Variance Ratio Tests for Panels With Cross‐Section Dependence

This paper develops panel variance ratio statistics to examine serial dependence in time series with cross‐sectional dependence. We derive asymptotic properties for panels where the cross‐section dimension is fixed or grows with T . Using a factor structure to explain cross‐sectional dependence, we propose a common correlation effects approach to eliminate latent factor influence. By leveraging an increasing N , our new pooled defactorized variance ratio statistic remains consistent over long horizons even when grows at the same rate as T , unlike univariate statistics. We compare its performance to a standardized pooled variance ratio across alternatives with distinct long‐run dependence patterns for both large and finite samples. Applying our tests to size and industry stock return portfolios, we find strong predictability in both short‐ and long‐term horizons, with the common factor primarily shaping serial dependence patterns.