Abstract
This paper provides an evaluation of five methods, proposed in the literature, for extracting factors used in the estimation of Gaussian affine term structure models. We assert that irrespective of the method used for extracting state variables, cross-sectional and serial correlations exist in measurement errors. However, using a simulation design which is consistent with the data, we show that parameter estimation using the Kalman filter and the model-free method are quite precise in the presence of serial and cross-sectional correlations in the error term, and, in the presence of different measurement errors, the Kalman filter demonstrates strong empirical tractability.
Similar content being viewed by others
Notes
Please see Yeh and Lin (2003) or Lundtofte (2013) for studies which implement interest rate models using a framework that is different from Duffie and Kan (1996). Lundtofte (2013) study interest rates using a general equilibrium model set up that enables consumers to learn within financial markets, while Yeh and Lin (2003) employ general equilibrium models and curve-fitting techniques (e.g. spline methods) to study the Taiwanese government bond market.
See Sect. 3 for more details on the CDGJ method.
Such restrictions (and the implied volatility specification) are consistent with the Vasicek (1977) term structure model. Ritchken and Sankarasubramaniam (1996) demonstrate that the Vasicek (1977) model is one of the only Markovian models which uses a volatility specification that allows for an analytical relationship between bond prices and interest rates.
Smaller values of λ produce slower decays and thus fit long-term maturities better, while bigger choices of λ produce faster decays and hence fit short-term rates better. We follow Diebold and Li (2006) in their selection of λ to match medium-term maturities.
Comparing the JSZ-NS technique against the JSZ-PC technique resulted in a Chi squared test statistic value of 112.796, comparing the JSZ-PC technique against the JSZ-B technique led to a chi-squared test statistic value of 136.962, and finally, comparing the JSZ-NS technique against the JSZ-B technique led to a chi-squared test statistic value of 121.248. In each case, we have 66 degrees of freedom and the resulting p-values were, respectively, 0.0003, 0.0000, and 0.0000.
\(\Updelta t\) is the time between observations and \(\Updelta z\) is a standard normal shock term.
The intuition behind this simulation design rests on the fact that within the class of affine models and as was explained in Sect. 3, one would construct yield data by constructing the factor (i.e., state variable) and then using the factor to generate yields. Hence, any persistence observed in the data arises from persistence in the factor. The remaining steps compute the factor persistence.
Since correlation is just a scaled version of covariance, the effect of this is to remove the necessity of a scaling factor.
We also run unpaired t-tests in which we assume that each simulated dataset is obtained from independent samples of normal distributions. However, while the main conclusions were essentially the same, we feel that reporting a paired-sample t test is more appropriate since we are examining parameter estimates before and after the addition of noise.
Employing maturities of up to 10 years for estimation is the norm in the literature (e.g., Dejong 2000; Duffee 2002; Tang and Xia 2007; Collin-Dufresne et al. 2008; Dempster and Tang 2011; Duffee 2011b; Duffee and Stanton 2012; Hamilton and Wu 2012). The reason for it is given by Dejong (2000), who notes that, “for the maturities over 10 years, the bond data are quite scarce, so the interpolation is not very accurate.” To overcome the belief that using up to 10 years of maturity may be ignoring the convexity risk in the term structure, and could be the reason why Gaussian models perform well, we investigated this issue further. We collected additional data corresponding to swaps with maturities of 15 and 20 years from Bloomberg and repeated every step of our data interpolation to construct additional zero-coupon yields with maturities ranging from 11 years through 20 years in increments of 1 year. Then, we re-estimated the JSZ normalization using the Kalman filter and the benchmark setting assuming that the 6-month, 2-year, and the 10-year are measured without error and all remaining bonds are measured with error and computed RMSEs corresponding to the difference between model-implied yields and actual yields. The results showed that for bonds with maturities greater than 10 years, the JSZ normalization of the Gaussian model still performed well.
References
Adrian T, Crump RK, Moench E (2013) Pricing the term structure with linear regressions. Staff report No. 340, Federal Reserve Bank of New York
Babbs S, Nowman KB (1999) Kalman filtering of generalized Vasicek term structure models. J Finan Quant Anal 34(1):115–130
Bauer MD, Rudebusch GD, Wu C (2012) Correcting estimation bias in dynamic term structure models. J Bus Econ Stat 30(3):454–467
Berndt E, Hall B, Hall R, Hausman J (1974) Estimation and inference in non-linear structural models. Ann Soc Meas 3(4):653–665
Boulder D (2001) Affine term structure models: Theory and implementation. Working Paper, Bank of Canada
Chacko G, Das S (2002) Pricing interest rate derivatives: a general approach. Rev Finan Stud 15(1):195–241
Chen R, Scott L (1993) ML estimation for a multifactor equilibrium model of the term structure. J Fix Inc 3(3):14–31
Chen R, Yang TLT (2002) A simple multi-factor, time-dependent-parameter model for the term structure of interest rates. Rev Quant Finan Acc 19(1):5–20
Christensen JHE, Diebold FX, Rudebusch GD (2009) An arbitrage-free generalized Nelson–Siegel term structure model. Econ J 12(3):33–64
Christensen JHE, Diebold FX, Rudebusch GD (2011) The affine arbitrage-free class of Nelson–Siegel term structure models. J Econ 164(1):4–20
Collin-Dufresne P, Goldstein R, Jones C (2008) Identification of maximal affine term structure models. J Finan 63(2):743–745
Dai Q, Singleton KJ (2000) Specification analysis of affine term structure models. J Finan 55(5):1943–1978
Dejong F (2000) Time series and cross-section information in affine term structure models. J Bus Econ Stat 18(3):300–314
Dempster MAH, Tang K (2011) Estimating exponential affine models with correlated measurement errors: applications to fixed income and commodities. J Bank Finan 35(3):639–652
Diebold F, Li C (2006) Forecasting the term structure of government bond yields. J Econ 130(2):337–364
Duan J, Simonato J (1999) Estimating exponential-affine term structure models by the Kalman filter. Rev Quant Finan Acc 13(2):111–135
Duarte J (2004) Evaluating alternative risk preferences in affine term structure models. Rev Finan Stud 17(2):379–404
Duffee G (2002) Term premia and interest rate forecasts in affine models. J Finan 57(1):405–443
Duffee G (2011a) Forecasting with the term structure: the role of no-arbitrage restrictions. Working Paper, Johns Hopkins University
Duffee G (2011b) Information in (and not in) the term structure. Rev Finan Stud 24(9):2895–2934
Duffee G, Stanton R (2012) Estimation of dynamic term structure models. Quart J Finan 2(2):1–51
Duffie D, Kan R (1996) A yield factor model of interest rates. Math Finan 6(4):379–406
Duffie D, Singleton KJ (1997) An econometric model of the term structure of interest-rate swap yields. J Finan 52(4):1287–1321
Fisher M, Gilles C (1996) Estimating exponential-affine models of the term Structure. Working Paper, Federal Reserve Board
Hamilton J, Wu C (2011) Testable implications of affine term structure models. J Econ (forthcoming)
Hamilton J, Wu C (2012) Identification and estimation of affine term structure models. J Econ 168(2):315–331
Harvey A (1989) Forecasting structural time series models and the Kalman Filter. Cambridge University Press, Cambridge
Hull J, White A (1990) Pricing interest-rate-derivative securities. Rev Finan Stud 3(4):573–592
Jennrich R (1970) An asymptotic χ 2 test for the equality of two correlation matrices. J Amer Stat Ass 65(330):904–912
Joslin S, Singleton KJ, Zhu H (2011) A new perspective on Gaussian DTSM’S. Rev Finan Stud 24(3):926–970
Joslin S, Le A, Singleton KJ (2013) Why Gaussian macro-finance term structure models are (nearly) unconstrained factor VAR’s. J Finan Econ 109(3):604–622
Joslin S, Priebsch M, Singleton KJ (2012) Risk premiums in dynamic term structure models with unspanned macro risks. Working Paper, University of Southern California and Stanford University
Lin WT, Sun DS (2007) Liquidity-adjusted benchmark yield curves: a look at trading concentration and information. Rev Pac Basin Finan Mark Policies 10(4):491–518
Litterman B, Schienkman J (1991) Common factors affecting bond returns. J Fix Inc 1(1):54–61
Lundtofte F (2013) The quality of public information and the term structure of interest rates. Rev Quant Finan Acc 40(4):715–740
Pearson N, Sun TS (1994) Exploiting the conditional density in estimating the term structure: an application to the Cox, Ingersoll, and Ross model. J Finan 49(4):1279–1304
Perignon C, Smith D, Villa C (2007) Why common factors in international bond returns are not so common. J Int Money Finan 26(2):284–304
Piazzesi M (2003) Affine term structure models. In: Hansen LP, Ait-Sahalia Y (eds) Handbook of financial econometrics, Ch. 12. North-Holland, Amsterdam
Ritchken P, Sankarasubramanian F (1996) Bond price representations and the volatility of spot interest rates. Rev Quant Finan Acc 7(3):279–288
Tang H, Xia Y (2007) An International examination of affine term structure models and the expectations hypothesis. J Finan Quant Anal 42(1):41–80
Taylor JB (2009) The financial crisis and the policy responses: an empirical analysis of what went wrong. NBER Working Paper 14631, National Bureau of Economic Research, Cambridge, MA
Vasicek O (1977) An equilibrium characterization of the term structure. J Finan Econ 5(2):177–188
Yang X, Cheng J (2010) A data-driven approach for building macroeconomic decision support system. IEEE Trans Sys Man Cyber Part A Sys Hum. doi:10.1109/ICSMC.2010.5641732
Yeh SK, Lin B (2003) Term structure fitting models and information content: an empirical examination in Taiwanese government bond market. Rev Pac Basin Finan Mark Polic 6(3):305–348
Acknowledgments
I would like to thank Chris Lamoureux, A. D. Amar, and seminar participants at the 2011 Financial Management Association Annual Meetings held in Denver, Colorado and the San Diego State University Workshop on Finance which also took place in 2011. I would also like to especially thank Cheng Few Lee, the associate editors, and the anonymous referees who are all affiliated with the Review of Quantitative Finance and Accounting for very helpful comments and insights that greatly improved the paper. I feel deeply indebted to them. The Data for the robustness check described in footnote 12 was obtained from the Bloomberg database via the Wells Fargo Financial Markets Laboratory at San Diego State University.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Juneja, J. An evaluation of alternative methods used in the estimation of Gaussian term structure models. Rev Quant Finan Acc 44, 1–24 (2015). https://doi.org/10.1007/s11156-013-0396-2
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11156-013-0396-2
Keywords
- Model evaluation
- Statistical simulation methods
- Financial econometrics
- Model estimation
- Model construction