Long Memory and Asymmetric Uncertainty Effects on Stock Returns and Volatility: A Fractional Integration Approach

Abstract

This paper investigates how Economic Policy Uncertainty (EPU) affects the returns and volatility (proxied by squared returns) of 448 S&P 500 stocks over the period January 2010–December 2020, and whether volatility persistence is related to EPU sensitivity. Persistence is measured with three semiparametric estimators of the degree of fractional integration: the Geweke–Porter-Hudak log-periodogram regression, the local Whittle, and the exact local Whittle. The linear EPU–volatility link does not survive market-level controls, but extreme EPU shocks generate significant volatility responses in roughly one-fifth of the stocks examined, with a slightly greater impact of adverse (bad) news. Long memory in volatility is pervasive. Persistence and EPU sensitivity are negatively related at the standard bandwidth (ρ = −0.182, p = 0.0001), a link that survives controls for size, liquidity, market beta, and sector fixed effects, and that holds for the subsample of stocks whose long memory is statistically significant according to the Qu (2011) test. This association is mainly concentrated at low frequencies, consistently with a low-frequency phenomenon possibly caused by level shifts.