Simple VARs cannot approximate Markov switching asset allocation decisions: An out-of-sample assessment

• Published in: Computational statistics & data analysis Vol. 56; no. 11; pp. 3546 - 3566
• Main Authors: Guidolin, Massimo, Hyde, Stuart
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-11-2012
• Subjects: Allocations; Computation; Markov processes; Markov switching; Mathematical models; Nonlinearity; Out-of-sample performance; Predictability; Risk; Strategic asset allocation; Switching; VAR; Vector autoregressive models; Predictability; Markov switching; Out-of-sample performance; Strategic asset allocation; Vector autoregressive models
• ISSN: 0167-9473; 1872-7352
• DOI: 10.1016/j.csda.2010.10.006

In a typical strategic asset allocation problem, the out-of-sample certainty equivalent returns for a long-horizon investor with constant relative risk aversion computed from a range of vector autoregressions (VARs) are compared with those from nonlinear models that account for bull and bear regimes. In a horse race in which models are not considered in their individuality but instead as an overall class, it is found that a power utility investor with a relative risk aversion of 5 and a 5 year horizon is ready to pay as much as 8.1% in real terms to be allowed to select models from the Markov switching (MS) class, while analogous calculation for the whole class of expanding window VARs leads to a disappointing 0.3% per annum. Most (if not all) VARs cannot produce portfolio rules, hedging demands, or out-of-sample performances that approximate those obtained from equally simple nonlinear frameworks.