[suj1th]

The short answer is I don't know. However, the convenience of simplifying the calculations is, in my opinion, a win. Again, I am myself not convinced. Probably, my premise on the ubiquitousness of normality assumption on the residuals in ARMA models is itself flawed.

[suj1th]

In a way, I am trying to link it with the central limit theorems. But, I may be woefully lost. I am just a novice in this.