Procedures of Granger causality tests
Granger causality tests mathematically calculate whether the lags of one variable (such
as media diffusion) enter into the equation for another variable (such as democracy)
and refer only to the effects of past values in variable dyads (Enders, 2004). In other
words, Granger causality tests statistically measure ‘‘that if X Granger-causes Y, then
X is a useful predictor of Y, given the other terms in the regression’’ (Stock & Watson,
2003, p. 449). In the sense of Humean causation, it is clear that Granger causality
testing does not intrinsically meet all four of the conditions for causality identified
by Cook and Campbell (1979). As described by McLeod and Tichenor (2003, p. 101)
these conditions are: (a) whether cause and effect covary, (b) whether the cause
precedes the effect, (c) whether the cause and effect do not appear independent of
one another, and (d) all potentially confounding third variables are controlled