I'm working with bond data and I want to get standardised residuals to conduct a copula analysis.

The problem is that often the prices, for consecutive days, are the same and this fact makes the log return equal to zero. When I study the autocorrelation in the squared returns, then, I find that those are not autocorrelated.

I Know that filtering the data using an ARMA-GARCH process model, I can obtain the desired standardised residuals, but, as in my case, the squared returns are not correlated and this model results unfeasible.

What can I do in this case?

  • $\begingroup$ If the squared return are not correlated and if the time series plot does not exhibit any volatility clustering then you probably don't have any ARCH effects in your residuals and hence, you do not need to fit an ARCH/GARCH model to model them. To be sure however conduct a test for ARCH effects! $\endgroup$ – Plissken Aug 5 '15 at 10:09
  • $\begingroup$ Here's a reference for the ARCH LM test: se.mathworks.com/help/econ/engles-arch-test.html $\endgroup$ – Plissken Aug 5 '15 at 10:11
  • $\begingroup$ Thanks Dan! I really appreciate your answer... But if neither the log returns nor the squared one are correlated, is it possible the returns are white noise? Even if they are not normal? $\endgroup$ – sciury Aug 6 '15 at 15:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.