Using Granger Causality to determine effect of time series on another I understand the lag of one time series on another can be determined using Granger Causality. But is it possible to determine the effect of one time series on another? For example, a 1 point increase in time series $X$ affects $Y$ by $n$%.
 A: You may test for Granger causality once you have built a model for the two time series. The estimated model will give you effect sizes, e.g. the effect of a 1 point increase in $x$ on $y$. (Regarding effect in percentage, a log-level model would give you a direct answer; it works for small changes where logarithmic change approximates percentage change well.)
A: You may be interested in Impulse Response Analysis (http://www.statsmodels.org/dev/vector_ar.html). 
Unfortunately, the statsmodels' package document does not clarify how to interpret the x-axis in the impulse response analysis graphs (check the mid-page in that webpage). However, in stata, I know that the vertical axis in the graphs reflect the estimated percentage point change in the response variable. 
On the contrary, according to this link, "The vertical axis is expressed in units of the Y variable."
So, in sum, the x-axis values in an impulse response graph do indicate the extent to which the response variable changes as a result of changes in X. However, I personally don't know the unit in which the changes are expressed in Python's statsmodels. Based on my observation of a number of such graphs, the range of values the x-axis ticks have got in all of them, and how standard they look, I would guess that the x-axis ticks reflect percentage change in the response variable.
