# Granger causality: need for different lag lengths for x and y

I am trying to do a Granger causality test. In the general form Granger causality analysis includes estimating the following equation:

$$y_{t}=a_{1}y_{t-1}+a_{2}y_{t-2}+\dotsc+a_{p}y_{t-p}+b_{1}x_{t-1}+b_{2}x_{t-2}+\dotsc+b_{q}x_{t-q}+e_{t}.$$

It is not necessary that $p=q$, so we can choose different values for $p$ and $q$ (using information criteria such as AIC or BIC). However, EViews forces $p=q$, look here.

Questions:

1. Is it possible to choose different lag values $p\neq q$ in Eviews?
2. Doesn't the lack of the ability to choose a lag contradict the theoretical background of Granger causality?
• Only the latter question doesn't it contradict theoretical background of Granger causality is on-topic here. Purely software-focused questions are off topic. – Richard Hardy Aug 4 '16 at 7:40
• Before closing as off-topic, consider my first comment: the question in italics is on topic. @Dave, there was not supposed to be any link in my comment. – Richard Hardy Aug 4 '16 at 8:19
• I was going through my old answers and noticed this one was not accepted. Do you perhaps need further clarification? – Richard Hardy Feb 20 '17 at 19:03

2. If the appropriate lags for $y$ and $x$ are different, $p \neq q$, forcing a common lag $r$ will be suboptimal.
If $r \geqslant \max(p,q)$, there will be unnecessarily many parameters in the model resulting in increased estimation variance and loss of power in the Granger causality test.
If $r \leqslant \min(p,q)$, there will be omitted variable bias resulting in inconsistent parameter estimates and messing up both the true significance and the power of the Granger causality test.
If $\min(p,q)<r<\max(p,q)$, there will be a combination of both effects.