One of the most visible macroeconomic successes in India over the last three years has been the sustained moderation of headline consumer price index (CPI) inflation. CPI (industrial workers) averaged 9.5% for six years between 2007 and 2013. For the last 30 months, however, it has averaged 6%—despite the fact that two of the last three years have witnessed a drought. Furthermore, it’s not just volatile food and fuel prices that explain the disinflation. Core inflation averaged 8.2% between 2007 and 2013, but has averaged 5.7% since 2014.
As celebrated as these numbers are, there has not been a rigorous analysis on what’s driven the disinflation. This is not a trivial question. There are many potential explanations. Oil and commodity prices have collapsed, the growth of minimum support prices (MSPs) in India has been curtailed sharply, the monetary policy regime has undergone a shift, fiscal deficits have edged down, and the government has tried to de-bottleneck food supply chains. So which of these factors, and in what combination, have contributed to the sustained disinflation? This is an important policy question because it points to the durability, or lack thereof, of the disinflation.
In a joint International Monetary Fund working paper, Pankaj Kumar and Prachi Mishra of the Reserve Bank of India and I try to get to the bottom of this question, by studying inflation dynamics over the last 15 years in India. We estimate an augmented Phillips Curve for India, where we model inflation as a function of the “output gap” (whether the level of activity is above or below potential), inflation expectations, global factors such as crude and food prices, and India-specific variables, including MSPs and rural wages that can have an impact on inflation, quite apart from the state of the business cycle. Some quick thoughts on this framework. The output gap is a summary indicator of slack in an economy, and therefore should be a key driver of pricing power and inflation. It is natural to ask why interest rates and fiscal policy are not entered as separate explanatory variables. Simply because, ultimately, inflationary pressures should stem from how much slack there is in an economy. Monetary and fiscal policies operate by changing the output gap and, therefore, their impact is embedded into the output gap.
Second, modelling inflation expectations is fiendishly challenging. We remain agnostic by assuming expectations are both backward and forward-looking. Backward-looking expectations are influenced by past inflation and, therefore, are modelled as lags of inflation, as is standard practice. Lags of inflation also capture the persistence of the inflation, and control for econometric problems such as “serial correlation”. Forward-looking expectations are even tougher to proxy. In India’s case, we exploit the fact that there was a seismic change in the monetary policy regime that may be influencing expectations. We, therefore, introduce a binary variable that gets switched on when the new regime came into being (January 2014) to capture its impact on inflation expectations. While this is admittedly crude, there are few easy options.
Third, we grapple with the fact that there is a lot of simultaneity between MSPs, wages and inflation. What’s driving what? We know MSP setting is influenced by inflation. But MSPs also drive future inflation! The same is true of wages. Wages clearly have an impact on inflation. But Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA) wages were indexed using past inflation. As if this is not enough, higher MSPs—by boosting demand for agricultural labour—potentially drive wages. But exogenous shocks to wages—by pushing up input costs for farmers—increase the clamour for higher MSPs. All told, MSPs, wages and inflation are all simultaneously determined, making estimation an econometrician’s nightmare.
So, what are our key findings?
First, that output gaps matter for inflation—a fact universally accepted but questioned in India. What this suggests is that monetary and fiscal policies influence activity and inflation.
Second, up to eight lags of past inflation influence current inflation. This reflects the persistence of the inflation pressures in India—households expectations are very “adaptive” and institutional mechanisms in India (MSPs and wage setting) are backward-looking and help propagate shocks to inflation.
Third, MSPs clearly matter, but less so, once past inflation is controlled, reflecting that MSPs are set taking past inflation into account.
Fourth, the new regime dummy is significant, suggesting that forward looking inflation expectations are being influenced by the regime change. That said, we remain agnostic about what this dummy variable represents, because it could also be capturing changing expectations on future oil prices that began a dramatic descent the same year, three-quarters after the new regime came into being.
Finally, oil prices matter, as do rainfall shortages, but only when shortages exceed 15%.
These results reflect the “average” effect in our 15-year sample. Next we use the estimated coefficients to decompose how much of the recent disinflation can be attributed to each of these factors. Almost 40% of the disinflation can be attributed to a moderation in “adaptive” expectations as well as the fact that disinflationary shocks were accentuated by backward-looking institutional mechanisms such as MSPs and MGNREGA wage indexation.
Twenty per cent of the disinflation can be attributed to a reduction in MSP growth, even after controlling for past inflation (i.e. the discretionary component of MSPs); another 30% can be explained by some combination of the new regime and commodity prices potentially influencing forward-looking expectations. Interestingly, and contrary to popular perception, the actual fall in global oil prices (apart from any impact on inflation expectations) explains very little of the disinflation, because such a small fraction was passed on to households. Finally, the output gap is estimated to have hardly changed over the last two years when inflation fell, and so cannot explain much of the disinflation.
In sum, modelling inflation is a mind-bogglingly complex process in India. There are no glib answers. The conventional explanations—inflation moderated because of good luck (the collapse in oil) or a large growth sacrifice (a sharp compression of output gaps)—is not supported by the data. Instead, we find the disinflation appears much more durable as expectations have begun to moderate, and Indian policymakers—in Delhi and Mumbai—deserve far more credit than they often get.
Sajjid Z. Chinoy is chief India economist, JP Morgan.
Comments are welcome at email@example.com