CAN ECONOMIC DATA BE USED FOR EQUITY MARKET TIMING? MANUFACTURING ISM CALLING FOR A REVERSAL IN STOCKS ONE-YEAR DOWN

Most economic data is useless for the purpose of predicting markets. Asset prices discount the future economic outlook and are often used as leading indicators. Forecasting financial instrument returns therefore requires predicting the course of leading indicators – basically two steps ahead of the economy.

Yet, leading indicators (excluding asset prices themselves) could potentially be used to differentiate between a “bear market rally”/“false dawns” from a sustainable turn in the outlook. The big question is the following: if there is a divergence between stock returns and the underlying economic outlook defined by a leading indicator, do fundamentals prevail?

I used the manufacturing ISM as a proxy for the future economic outlook. The S&P composite monthly returns are positively correlated with coincident changes in the ISM. Also, the monthly divergence indicator does not add any value to intra-year investment decision-making. However, not all is lost.

The level of the ISM predicts 1-year forward S&P performance. The relationship is inverse and statistically significant. The current level of the ISM suggests a good 2009 for stocks.

ISM AND S&P 500 ARE COINCIDENT

Monthly equity returns covary positively with the change in ISM.

S&P Comp. = 0.60 + 0.15*ISM

Data: Monthly % Changes from January 1948 to December 2008
R^2 = 5%
t-statistic = 6.8

While this information is obvious and good to know, it does not help us too much in our market timing endeavor.

ISM AND S&P COMPOSITE DIVERGENCE INDICATOR NOT USEFUL

To test my divergence hypothesis, I conducted the following experiments:

1. If monthly return on the S&P is positive but the 3-month average (to filter out noise) change in the ISM is negative, I short the index.

Average monthly return: -0.72
Standard deviation: 3.32
No. of trades: 209

2. If monthly return on the S&P is negative but the 3-month average change in the ISM is positive, I go long the index.

Average monthly return: -0.13
Standard deviation: 3.71
No. of trades: 124

The results are disastrous and suggest that either there is momentum in monthly returns or that stocks predict the economy better than the ISM. The lead-lag structure of correlations indicates that S&P composite’s variations predict the future movements in the purchasing manager’s index.

Correlation (S&P (t), ISM (t)) = 0.24
Correlation (S&P (t-1), ISM (t)) = 0.19
Correlation (S&P (t), ISM (t-1)) = 0.02

While these results reinforce my belief in the futility of economic indicators in devising profitable trading strategies, all is not lost.

YEARLY RESULTS PRESENT A DIFFERENT PICTURE

There exists an inverse relationship between the level of the ISM and 1-year forward returns. The reason for this covariance, in my view, is that the current in point the cycle has implications for the likely evolution of the future economic landscape.

Medium to long-term, the invisible hand operates to correct deviations from equilibrium. For instance, high oil prices set into motion various forces to reverse the trend. At steeper valuations, new sources supply such as tar sands become viable. Certain alternative fuels also gain traction. Elasticity of demand also rises with time as people adjust behavior and switch to processes which economize on fuel. In addition, governments reinforce these processes by actual policy or pronouncements which affect expectations of economic agents.

Similarly, a high level of the purchasing manager’s index indicates a mature business cycle with incipient inflationary pressures. In a money targeting regime, rising money demand forces interest rates to rise, working to moderate the cycle. At economic extremes, public outcry makes the authorities more responsive - Central Banks reinforce market forces to increase the speed of reversion to equilibrium.

S&P = 40.63 – 0.60* ISM (t-1)

Data: Yearly from 1948 to 2008, ISM: levels; S&P Comp.: Changes
R^2 = 9%
t-statistic (beta) = -2.51

The coefficient of ISM is statistically significant at the 1% level. The latest manufacturing PMI reading indicates the possibility of solid returns (about 19%) over the next year.

As a next step, I tested the strategy of buying the index when ISM fell below 45. I only considered year-ends to get non-overlapping time intervals.

Mean: 19%
Standard deviation: 15%
Maximum drawdown: -14%
Maximum: 41%
Number of trades: 10

Importantly, 9 out 10 trades made money. The strategy of picking the bottom lost money in 2001. I also tested whether these results represent a chance outcome devoid of any economic significance. A t-test for the difference in mean reward of a buy-hold strategy and the one presented above point to superiority of the PMI based trading scheme and we can assert this with 95% statistical confidence.

THERE ARE ALWAYS CAVEATS; IS THIS TIME DIFFERENT?

Understanding of why the rule worked in the past can also highlight circumstances where it might fail. The only failure occurred in 2001. September 11 was an exogenous shock that severely dented confidence. Additionally, in early part of the decade, equities sported stratospheric valuations. Stocks are clearly no longer exuberantly priced.

The second caveat is that the equilibrating forces I mentioned earlier, fail to operate as hypothesized or there is a longer lag than usual. Indeed, there is reason to believe that this scenario could be relevant today. The transmission mechanism of monetary policy has been rendered dysfunctional by the financial sector crisis.

Yet, given that the authorities are willing to do everything possible to avert debt deflation, I think it is imprudent to bet for a significant decline from current levels. Not that another downdraft is not possible, just that it requires another Great Depression-like outcome. I would bet against that low probability scenario.

Data Source:
http://www.irrationalexuberance.com/
http://www.ism.ws/