Comparing Multi-State Kalman Filter and ARIMA forecasts: an application to the money multiplier

This paper derives one-month ahead forecasts of the money (M I) multiplier using the Multi-State Kalman Filter and Box-Jenkins ARIMA methods. A comparison of the forecasts far the period 1980-82 reveals that the Multi-State Kalman Filter procedure was generally superior to the ARIMA procedure In terms of most summary statistics. The superiority is traced to the turbulent period of 1980-81. This paper also compares aggregate and component forecasts of the multiplier. The aggregate Multi-State Kalman Filter was the most accurate in predicting the one-month ahead money multiplier.