Movements in expected returns (ER) can cause a bias in measured autocorrelations, and the resulting spurious component is positive for infrequent regime shifts. We demonstrate this point analytically and investigate its empirical prevalence. In a key contribution, we use shifts in ex ante ER...