RDP 8910: An Analysis of the Determinants of Imports 5. Methodology

The estimation procedure involves a number of steps.

Firstly, all time series used in the estimation process are tested for stationarity[10]. They are each tested for a maximum of three unit roots down, using three different tests; the Dickey-Pantula, Stock-Watson, and Dickey-Fuller tests.

The next step in the estimation procedure involves estimating the long run, or cointegrating, relationships between the variables, noting the properties of the data suggested above. Thus the alternative measures of imports are regressed on each of the proposed demand variables and the price variables. An augmented Dickey-Fuller test (ADF) is used to test for stationarity of the residuals.

Finally, the full error correction models (ECM) are estimated. Initially four lags of each of the differenced explanators are included, and then the insignificant explanators are dropped from the equation to leave a preferred equation.

Footnote

Rob Trevor's “Unitroot” procedure was used to do these tests. [10]