Unit Root Testing in the Presence of Mean Reverting Jumps: Evidence From US T-Bond Yields

Abstract

Mean reversion of financial data, especially interest rates is often tested by linear unit root tests. However, there are times where linear unit root test results can be misleading especially when mean reverting jump formations are at stage. Considering this framework, we provide a new unit root testing methodology and compute its asymptotic critical values via Monte Carlo simulation. Moreover, we numerically compare the power of this generalized mean reversion test with the pioneering linear unit root test in the literature namely the Augmented Dickey Fuller (ADF) test. We deduce that our test is a refinement of ADF test with a higher power. Weapply our findings to US 10-year Treasury bond yields. We aim to shed light to the discussion among researchers whether interest rates can sometimes revert to a long-term constant mean or not from an unorthodox point of view.