The stock market eagerly anticipates the release of various macro-economic data and associated policy announcements. The annual Budget and Economic Survey are of course, most important. But other numbers such as the Wholesale Price Index (WPI) and the Consumer Price Index (CPI) and the Index of Industrial production (IIP) also have an impact. So does the RBI credit policy reviews, and occasional updates on the trade gap, sovereign credit ratings, etc.
The market reacts quite strongly to some of these figures and there is higher price volatility on the days when data are released. Analysts use these lagged numbers to make forward projections. Everybody in the market follows inflation data, and the IIP.
Undoubtedly, there is short-term correlation with the newsflow on days when data are released. This can offer 2-3 session trading opportunities. More interestingly, is there a long-term correlation?
Logically, long-term correlation, or inverse correlation in the case of WPI, should exist. Macro-economic data offers valuable clues as to overall business cycle trends. Also, in theory, macro-economic data has more predictable trends than stock prices. So if a trader or an investor can work out correlations, he or she may be able to develop trading/ investment guidelines.
Let's take a deliberately non-mathematical look at the Index of Industrial Production (IIP) and the Wholesale Price Index (WPI), which are the two most widely followed series. Both IIP and WPI have data-gathering problems. In both series, a preliminary number is released and later revised. The revisions can be large and aren't well publicised. The prelim number is high-profile.
In the WPI, price changes are often incorporated late. In the IIP too, manufacturing unit numbers may be left unchanged for months. There may be methodology issues as well. These are both weighted baskets of components. The weights could be misleading in terms of actual importance to inflation, or the influence on industrial activity. There are a couple of other issues with presentation. The market looks at the point-to-point (P2P) change over the levels of a year ago. This P2P method is easy to compute and understand. It deseasonalises data, since winter months are compared to winter months, etc.
But P2P produces odd results if the base month had unusually high or low levels. It can introduce distortions and often masks trends. One typical issue is with Diwali and other festivals, for instance. Hindus and Muslims use lunar calendars, which means major festivals shift in solar terms. Festivals cause blips on both WPI and IIP. If Diwali falls in different months of two successive years, as often happens, the P2P numbers show funny results.
As an illustration look at Chart 1, WPI Vs P2P, where the WPI has been mapped (left axis) against its own P2P changes (right axis) since April 2006. The P2P obviously fluctuates much more, and it has been misleading at times.
In many periods, the P2P started trending down (suggesting lower inflation) while the WPI has moved up with a steady slope. If one follows only the P2P line, inflation started reducing in June 2010! It's only in August 2012 that the WPI trendline has started correcting (moving sideways rather than up). The nearly linear slope of the WPI for years at a time makes it highly predictable and it shows the RBI has managed to contain inflation though it hasn't reduced it.
There are many better ways to track trends in time-series like the WPI or IIP. But those are a little more complex, even if more accurate. For example, moving average systems, or regression analysis, presents smoother, more accurate views of lasting trends.
The bulk of market players don't bother and you have to take note of the P2P change numbers because that will influence sentiment. But if you compare with other tools, you will find instances where interpretation or action based on P2P would be wrong.
Both WPI and IIP are composites of several broad groups, and include many individual items with different weights. Specific segments of IIP and WPI components have relevance to different industries. It is possible to discover relationships, say between cement prices and the construction industry, even if there is not much correlation of the broad stockmarket index versus the overall macro-economic series.
Comparing the WPI series to the Sensex, runs into a problem of historicity. The current WPI extends back to April 2005. Since 2005, the stock market has gone through two complete cycles of bull-bear markets and is now into the bullish stage of a third cycle.
But the WPI has fallen only for two months during the global recession of 2008 and 2009. At most times, inflation has been high. Sometimes it has been moderate or low, but negative only for two months out of 92. We don't know what happens to stock prices during an extended period, when inflation is low or downtrending. (Earlier evidence from 2003-05 suggests the market goes up).
There is some correlation and some inverse correlation between changes in inflation (Chart 2, Sensex Vs WPI-P2P) and the Sensex but that is also weak. The best one say is that during this 7-year period, there were some periods when inflation and the Sensex moved in opposite directions, and some periods when they moved in the same direction.
The IIP is more interesting. Since IIP tracks business cycles, it has also seen several periods of negative movements like the stockmarket and it has a volatile profile.
Take a look at Chart 3 (Sensex Vs IIP), where Sensex values have been superimposed across the IIP time-series. These series show long periods of definite positive correlation between the Sensex and the IIP. The stockmarket usually leads the IIP, sometimes by several months. This is true both in uptrends, as well as in downtrends. Specific sectors like capital goods manufacturers and power utilities are also more sensitive to changes in the IIP.
The correlation and the lead-lag factors are both more obvious if one uses smoothed values. Chart 4, Sensex Vs IIP Smoothed, shows the Sensex's 6-month moving average mapped versus the IIP 6-month moving average. Periods of positive correlation when the Sensex and the IIP have moved in the same direction are marked. So is the lead-lag factor. The IIP is therefore, usually a lagging indicator and a good one. It confirms if the market is trending in a sustainable fashion or not, as the case may be. If there are divergences when the market is trending up, and the IIP values are negative or flat, one has to be watchful and wait for the IIP to move up, or shareprices to move down.
A look at the charts will show this sort of inverse correlation has occurred now for several months. For greater detail, check Chart 5, Sensex VS IIP P2P, where trends have been smoothed out, and the leads-lags are marked and numbered. Interestingly, this says, IIP started falling in July 2010 and fell until September 2012, when the trend flattened out. (Nominal P2P change in the IIP was negative in November 2012 versus November 2011 due to the Diwali effect). The aberration where the IIP dropped in July 2010 while the Sensex continued to rise till December 2010 was caused by excessive liquidity.
The Sensex has trended up since August 2012, whereas the IIP has been flat or rangebound between August-November 2012 (our last IIP data is November 2012). That's a 4-month long divergence.
Share market gains through December and January (The Sensex has gained about 2.5 per cent since November 30,2012) suggests economic activity picked up in those months.
If the IIP doesn't jump in December or January, the market could correct down. This is another way to say that the market is betting on industrial activity has started to pick up and if this hope is not confirmed soon, disappointed bulls may turn bearish.
This very preliminary look at the data suggests that the Sensex does have useful positive correlation with the IIP. There are plenty of other ways to test IIP data and also WPI data for potential predictability and possible market signals.
I've deliberately avoided statistical detail because the analyst must understand what he's looking for before he hunts for the appropriate statistical tools. Otherwise it is easy to run batteries of tests and end up simply with a confused mass of numbers that don't offer much insight.