Cyclical Consumption and Expected Returns: New Evidence - How Long Does it Take to Form the Habit in Habit Model?

Atanasov, Møller, and Priestley (2019) find that cyclical consumption at 5-7 year frequency can predict (excess) returns at market level and they argue that low-frequency fluctuations in consumption capture slow-moving counter-cyclical variations in expected returns. Based on cross-sectional evidence, I find that their results are mainly driven by the large-capitalization stocks and cannot be extended to other sorted portfolios. Meanwhile, all firms' returns can be predicted by cyclical consumption at 1-2 year frequency and it suggests cyclical consumption may capture the risk premia at shorter business cycle frequency. I also find cyclical consumption growth is persistent and the persistence increases with time horizon and the future dividend-price ratio can be predicted by cyclical consumption. To rationalize the stylized facts, I modify the Campbell-Cochrane habit model by allowing persistent consumption growth and finite-horizon habit formation. The modified model can reproduce the inverse relation between cyclical consumption and future expected stock returns, consistent with empirical findings.