The term structure of liquidity premia and the macroeconomy in Canada: A dynamic latent-factor approach

This study derives the term structure of liquidity premia using the stationary vector-stochastic process for the expectation hypothesis of Campbell and Shiller (1987, 1991). The bivariate VAR model identifies liquidity premia for 12 bond yields that are mean-stationary. A state-space representation for a dynamic factor model captures three well-known latent factors for the term premium curve - level, slope, and curvature. The latent factors are able to aggregate the information from the 12 term premia to accurately identify a term premium curve that increases in maturity on average to 117 basis points. The dynamic factor estimations with macroeconomic variables capture important links from the macroeconomy to the term premium factors that indicate the term premium curve responds to higher inflation and interest rates and to higher expectations of future inflation due to excess demand. Regime-switching estimations for the variances of the latent factors stochastically divide the sample period into high- and low-variance regimes. The dynamic factor models for the two regimes indicate that the term premium curve has steepened by almost 50 basis points in the current low-variance regime relative to that in the previous high-variance regime.