When it comes to predicting inflation, estimates from forecasters and economists at India’s leading firms seem, well, inflated. A Mint analysis of weekly inflation estimates in the last 19 weeks, starting January, by a group of 12 firms that are regularly polled by Bloomberg, showed that in seven out of 19 weeks, a forecaster has been bang on target with its estimate. That translates into just 8.8% out of 180 individual estimates.
Inflation is used as a particularly important political tool in India because roughly 85% of its workforce is in the informal sector, where salaries aren’t tied to cost-of-living increases. Weekly inflation numbers continue to grab the headlines with the ruling United Progressive Alliance government on the defensive when inflation touched a two-year high of 6.6% mark in January this year.
Despite the importance of the indicator, both as a political tool and as an indicator of cost of living increases, getting an accurate handle on inflation remains a challenge in India because of the inaccuracy of price data for the approximately 400 items that make up the wholesale price index and the fluctuations that come with forecasting inflation on a weekly basis.
In nine out of 19 weeks, the median forecast of these firms was within 23 basis points of the actual number released the following day by the ministry of commerce and industry, with just one week (13 January) when the median was the same as the actual weekly inflation number. Some nine times out of the 19 weeks, the median was above the actual figure.
Consider last week. Not a single analyst out of 12 polled by the news service, which provides the data to bankers, traders and the media, estimated the inflation rate correctly for the week of 26 May, when wholesale prices decelerated to 4.85%, from 5.06% the previous week. The median for that week was 5.05 and some analysts missed the inflation estimate by as much as 35 basis points.
The wholesale price index is released by the commerce ministry every week. It initially releases the data as provisional and revises them as final data after a lag of eight weeks.
Of the companies that were polled at least 10 times, Kotak Mahindra got it exactly right or were within seven basis points of the government numbers at least eight out of the 17 times it was polled.
Thomson IFR did not get it right even once. The company was off by an average of 13 basis points in the 19 times it was polled, while ING Vysya Bank, which was polled 13 times, was off by an average 12 basis points. Forecast Singapore Pvt. Ltd and Lehman Brothers, which were polled 17 and 12 times, respectively, were off by 14 and 12 basis points each. As a group, the analysts estimates averaged at least five basis points below the provisional numbers.
While some economists maintain that the repeated error in forecasting is due to a fundamental flaw in the way they, the forecasters, do their arithmetic, the analysts themselves insist the problem lies in the accuracy of the government base data.
“The world has globalized but our analysts and reportage are still in the dark ages,” said Surjit Bhalla, a principal with emerging markets research outfit Oxus Research and Investments. Bhalla is also a member of the National Statistical Commission of India. He does not participate in the poll.
“If you are looking at a bang-on estimate, nobody would get it right,” said Harish Menon, an economist with ING Vysya Bank, which is one of the companies polled regularly by Bloomberg. “When it comes to the wholesale price index, you are looking at a basket of some 400 items. It is impossible to have price information for all those items.”
Menon said analysts typically looked at broad indices such as manufacturing and then used them to extrapolate the direction of price growth. “If somebody predicts inflation at 5.07 and it turns out to be 5.07, then it is purely a fluke.”
A New Delhi-based news executive of Bloomberg News declined to comment when asked about the usefulness of such surveys to its customers. The analysts said their estimates are to be used more to identify the broad trend than the exact numbers.
“These are short-term things. Sometimes, you have an instinct and you make assumptions and sometimes they don’t bear out,” said one of the analysts, Soumitra Choudhury, an economic advisor and research co-ordinator with ratings firm Icra Ltd. Another problem is the difficulty in verifying government data on prices of the basket of goods that are used to calculate inflation, he said. Icra got it wrong 16 out of 17 times it was polled.
Leslie Khoo, an economist with Forecast Singapore, said one of the major constraints with forecasting inflation in India was the lack of suitable econometric models that would take into account weekly fluctuations in prices. “For India, sometimes I get close, sometimes I go way off,” conceded Khoo. “In Taiwan, for which I do this as well, it is quite all right, because their estimates are on a monthly basis. Some of the weekly fluctuations even out.”
“In order to project inflation year-on-year, you need to take into account seasonal factors such as harvest-linked supply and demand and seasonal variations in the prices of inputs such as oil. There are a lot of people who do not take that seasonality into account. In the US, most indices use seasonally adjusted data,” Bhalla said.