Adaptive Expectations and Commodity Risk Premiums

We analyse a novel time series of investors’ expectations on future commodity spot prices, and show that a model with adaptive learning can replicate investors' forecasts. We use this framework to back out the dynamics of the (ex-ante) risk premia for different commodities and maturities, and provide evidence that commodity risk premia are time-varying and their dynamics is predominantly due to the changing nature of risk sharing and appetite, as proxied by open interest, hedging pressure and time-series momentum. Finally, we show that the explanatory power of alternative factors is not constant over time, both across commodities and time horizons.