Time Series Forecasting: ARMAX by Thomas H. Lincoln
As an avid reader of stock market books and newspapers, I have observed the overwhelming number
of graphic illustrations of economic and market indicators. Oftentimes some indicators are contradicted
by others and leave the future direction of the market probable, but not always quantitatively definite.
Determined to make a useful forecast from the many graphs and indicators, I have combined a multiple
regression model of several economic indicators and related market indicators with the autoregressive
moving average (ARMA) models of Box-Jenkins to forecast the Standard & Poor's 500 cash index six
months ahead. The final model, often called ARMAX (autoregressive moving average exogenous variables
model), provides a fundamental and technical time series approach to market forecasting.
There are clear and definite steps involved in building an ARMAX forecasting model, which we will
discuss here, and I can also provide a sample model with good regression statistics and good forecasting
statistics. The economic and stock data were retrieved from the two public domain Federal Reserve
bulletin boards. The indicators I used are noted in Figure 1.