Fast Forward Forecasting

Mark Masuoka is a successful money manager based in Newport Beach, California. His firm, Fast Forward Forecasting is best known for using superior macro-economic forecasting to time the stock market. Currently, he is bearish on the stock market and is invested 50% invested in the S&P and 50% invested in T-bills and high-grade commercial paper. He is concerned that the U.S. market is overvalued at current levels, but he does not want to be out of the market entirely.

Mark  has read  a lot about the predictability of stock market returns using dividend yields, past returns and interest rate variables.   He is aware of claims that dividend yields are a strong indicator of whether the market is over-valued, however he has no specific quantitative model that measures this relationship. Mark hired Dick Cassama as an intern for the summer. He gave him what he believed to be a straightforward task: to test whether  any variables such as dividend yields or deviations from the six-month moving average of the treassury bill rate or even past long-horizon returns have predicted stock market movements in the past and if they might do so in the future.    Masuoka not only wanted some statistical evidence -- he wanted some clear rule of thumb that might help him manage money  -- thuis required the identification of key thresholds that should indicate when to move out of stocks into T-bills or bonds. Masuoka also asked Dick to provide some estimates of the reliability of the prediction model, and some measure of how much better a client will perform if they use his management services.

 Dick decided to use historical data, and some statistical methods in order to address Masuoka's request. I He decided to build his own tactical asset allocation model based on past data.
 

Assignment

Prepare an analysis of dividend yields  as predictors of the S&P 500.  Perform the test through 1990 data aas well as through 2002 data. In addition,  consider  one other tactical asset allocation model, based upon either past returns or some interest rate variables that appear to have been good predictors  in the literature on forecasting long-horizon returns.  In your analysis  you should address the following issues:
 
 

1) How significant is the prediction power for short-horizon returns (one-year) and for long-horizon i.e. multiple-year returns?

 2) How would you exploit this strategy to make money?

 3) Is the effect confined to any specific time periods? Is it only useful when yields are high or low?

 4) How can you reasonably estimate trading profits?

 5) What is an appropriate benchmark against which to measure the portfolio performance? How can you describe the systematic risk of the strategy?

 6) How can you use your model in a mean-variance framework to improve the inputs?

 7) Based upon your model, what do you expect for today's market? How reliable is your prediction?