my ARDL model has heteroskedasticity

[Varaque Sabounjian]

hello all,

basically I am trying to find the presence of cointegration between German banks CDS score and German Sovereign CDS score so that I can look who is leading the price discovery.

My data starts in 2007 and ends in 2021 and its daily.

the following are the commands I used:

gen logGermanbanksCDSscore = log( GermanbanksCDSscore)
gen logGermany_Soveriegn = log( Germany_Soveriegn)

gen ChangelogGermanbanksCDSscore = logGermanbanksCDSscore - logGermanbanksCDSscore[_n-1]
gen ChangelogGermany_Soveriegn = logGermany_Soveriegn- logGermany_Soveriegn[_n-1]

varsoc ChangelogGermanbanksCDSscore
varsoc ChangelogGermany_Soveriegn

ardl ChangelogGermanbanksCDSscore ChangelogGermany_Soveriegn, lags(4,4) aic

ardl ChangelogGermanbanksCDSscore ChangelogGermany_Soveriegn, lags(4,4) ec btest

ardl ChangelogGermanbanksCDSscore ChangelogGermany_Soveriegn, lags(4,4) regstore(ecreg)

estimates restore ecreg

regress

estat dwatson

estat imtest, white

cusum ChangelogGermanbanksCDSscore ChangelogGermany_Soveriegn

[Varaque Sabounjian]

So as you can see in the btest my F statistics 498 which is very high but its more than I(1) so there is cointergration. The white test is indicating that there is heteroskedasticity. And the cusum test is not working. Can someone please help point what I am doing wrong or how to fix it ?