|•||While not useful in predicting future inflation, the slope of the yield curve is useful in predicting changes in interest rates.|
|•||Commodity prices (i.e., gold and oil) do not appear to contribute any additional information to that contained in the other variables included in the analysis.|
|•||Base growth is useful in forecasting interest rates.|
|•||Excess base growth is useful in forecasting interest rates and inflation, and vice versa.|
The result of a model that combines financial prices, base growth, and excess base growth does quite well in explaining the inflation rate and changes in the three-month T-bill (Figures 5.2 and 5.3). Given that the empirical specification of the model uses only quarterly values of financial prices and that monetary aggregates lagged from one to five quarters, the estimated equation may be used to forecast one step ahead (i.e., one quarter) changes in interest rates and inflation. For the sample considered, the model correctly forecasts the direction of change in interest rates approximately 75 percent of the time. The projections for 1989 are equally encouraging. 11
During the first quarter of 1989, the three-month T-bill rose 75 basis points. The model projected a 30 basis point rise. In the second quarter the model projected a 116 basis point decline. The three-month T-bill declined 67 basis points. For the third quarter, the model projects a 96 basis point decline. As of September 1, the three-month T-bill had declined 29 basis points.
The inflation forecasts are just as encouraging. During the first and second quarters, the model projected a 5.7 percent inflation rate which was close to the realized inflation rate. For the third quarter the model projects a decline in the inflation rate to 4.5 percent.
The formulation of the empirical analysis assumes that expectations regarding the different endogenous variables are formed rationally in the sense that, given the current available information, participants in each market use the structure of the economy to form optimal forecasts. Furthermore, actions based on these forecasts do, in fact, generate the economic structure. Our approach to estimating the relationship is to estimate a vector autoregression model of the series in question. The vector autoregression technique is based on the work of Sims and consists of a system of equations, one for each of the variables included in the analysis. 12
Each of the equations contains the same set of explanatory variables. Originally, the empirical investigation included real GNP growth, the budget deficit as a percent of GNP, the three-month T-bill, the slope of the yield curve, the inflation rate, the foreign exchange value of the dollar, the S&P 500, the price of oil, the price of gold, M1, and base and excess base growth. Variables that did not