Sandip Sabherwal makes the point that our WPI inflation is based almost entirely on commodities, and is seriously correlated with the CRB/Reuters index (a global index on commodities).
– In my first iteration I just took the WPI figures as they are over the last 12 years and did a correlation with the Reuters CRB Index on a like to like basis. This correlation did work well however there was a lag impact that was apparent.
-Secondly I did an analysis of the WPI with the Reuters CRB Index taking a 3 month lag i.e. the impact of the movement of global commodity prices is felt in India with a lag of 3 months. This correlation came out to be pretty strong with a correlation of 72%.
Thirdly in order to further refine the study I broke up India’s WPI inflation into two parts where I removed food price inflation out of the WPI as that is largely driven by domestic production and policies. As such I did a correlation of WPI (Manufacturing+Primary ex of food+Fuel) with the Reuters CRB Index on a 3 month lag basis. This analysis shows a more or less perfect correlation as is reflected in the chart given above. The correlation for these series is 81%. The impact is stronger if the Rupee appreciates during the process and has a greater lag if the Rupee depreciates.
I decided to analyze the data myself. I have India’s WPI since 1994 from http://eaindustry.nic.in , and I downloaded the CRB/Reuters Total Returns Index data. Then I worked the CRB data into monthlies, put in a three month lag, divided by three to “harmonize” the series, and plotted the CRB inflation (year on year change) with the WPI.
While it looks similar in many cases, this is a data check and correlation has to be verifiable in the actual underlying data. Does a trend change in CRB inflation mean that the trend of Indian WPI changes?
Using basic excel formulas the correlation comes to 50%, checking the inflation figures for correlation. The CRB data is extremely volatile and one or two months of sharp rises and falls haven’t shown appropriate responses in India. But that’s not the point. The mathematical correlation doesn’t always mean causation.
A sharp variation in CRB hasn’t always resulted in a large change in the Indian WPI inflation figure. In 2010, the CRB index inflation went to near zero, and our inflation remained steadfastly high. In 2006 (another period of interest rate hikes) CRB inflation fell dramatically – the above chart for CRB is lagged three months, remember – yet, India’s WPI only went down slowly, and took nearly a year to come down to 5%. In the 90s there were many instances of divergence (Sandip mentioned on twitter that this was due to to high customs duties).
What this means is: A CRB rise or fall may not result in a dramatic increase or decrease in inflation in India.
The CRB is a dollar measured index so would the numbers change with heavy changes in the dollar rupee equation? Too little data. The rupee has remained in a close range for most of the last decade, but it’s apparent now that the rupee dollar equation impacts inflation quite a bit. The rupee depreciation in 2008 would have resulted in quite a bit of inflation without the CRB change impact.
Finally, he makes this point:
As such given the fact that the Reuters CRB Index has crashed by over 15% in the month of September 2011 the impact of this on the WPI inflation will be seen clearly by December. Inflation will fall much faster than being currently estimated and given the way global commodities have corrected we could be at a 5.5% figure by March 2012.
While I’m not sure of the numbers, I think Sandip is right in that inflation will fall. There are many reasons for it:
We can’t say for sure that the one reason (the CRB index drop) will be solely responsible.
Whatever the case is: the RBI can’t base their decisions on this shaky correlation. They did that the entire of 2010, hoping that the commodity correction in the west would help us. It didn’t. They aren’t going to make that mistake again. Assuming correlation for causation can be a big mistake, especially for policy decisions.