We noted in the Introduction that forecasting is an important part of econometric analysis, for some people probably the most important. How do we forecast economic variables, such as GDP, inflation, exchange rates, stock prices, unemployment rates, and myriad other economic variables? In this chapter we discuss two methods of forecasting that have become quite popular: (1) autoregressive integrated moving average (ARIMA), popularly known as the Box-Jenkins methodology,1 and (2) vector autoregression (VAR).
In this chapter we also discuss the special problems involved in forecasting prices of financial assets, such as stock prices and exchange rates. These asset prices are characterized by the phenomenon known as volatility clustering, that is, periods in which they exhibit wide swings for an extended time period followed by a period of comparative tranquility. One only has to look at the Dow Jones Index in the recent past. The so-called autoregressive conditional heteroscedasticity (ARCH) or generalized autoregressive conditional heteroscedasticity (GARCH) models can capture such volatility clustering.
The topic of economic forecasting is vast, and specialized books have been written on this subject. Our objective in this chapter is to give the reader just a glimpse of this subject. The interested reader may consult the references for further study. Fortunately, most modern econometric packages have user-friendly introductions to several techniques discussed in this chapter.
'G. P. E. Box and G. M. Jenkins, Time Series Analysis: Forecasting and Control, revised ed., Holden Day, San Francisco, 1978.
836 PART FOUR: SIMULTANEOUS-EQUATION MODELS
The linkage between this chapter and the previous chapter is that the forecasting methods discussed below assume that the underlying time series are stationary or they can be made stationary with appropriate transformations. As we progress through this chapter, you will see the use of the several concepts that we introduced in the last chapter.
Was this article helpful?