We use a comprehensive database of inter-dealer quotes to conduct the first empirical analysis of the dynamics of the swaption cube. Using a model independent approach, we establish a set of stylized facts regarding the cross-sectional and time-series variation of conditional volatility and skewness of the swap rate distributions implied by the swaption cube. We then develop and estimate a dynamic term structure model that is consistent with these stylized facts, and use it to infer volatility and skewness of the risk-neutral and physical swap rate distributions. Finally, we investigate the fundamental drivers of these distributions. In particular, we find that volatility, volatility risk premia, skewness, and skewness risk premia are significantly related to the characteristics of agents' belief distributions for the macroeconomy, with GDP beliefs the most important factor in the USD market, and inflation beliefs the most important factor in the EUR market. This is consistent with differences in monetary policy objectives in the two markets.
