A decomposition of the predictive content of the term structure for output growth in Canada

Abstract

The purpose of this study is to identify the underlying economic disturbances that drive the predictive content of the term structure for future output growth and those that may distort its information content. The study uses a structural vector autoregressive (VAR) model of a small and open economy for Canada that takes into account its relationship with financial markets in the USA and that Canada is a relatively large exporter of commodities. The model is used to decompose the sources of the variation of the slope of the yield curve and the correlation between the term spread and output growth. Monetary policy disturbances in both Canada and the USA, as well as short-term interest rates, are found to trigger excessive volatility in short-term rates and the term spread that do not contribute to the predictive content of the term spread for future output growth at horizons relevant for monetary policy analysis. However, innovations in output growth, inflation and other macroeconomic variables do not distort the forecast power of the term spread. Unlike the evidence for the USA, disturbances in nominal long-term yields are found to contribute about the same amount to the predictive content of the term spread as unexpected movements in monetary policy.

Notes

¹ See Hassapis (2003), Atta-Mensah and Tkacz (2001), Harvey (1997) and Cozier and Tkacz (1994) for Canada; and Hamilton and Kim (2002), Galbraith and Tkacz (2000), Estrella and Mishkin (1997, 1998), Estrella and Hardouvelis (1991), among others, for the United States and G-7 countries.

² See Lange (2005) for a more detailed presentation of the VAR model and discussion of the impulse responses to the shocks from the model.

³ Cushman and Zha (1997) also apply block exogeneity to US and foreign variables in order to identify an explicit monetary policy function in a structural VAR model for a small-open economy like Canada. Block exogeneity has also been applied by Racette and Raynauld (1992) and Lange (2005) for Canada, and by Genberg et al . (1987) for Switzerland. Evans and Marshall (2001) and Wu (2003) use block exogeneity to restrict bond yields from affecting macroeconomic variables.

⁴ The null hypothesis of the block exogeneity test is that the lags of the variables in the endogenous block do not enter the equations for the exogenous variables.

⁵ See Lange (2005) for a detailed description of the data.

⁶ The over-identifying restrictions on the contemporaneous coefficient matrix relative to a just identified model could not be rejected at the 95% confidence level.

⁷ The construction of the impulse responses of the term spread differs from Smets and Tsatsaronis (1997) in that the response function for the term spread is calculated as the difference between the impulse responses of the long- and short-term rates, and not the response of the term spread itself. The forecast-error variances in Table 1 below indicate that the two specifications are quite similar, but not identical.

⁸ The innovations to the real exchange rate are quite small and excluded from the tables.

⁹ In the bottom panel, the column for the short rate is equal to the sum of the term spread and the long-term yield.

¹⁰ The unconditional covariance between the current term spread and future output growth at the various forecast horizons is calculated as the accumulated covariance over 100 quarters.