Hello,

I am working on a time series project and I try to verify the Permanent Income Hypothesis with quarterly US data.

Therefore I regress first difference of consumption on first difference on income. If I predict the residuals and test them for serial correlation it indicates that I have positive serial correlation.
I have tried to include more lags in the regression but this doesn't solve the autocorrelation issue in the error term.
Now I have read something about the Newey West Standard Error and I calculated the lags using the formula from Stock and Watson m=0.75*Obs^1/3.
However if I predict graph the residuals (and additionally test them) there is still higher order serial correlation in the error term.


[newey c1 l2.y1 l3.y1 l4.y1, lag(3)

predict uhat, resid
reg uhat l.uhat l2.uhat l3.uhat, r
test l.uhat l2.uhat l3.uhat

( 1) L.uhat3 = 0
( 2) L2.uhat3 = 0
( 3) L3.uhat3 = 0
( 4) L4.uhat3 = 0
( 5) L5.uhat3 = 0

F( 5, 104) = 11.79
Prob > F = 0.0000



tsline uhat]

Does anybody know how I should proceed?
I really would appreciate some help, thank you!