Changing Expected Returns Can Induce Spurious Components in Autocorrelations

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 estimates from options prices, factor models, and analysts’ forecasts to investigate our premise. As predicted, absolute shifts in ER are indeed strongly and positively related to observed autocorrelations. We also show that shifting ER implies biased cross-autocorrelation, and find supporting evidence for this phenomenon as well