Arvind Subramanian’s method of estimating GDP suggests that Germany overestimates the most; brazil underestimates GDP the most—India only a mild outlier.
Recently measured GDP statistics for India (and the world) suggest trouble. No mis-estimation here. Indian GDP growth has declined from 8.2% in 2Q2018 to 5.8% in 1Q2019—one of the largest three-quarter declines in the last 15 years—and if fiscal years 2009, 2010 and 2012 are excluded, it is the third worst decline.
In the recent debate over ex-CEA Arvind Subramanian’s (AS) allegations that GDP growth in India was overestimated by an average 2.5% a year, many commentators have commended AS for his astuteness and bravery in making a much-needed and correct call over the “fudging” (there is no other word) of official GDP statistics in India. The argument goes as follows—motor vehicle sales down, two-wheeler inventory at highest levels ever, no private investment, animal spirits have disappeared; see, AS is right, GDP growth is being overstated. But, as just documented above, official GDP data is documenting the reality of GDP growth being way down.
The government also gets it. Every day, there is an announcement of concern and admission that the economy is in trouble. All eyes are rightly on the Budget to be presented on July 5. It is to be seen whether the finance minister Nirmala Sitharaman listens to the voices of “old” economists and bureaucrats who want to continue with business as usual, be concerned with the minutiae of the fiscal deficit, and ask for restraint on changing course on three world-records that India holds—highest real interest rates, highest effective corporate tax rates, and the worst labour laws.
The same “experts”, bureaucratic or otherwise, asserting and demanding that international experts be called in to look at Indian statistics (because they are allegedly not capturing one of the worst domestic, and global, downturns!) also argue for restraint on any policy action, e.g., don’t change policy rates, don’t lower tax rates—indeed, raise them to gather more revenue to finance the increased fiscal deficit brought about by the slowing economy. It doesn’t get any crazier.
But, maybe it does, in the form of AS’s “academic” calculation that Indian GDP growth is being overstated since 2011. Before looking at this AS miscalculation, I must remind readers that AS was among the few (along with self!) who had the courage to point out that the MPC under Governor Urjit Patel was leading India on a downward growth spiral; that real policy rates in India needed to be 200 bps lower than where Patel’s MPC had kept them. That was in June 2017; in June 2018, the Patel MPC was busy hiking rates and expecting growth (and inflation) to accelerate; a year later, Governor Shaktikanta Das’s MPC has cut rates by 75 bps, but real rates, at 3%+, are where they have been for >2 years.
Real interest rates in India are high(est) because of three policy failures—failure of the finance ministry to reduce government-controlled deposit rates (e.g., on small savings); RBI’s decision to keep repo rates high in the mistaken belief that there is an inflation dragon waiting to be slayed; and the policy makers’ belief that capital markets, including well-capitalised NBFCs, must not be opened up to foreign investors.
GDP growth is low because of policy failures—true today as well as before. Hopefully, recognised by all. But, now to AS’s allegation that Indian GDP between 2012 & 2017 had been mis-estimated and that the “actual” GDP growth in these years was as little as 3.5-5.5%, not the official 7%, i.e., an overestimation of 2.5% a year (7% – 4.5%).
I want to examine AS’s hypothesis and results with the assumption/view that AS is entirely correct in his assumptions, and method of analysis. AS’s assertions rests on three pillars:
Pillar 1: Growth in four real variables (exports, imports, credit and electricity-hereafter X variables) can more than adequately proxy real GDP growth for all non-oil exporting countries with population greater than 1 million;
Pillar 2: That, for all countries, the relationship is robust for two different time-periods; period I being 2001-2011 and period II being 2012-2017;
Pillar 3: Only for India is there a problem with official GDP data. Hence, AS’s analysis is geared to examine how much Indian GDP in period II veered off the (AS) predicted path.
AS brings all his statistical acumen to confirm that the gap between actual and predicted GDP was as much as 2.5% and that this gap was statistically significant (i.e., could not have happened by chance). Since AS believes his model can proxy growth, he is broadly right in also believing that the “only” explanation for the gap between official and predicted GDP growth is that the former, and not the latter, is in error—fudged either by political masters, or incompetence of statistical authorities around the world that vetted India’s GDP measurement, or both. I want to accept AS’s method and conclusions, if only because the two of us were lonely warriors against the Patel MPC crusade against inflation and growth. For 89 countries, I relied on World Bank data for the four AS variables (as does AS). I successfully reproduced his estimate of 2.5% (baseline, column 1, page 11 of his paper, India’s GDP Mis-estimation…).
When I first read AS’s paper a week ago, I was struck by the absence of any discussion on the statistical possibility that his method could yield mis-estimation errors for other countries. He does have a throwaway line that there were four outlier countries (Cambodia, Tajikistan, Ireland and Ukraine) excluded from analysis, but no more. What is sauce for the goose is also sauce for the gander; utilising that truth, I decided to estimate the AS model for all 89 countries, i.e., estimated the gap between measured GDP and AS predicted GDP in period II.
Here is what I found. Out of 89 countries, for 46, the AS country dummy was not significant. For 22 of these 46, the individual country effect was negative, i.e., measured GDP was less than predicted GDP by an average 0.5 ppt; for the remaining 24 countries, measured GDP growth was above predicted growth by 0.4 ppt. The remaining 43 countries, with significant individual country effects, were almost equally divided between overestimation (1.7 ppt) and underestimation (1.8 ppt).
This last result is significant. There is equal overestimation and underestimation of GDP in the world (at least for 43 countries). AS is concerned with overestimation. There are 21 such countries, and Germany tops the list, i.e., according to AS, German (ECB take note) GDP data is being overestimated the most. AS (and his “intellectual” supporters) should train their statistical guns at Germany for systematically overestimating GDP by an average of 1.8 ppt a year in period II.
One final calculation. Reported GDP growth for Germany in period II was 1.3%; AS overestimation number, 1.8%; hence, excess GDP over reported GDP (ratio of 1.8 and 1.3) is 135%, the highest in the world. Number 8 is Bangladesh with an excess of 60%; India is 16th (out of 21 countries) with an excess magnitude of 38% (ratio of 2.5 and 6.7%).
The accompanying table also reports the excess magnitude, etc, calculation for several other countries. For example, reported GDP growth in Brazil was –0.4%; AS method suggests that Brazil growth is underestimated by as much as 3%; Jamaica has a positive average growth in period II of 0.6% and AS deems it to have underestimated growth each year by 2.4%. Maybe, a happiness index can be constructed on the basis of the AS methodology, rather than a GDP mis-estimate. One strong result—Brazilians are a lot happier than the Portuguese; and Jamaicans are the happiest. I would have believed that if the West Indies were doing well in the World Cup.
(The author is Contributing Editor, Financial Express. Views are personal. Twitter: @surjitbhalla)