Chapter Summary:
In this chapter different type of methods for forecasting are discussed. Basically, these methods are divided into two different types of categories, i.e. conditional (theoretical) and unconditional (a-theoretical).
The first method of forecasting is ARIMA model, which is a-theoretical model. This model is used when there is no other significant predicator is not available. On basics of time series literature many time theoretical model can converted into the a-theoretical models (difference equations). Mostly the ARIMA models are used for high frequency variables (such as exchange rate).
Second method for forecasting is conditional approach. For example, consumption is function of income (for simplicity consider a single equation model). As consumption depends upon level of income, that’s why these model are called conditional model, consequently, these model are based upon proper theory that’s why they are called theoretical models. The benefit of this approach is that a forecaster can generate different forecasts of the target variables conditioned on different values of the predictors.
In many cases, the dependent variable is non-traditional (binary). When the dependent variable is non-traditional then mostly the objective of the analyst it to find probability of happening of specific event verses not happening.
All three methods are decided by the analyst according to their objective, these methods are useful at different times and in different circumstances. An analyst must be familiar with all of these forecasting approaches.