14 October 2014

Wanted: dogma-free models of inflation

October 14, 2014 

The target of inflation experts at the RBI is 8 per cent for December.

Inflation data for September 2014 was released yesterday and came in at 6.5 per cent, making it the best reading since February 2008, when it was 6 per cent. Last October, the inflation gauge measured 10.2 per cent. That is close to a 4 percentage point decline in one year. But I guarantee you that my brethren in the profession, and experts at the RBI, will still be talking about how this declining inflation estimate is distorted, like many others this year, and how base effects, seasonals, etc explain the entire reduction. And how, just you wait, inflation is going to shoot back up.

In a recent note, monetary experts at the IMF argued that India needs to increase the repo rate in order to successfully reduce double-digit inflation. The target of inflation experts at the RBI is 8 per cent for December. We all get our forecasts wrong every now and then, but less than three months ago, RBI experts were talking about the dangers of inflation not slipping below 8 per cent. And just 10 days ago, RBI Governor Raghuram Rajan talked about the dangers of inflation being above 6 per cent 15 months from now, in January 2016.

If the RBI knows best about inflation, why is it consistently missing near-term forecasts, and by a large margin? One explanation is that it does not have a correct model of inflation or growth, or a model relating the two. If this is true, why should the RBI be trusted with its present model of inflation? It could be that the RBI policy of substantially high repo rates is working to reduce inflation; however, the model needs to communicate the correct trend. A miss by such a magnitude is a structural miss — remember, the “informed’ RBI target for December is 8 per cent. Three months from now, inflation is likely to be closer to 6, not 8 per cent.

But what is the “correct” model of inflation? In several articles since July 2011, including (‘Where monetary policy is irrelevant’, IE, September 13), I have argued the following: that a primary determinant of Indian inflation is the policy variable of minimum support price (MSP) inflation set by the agriculture ministry. In addition, I have argued, indeed challenged, the monetary authorities and/ or researchers to empirically document a relationship between any non-MSP variable and inflation.

So far, this challenge has not been met. However, in an article written by a consultant to RBI (CAFRAL division), senior economist Amartya Lahiri, without offering an alternative model of inflation, claims that MSP inflation was just mirroring lagged CPI inflation. Technically, Lahiri does this by purging MSP inflation of past CPI inflation; interestingly, he does not do the counter-purge. If he did, as discussed below, he would not have obtained the result he did. (‘Don’t blame MSP for inflation’, IE, October 7).

There are two major problems with Lahiri’s conclusion. First, at a conceptual and/ or policy level, how can one argue that the politically and Sonia Gandhi-inspired change in the terms of trade in favour of agriculture were all due to the fact that the UPA experts were just making up for past CPI inflation? Just a simple check of the numbers would have shown Lahiri that he was following the wrong “garden” path. Second, I have argued, and still argue, that MSP affects inflation with a one-year lag. CPI inflation was averaging 4.5 per cent for the nine years from 1999 to 2007 and MSP inflation averaged 5.2 per cent. In 2008, MSP was increased by 27 per cent; and CPI inflation averaged 12.4 and 10.4 per cent in the subsequent two years, and in double digits for each of the next three years. He should also note that CPI inflation is down in 2014 and likely to go down further, because MSP inflation has averaged less than 5 per cent in 2013 and 2014.

Any knowledge of Indian policymaking would dictate that while past inflation is one of the factors entering a policymaker’s mind, it is but one factor; others, especially vote-getting factors, are more important. One is tempted to add international food inflation as an additional factor determining MSP, and therefore CPI inflation. But one should resist such temptations because international food prices go up and down. After all, when was the last time a policymaker suggested a reduction in food procurement prices in India?

Thankfully, we do not have to conduct casual empiricism in these matters. Nobel prize-winning economist C.W. Granger tackled this issue some 30 years ago and, since then, it is standard practice for graduate students and most senior economists to run these tests before they pronounce judgment on causality. According to Lahiri, past inflation causes MSP inflation, and MSP inflation does not cause future inflation. Interestingly, for the period for which we have both MSP and CPI (industrial workers) data, 1976-2013, the results are just opposite to those assumed by Lahiri — causality tests indicate that while MSP causes CPI inflation, CPI inflation does not cause MSP inflation.

All these results, plus the data, have been posted in a post (“FAQs on Indian Inflation”) on my website, ssbhalla.org. For a total of 19 monetary and real variables (for example, money supply growth, growth in mining output, fiscal deficit, rural wages, etc) two regressions were run — the first a simple regression between inflation and the determinant in question and the second a more correct estimation of the model with inflation as a function of lagged inflation (to control for persistence) and lagged determinant (what can the variable explain of inflation that is not accounted for by past inflation). Interestingly, only one determinant, among 19 contenders, is significant in the “correct” regression —
MSP inflation.

In his thoughtful Statistics Day speech (September 25), Rajan wondered aloud whether my results were affected by the use of “strategic dummies”. As the blog makes clear, the results are entirely unaffected by the use or non-use of any dummies, and all the determinant tests reported were conducted without any dummies.

It is important that the determinants of inflation debate be carried out transparently, and to its logical end. Rajan states, citing my results, that “if government-determined MSP drives inflation in India and nothing else, then forget expectations, forget output gap, and the whole exercise of monetary policy is pretty worthless and if that is so, then why doesn’t the RBI cut rates immediately?” The cutting rates part can come later but for the moment, can the RBI, or any RBI consultant, or any researcher, prove a statistical link between inflation in India and any monetary or interest rate, or output gap, or any variable besides MSP? That would be useful to take the analysis and debate forward. Then we can talk about the desperate need of the Indian economy for lower interest rates. Considerable evidence exists that lower real rates help growth — and considerable evidence exists that lower rates do not affect inflation in India. In this disinflation age, there is precious little evidence of lower interest rates affecting inflation anywhere.

Finally, Lahiri wants to shut up commentators by appealing to authority, a well-known inductive fallacy. Here is how he concludes his article on my work: “Intemperate and ill-thought-out assertions about the efficacy of monetary policy can unhinge private expectations of inflation. The RBI has been introducing a new monetary regime over the past year. The changes need time to work. It may be worth letting the experts do their job, rather than complicating matters by creating unnecessary confusion.” (emphasis added).

The writer is chairman of Oxus Investments, an emerging market advisory firm, and a senior advisor to Zyfin, a leading financial information company

- See more at: http://indianexpress.com/article/opinion/columns/wanted-dogma-free-models-of-inflation/99/#sthash.E7vYhwMQ.dpuf

No comments: