Forecasting Asset Returns in State Space Models

Boos, Dominik
October 2010
University of St. Gallen, Business Dissertations;10/26/2010, p1
Expectations about the future evolution of the economy are of immense importance for taking the right decisions in a stochastic environment. Econometricians have long been studying forecasting techniques to this end. Most of this work is based on regression techniques such as OLS or GMM. I propose a different approach: state-space models and their estimation by means of maximum likelihood using the Kalman filter. While the two techniques are often identical in an environment with clean data; state-space models are clearly superior if the observed data is affected by measurement error or displays a seasonal pattern. In this case, state-space models allow the separation of the true underlying signal from the measurement noise. As only the signal is relevant for prediction, this can considerably improve the quality of the forecast. In particular, I use the state space framework to estimate affine yield curve models and find that the implied return forecasts for long bonds is much more reliable than that implied by a linear regression, although the implied insample R2 is lower. Moreover, I detect substantial predictability of long/short portfolios not properly revealed by a linear regression. I then generalize the affine yield curve models such that they can include persistent shocks or state variables not spanned by yields. Firstly, these unspanned factor models are used to further improve the yield-curve forecast by including expected inflation as an additional state variable. In this model, the R2 of the annual term premium forecast is above 30 percent. Secondly, I build a joint stock-bond model that merges the yield curve model with a stock market model using the price-dividend ratio as an additional variable. This is achieved by linearization using the Campbell-Shiller approximation. Thirdly, the cross-section of assets is enlarged by including size and book-to-market sorted portfolios. This model provides evidence for substantial variation in the dividend growth rate. Once the model captures this feature, it is able to explain a large fraction of the value premium by a higher exposure of value stocks to the single persistent shock of the system. Finally, this thesis uses rank-reduction techniques to explore the return predictability pattern. This analysis provides strong evidence for at least two independent predictability factors: the term premium and the equity premium.


Related Articles

  • Decomposition of Optimal Portfolio Weight in a Jump-Diffusion Model and Its Applications. Jin, Xing; Zhang, Allen X. // Review of Financial Studies;Sep2012, Vol. 25 Issue 9, p2877 

    This article solves the portfolio choice problem in a multi-asset jump-diffusion model. We decompose the optimal portfolio weight into components that correspond to a collection of fictitious economies, one of which is a standard diffusion economy, and the others of which are pure-jump...

  • Special Properties of Private Equity in the Context of Portfolio Optimization. Fürstenberger, Mattias Thomas // University of St. Gallen, Business Dissertations;5/13/2011, p1 

    The following study develops a continuous time model evaluating the return loss from the delayed investment flow into Private Equity investments. Using several assumptions, an optimal rule to invest the committed but not invested capital can be derived analytically. The model context is extended...

  • VaR Only as Good as the User's Imagination. Rossi, Clifford // American Banker;6/20/2012, Vol. 177 Issue 95, p8 

    The article opines the value-at-risk (VaR) model, a metric to estimate bad outcomes in market losses or "tail risk" in portfolio management, needs refinement because it is vulnerable to errors caused by unforeseen market conditions and mistaken assumptions.

  • Econometric Models for the Analysis of Financial Portfolios. ANGHELACHE, Gabriela Victoria; ANGHELACHE, Constantin; DINCĂ, Zoica // Romanian Statistical Review;2013, Issue Sup, p92 

    Using a factorial design for explaining the rentabilităţilor allows to reduce the volume of such calculations as long as the number of factors is less than the number of assets. Under these circumstances, rather than to introduce wording ARCH directly into rentabilităţilor, estimate...

  • EFFECTS OF PURCHASING POWER RISK ON PORTFOLIO DEMAND FOR MONEY. Chen, Andrew H. // Journal of Financial & Quantitative Analysis;Jun79, Vol. 14 Issue 2, p243 

    The article addresses the effect of purchasing power risk on liquidity preference. The author addresses previous work from Arrow, Tobin, and Pratt regarding risk aversion. The article considers liquidity preference with purchasing power risk in a static model, liquidity preference with...

  • Optimization strategies in credit portfolio management. Ivorra, Benjamin; Mohammadi, Bijan; Ramos, Angel Manuel // Journal of Global Optimization;Feb2009, Vol. 43 Issue 2/3, p415 

    This paper focuses on the application of an original global optimization algorithm, based on the hybridization between a genetic algorithm and a semi-deterministic algorithm, for the resolution of various constrained optimization problems for realistic credit portfolios. Results are analyzed...

  • GURU. Harris, Will // Marketing (00253650);4/18/2012, p21 

    The article presents an answer to a question about the development of an econometric model to quantify a brand's return on investment (ROI) rate.

  • Asset-Pricing Implications of Dividend Volatility. Yan Li; Liyan Yang // Management Science; 

    This paper establishes dividend volatility as a fundamental risk metric that prices assets. We theoretically incorporate dividend volatility clustering into a model in which narrow-framing investors are loss averse over fluctuations in the value of their investments. Our model shows that...

  • Portfolio Optimization Theory Versus Practice. Ballentine, Roy // Journal of Financial Planning;Apr2013, Vol. 26 Issue 4, p40 

    The article discusses the topic of investment portfolio construction and optimization, and comments on problems which arise from applying theoretical portfolio optimization models to individual investors. Aspects of portfolio theory including the representation of asset classes by indexes, the...


Read the Article


Sorry, but this item is not currently available from your library.

Try another library?
Sign out of this library

Other Topics