% of overall GVA. As WPI was significantly lower than CPI post FY15 (Exhibit 7), it lead to overestimation of this particular segment of GVA.
• Using formal activity as a proxy to informal sector growth: As per the author, informal segment accounts for 30% of manufacturing GVA in normal times (5% of overall GVA). As demonetisation and GST severely impacted the informal segment of the economy, there is a fair case of overestimation here, particularly for nominal GDP series as well.

Evidence of overestimation using lead indicator data: The paper presents two correlation scatters for 17 standard “real” indicators with GDP growth pre and post FY11, which are discussed below:

• Exhibit 1, 16 out of 17 indicators show positive correlation to GDP growth before 2011. However, post-2011, 11 out of 17 indicators are negatively correlated with GDP (they fall below the green, horizontal line).
• The same Exhibit indicates that while real GDP growth in the two periods are reported to be similar at 6.9% and 7.5% on average, 11 out of 17 real indicators are below the 45 degree line in the second period, indicating much lower growth. 
• Performance of key indicators, pre and post 2011:
• Exports (goods and services) grew at 14.5% before 2011 and 3.4% thereafter
• Imports (goods and services) grew at 15.6% prior to 2011 and 2.5% after 
• Production of commercial vehicles grew at 19.1% before 2011 and -0.1% thereafter
• Only petroleum consumption and electricity grew marginally faster post 2011 
• Exhibit 1,2, specifically the lower average values for nearly all indicators and the negative correlations post-2011, corroborate the hypothesis that GDP growth was substantially over-estimated during this period.

Policy implications - Indian policy automobile has been navigated with a faulty or even broken speedometer:

• Need for urgent and enhanced fiscal and monetary response: If we were to assume that in fact average growth was 4.5% instead of 7% over FY12-17, it definitely would explain the weak corporate performance, banking sector ailments, rising unemployment, lack of investments and weak exports in recent years. Hence, given the dire state of growth, easier counter-cyclical monetary and fiscal policies are required. We believe this calls for urgent reform measures, as current measures may not suffice for an effective growth revival.
• Quality and integrity of data needs to be improved; India must restore the reputational damage across the board: Indian data on GDP to employment to government accounts has suffered in recent years. Consequently, there is a need for robust central statistical organisation (CSO) institutions and appointment of people of stellar technical and personal reputation.

The detailed paper can be accessed at: and CSOs response to this can be found ​