The path coefficient of, for example, the 1st order path from region A to B in a Granger causality model is the amount of change in region B when region A changes one unit, just like the meaning of a coefficient in a regression model. So a negative path of, for example, -0.5 indicates region B decreases by 0.5 when region A increases one unit.
Gang