Nominal term structure and term premia: evidence from Chile

ABSTRACT

The downwards trend exhibited in Chile’s nominal term structure since 2003 has been a common pattern shared by other developed and developing economies. To understand the behaviour of the nominal yield curve in Chile, we rely on an affine dynamic term structure model which allows the term structure to decompose into the expected short-term interest rate (related to the monetary policy expectation) and the term premium. We show that most of the fall of long-term interest rates as well as its dynamics are related to the term premium rather than the expected short-term interest rate. Moreover, we find evidence that term premium is driven primarily by the US term premium and domestic nominal uncertainty derived from expected inflation.

KEYWORDS:

• Affine models
• term structure
• term premium
• uncertainty

JEL CLASSIFICATION:

• E43
• E44
• G12
• G15

Acknowledgments

The authors are thankful to seminar participants at the Central Bank of Chile for useful comments and suggestions. The opinions and remaining mistakes are of exclusive responsibility of the authors and do not necessarily represent the opinion of the Central Bank of Chile or its Board.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 This section is based on Gürkaynak and Wright (2012).

2 In general, the R² is quite moderated.

3 Ceballos (2014) presents the main empirical facts for the Chilean nominal term structure.

4 Our measure of core inflation is the CPI excluding energy and food. The 236 products included in this index total a combined 72.29% of the total weight of the CPI.

5 Computed as the cyclical component of the IMACEC derived from a Hodrick–Prescott filter with smoothing parameter of 14 400.

6 Ceballos and Romero (forthcoming) report that these components are relevant in the determination of output growth and inflation.

7 This section uses standard notation in the literature, specifically based on Bauer, Rudebusch and Wu (2012).

8 This result is standard in the literature. See the appendix in Ang and Piazzesi (2003) for a proof.

9 In fact, Diebold and Li (2006) find that macroeconomic variables are relevant to forecast the yield factors (defined as level, slope and curvature).

10 In Diebold, Piazzesi and Rudebusch (2005), the authors refer to different risk factors used in recent literature as well as the link between macroeconomic variables and yield curve factors.

11 In particular, in the single-factor model the determinant of bond yield is the short-term interest rate. The Vasicek model assumes that short rate follows an Ornstein–Uhlenbeck process and continuously compounded yields are Gaussian with constant volatility. In this model, the risk premium is constant and invariant. The Cox–Ingersoll–Ross model has a similar process than the Vasicek model but short rate is no Gaussian. In both cases, a single factor has not been able to capture empirically the dynamics of term structure models.

12 However, in Wright (2014), the author argues that the VAR process with standard OLS and no bias-correction leads an expected path of short-term yields comparable with surveys, although in both cases the confidence intervals are wide reflecting uncertainty in the parameter estimation.

13 We compare our baseline model with the alternative methodology proposed by Adrian, Crump, and Moench (2013). In Appendices 1 and 2 we report the methodology and compare the DTSM fitted with the baseline model respectively. In Appendix 3 we report the estimated parameters of the DTSM model.

14 In Appendix 4, we report the robustness estimations which leads similar results.

15 In Appendix 5, we report the estimation procedure and VAR coefficient and standard errors.

16 We estimate the term premium for the USA according to our baseline model considering as macroeconomic factors inflation and GDP.

17 For example Fama and Bliss (1987), Campbell and Shiller (1991) and Cochrane and Piazzesi (2005, 2008). See Gürkaynak and Wright (2012) for a review.