Can GDP ‘nowcasts’ help a data-strapped RBI formulate monetary policy? It’s worth a try

Reddy was making an important point in his own witty way. RBI governors are not clairvoyant. They need data to make good policy decisions. Whether to raise or lower interest rates, or hold them steady, depends on the economy’s macro fundamentals; in particular, whether GDP growth needs support or not, juxtaposed with RBI’s primary responsibility to ensure price stability.

But data needs to be both timely and reasonably accurate—not subject to frequent or wide revisions, i.e.—if it is to serve any meaningful purpose.

For RBI governors, tasked with framing policy against a host of macro indicators and criticized at times for their failure to read signals correctly, the importance of clarity on the state of the economy cannot be over-emphasized.

Take GDP estimates, a critical input for RBI’s monetary policy panel. Apart from the inherent handicap of these numbers offering a mirror to the past while monetary policy is forward-looking, they arrive after a long lag. In the case of quarterly GDP, data appears two months after the quarter is over. For annual data, the lag is much longer.

Admittedly, collecting data on macro variables like GDP in a large and complex economy with a large informal sector is a Herculean task. To give credit where it is due, the wait for final GDP numbers has been reduced from three years to two. But we still have as many as five iterations, with the figures moving in both directions. All this complicates the job of policymakers.

India’s first advance estimates, released in January, are followed by second advance estimates a month later, followed in turn by provisional estimates and then by the first revised estimates before the final estimates are released two years down the line.

Sure, it is possible to track high-frequency indicators, but these are often in conflict, making it hard to form an accurate assessment of how the economy is doing.

This is where ‘nowcasting’ could offer a way out. The idea is to capture information from various recent indicators and then use modelling techniques to roll it all into a composite index. In scenarios of high economic uncertainty, it may prove especially helpful.

As a paper by Indrajit Roy and K.M. Neelima in the latest RBI Monthly Bulletin points out, “There are many high frequency coincident indicators which are correlated with the targeted macro-economic indicator that are available at much shorter time lags."

However, “separating meaningful information from noise" is a “humongous task." Done diligently, information extracted from more contemporary data-sets could deliver an early estimate—or ‘nowcast’—of the GDP reference series before it is published by India’s ministry of statistics.

Unlike GDP compilation, for which we have well-set modalities, nowcasting is still experimental. Nowcasts need to be tested for how closely they track the actual data before they can be deemed useful. Of several alternative models, the paper’s authors suggest using a “two-step-maximum-information" model for accuracy.

While this may be the most robust of the lot, it might be a while before its predictive power gains credibility. Willy-nilly, statistical constructs can have weak links even in the best of hands. What’s beyond doubt is that nowcasts can serve as a valuable aid for monetary policymaking—so long as they supplement existing data-sets and don’t supplant them.