# How to correct an error correction model

I have found ARCH effects at the 10% significance level.

What could I do to correct this?

• There is no formula for correcting misspecification in a regression. A 10% significance level can be a statistical fluke that one could ignore or it could be a sign of something else. The ARCH test implies that the squares of the lagged residuals are correlated with the contemporaneous residuals. This can be occur if there is additional unmodelled dependence, nonlinearities, time varying heteroskedasticity, outliers or for many number of other reasons. – Andrew M Apr 24 '20 at 15:03