Hi,

I know that wild bootstrap make tests robust to heteroscedasticity, and sometimes also to serial correlation (block wild bootstrap and a dependent wild bootstrap). Is the Small-sample Johansen coint. rank tests with wild bootstrap also robust to serial correlation? 

Hi Lars, co-author of the johansensmall package here. The simple answer is: no, this is a "plain" wild bootstrap. The longer-term plan is to re-use resampling code that allows also other variants. That more general code partly lives in the "SVAR" addon and/or in the "extra" addon. But this hasn't happened yet.

Apart from that, the best solution of course is to capture the serial correlation in the VAR model itself. If this doesn't work (and I know it often doesn't in practice), then I wouldn't be very confident about healing the problem with a certain bootstrap variant, to be honest. But I admit this depends on the context and application.

cheers

sven