Speaker
Description
This paper examines whether the predictability of WTI crude oil futures returns is a regime dependent feature of the market. Building on the theory of storage and risk premium explanations of commodity returns, we test whether a simple model using basis changes and realized volatility can deliver economically meaningful out of sample forecasts and whether its performance is concentrated in high volatility states.
The theoretical framework combines two channels. Changes in the spot–futures basis proxy shifts in inventories, scarcity and convenience yield, which should translate into subsequent futures returns. Realized volatility captures periods of heightened uncertainty and risk aversion, when pricing relations become tighter and risk premia more pronounced. These arguments imply that basis and volatility should be especially informative in turbulent markets, generating regime dependent predictability.
Empirically, we construct a monthly data set for WTI front month futures and spot prices over 1985–2024, resulting in 471 observations after cleaning and transformations. Futures returns are computed from front month settlement prices, while the predictors are the first difference of the spot–futures basis and 12 month realized volatility. The baseline model is a linear OLS regression with robust standard errors, estimated in a recursive expanding window scheme to mimic real time forecasting and avoid look ahead bias. Volatility regimes are defined exogenously by terciles of realized volatility.
Preliminary results show that, relative to a historical mean benchmark, the two variable model attains an out of sample R2R2 of about 40%, with a substantial reduction in mean squared forecast error and statistically significant Clark–West test statistics. This forecasting power is strongly regime dependent: in low volatility states, out of sample R2R2 is modest, while in high volatility regimes it exceeds 50%. These findings suggest that WTI futures returns are only weakly predictable in tranquil markets but become strongly predictable when volatility is elevated, highlighting the central role of regime dependent dynamics in crude oil futures pricing.
