User’s Guide : Basic Data Analysis : Groups : Granger Causality
  
Granger Causality
Correlation does not necessarily imply causation in any meaningful sense of that word. The econometric graveyard is full of magnificent correlations, which are simply spurious or meaningless. Interesting examples include a positive correlation between teachers’ salaries and the consumption of alcohol and a superb positive correlation between the death rate in the UK and the proportion of marriages solemnized in the Church of England. Economists debate correlations which are less obviously meaningless.
The Granger (1969) approach to the question of whether causes is to see how much of the current can be explained by past values of and then to see whether adding lagged values of can improve the explanation. is said to be Granger-caused by if helps in the prediction of , or equivalently if the coefficients on the lagged ’s are statistically significant. Note that two-way causation is frequently the case; Granger causes and Granger causes .
It is important to note that the statement “ Granger causes ” does not imply that is the effect or the result of . Granger causality measures precedence and information content but does not by itself indicate causality in the more common use of the term.
When you select the Granger Causality view, you will first see a dialog box asking for the number of lags to use in the test regressions. In general, it is better to use more rather than fewer lags, since the theory is couched in terms of the relevance of all past information. You should pick a lag length, , that corresponds to reasonable beliefs about the longest time over which one of the variables could help predict the other.
EViews runs bivariate regressions of the form:
(12.44)
for all possible pairs of series in the group. The reported F-statistics are the Wald statistics for the joint hypothesis:
(12.45)
for each equation. The null hypothesis is that does not Granger-cause in the first regression and that does not Granger-cause in the second regression.
We illustrate using data on consumption and GDP using the data in the workfile “Chow_var.WF1”. The test results are given by:
For this example, we cannot reject the hypothesis that GDP does not Granger cause CS but we do reject the hypothesis that CS does not Granger cause GDP. Therefore it appears that Granger causality runs one-way from CS to GDP and not the other way.
If you want to run Granger causality tests with other exogenous variables (e.g. seasonal dummy variables or linear trends) or if you want to carry out likelihood ratio (LR) tests, run the test regressions directly using equation objects.
Panel causality tests are described in “Panel Causality Testing”.