Yarema Okhrin

Multivariate Shrinkage for Optimal Portfolio Weights

Abstract This paper proposes a multivariate shrinkage estimator for the optimal portfolio weights. The estimated classical Markowitz weights are shrunk to the deterministic target portfolio weights. Assuming log asset returns to be i.i.d. Gaussian, explicit solutions are derived for the optimal shrinkage factors. The properties of the estimated shrinkage weights are investigated both analytically and using Monte Carlo simulations. The empirical study compares the competing portfolio selection approaches. Both simulation and empirical studies show that the proposed shrinkage estimator is robust and provides significant gains to the investor compared to benchmark procedures.

Properties of hierarchical Archimedean copulas

Abstract In this paper we analyse the properties of hierarchical Archimedean copulas. This class is a generalisation of the Archimedean copulas and allows for general non-exchangeable dependency structures. We show that the structure of the copula can be uniquely recovered from all bivariate margins. We derive the distribution of the copula values, which is particularly useful for tests and constructing confidence intervals. Furthermore, we analyse dependence orderings, multivariate dependence measures, and extreme value copulas. We pay special attention to the tail dependencies and derive several tail dependence indices for general hierarchical Archimedean copulas.

ESTIMATION OF OPTIMAL PORTFOLIO WEIGHTS

The paper discusses finite sample properties of optimal portfolio weights, estimated expected portfolio return, and portfolio variance. The first estimator assumes the asset returns to be independent, while the second takes them to be predictable using a linear regression model. The third and the fourth approaches are based on a shrinkage technique and a Bayesian methodology, respectively. In the first two cases, we establish the moments of the weights and the portfolio returns. A consistent estimator of the shrinkage parameter for the third estimator is then derived. The advantages of the shrinkage approach are assessed in an empirical study.

Bayesian Estimation of the Global Minimum Variance Portfolio

In this paper we consider the estimation of the weights of optimal portfolios from the Bayesian point of view under the assumption that the conditional distributions of the logarithmic returns are normal. Using the standard priors for the mean vector and the covariance matrix, we derive the posterior distributions for the weights of the global minimum variance portfolio. Moreover, we reparameterize the model to allow informative and non-informative priors directly for the weights of the global minimum variance portfolio. The posterior distributions of the portfolio weights are derived in explicit form for almost all models. The models are compared by using the coverage probabilities of credible intervals. In an empirical study we analyze the posterior densities of the weights of an international portfolio.

Time Varying Hierarchical Archimedean Copulae

There is increasing demand for models of time-varying and non-Gaussian dependencies for mul- tivariate time-series. Available models suffer from the curse of dimensionality or restrictive assumptions on the parameters and the distribution. A promising class of models are the hierarchical Archimedean copulae (HAC) that allow for non-exchangeable and non-Gaussian dependency structures with a small number of parameters. In this paper we develop a novel adaptive estimation technique of the parameters and of the structure of HAC for time-series. The approach relies on a local change point detection procedure and a locally constant HAC approximation. Typical applications are in the financial area but also recently in the spatial analysis of weather parameters. We analyse the time varying dependency structure of stock indices and exchange rates. We find that for stock indices the copula parameter changes dynam- ically but the hierarchical structure is constant over time. Interestingly in our exchange rate example both structure and parameters vary dynamically.

Dynamic structured copula models

Abstract There is an increasing demand for models of multivariate time-series with time-varying and non-Gaussian dependencies. The available models suffer from the curse of dimensionality or from restrictive assumptions on the parameters and distributions. A promising class of models is that of hierarchical Archimedean copulae (HAC), which allows for non-exchangeable and non-Gaussian dependency structures with a small number of parameters. In this paper we develop a novel adaptive estimation technique of the parameters and of the structure of HAC for time-series. The approach relies on a local change-point detection procedure and a locally constant HAC approximation. Typical applications are in the financial area but also recently in the spatial analysis of weather parameters. We analyse the time varying dependency structure of stock indices and exchange rates. Both examples reveal periods with constant and turmoil dependencies. The economic significance of the suggested modelling is evaluated using the Value-at-Risk of a portfolio.

Surveillance of the covariance matrix based on the properties of the singular Wishart distribution

A methodology which allows applying the standard monitoring techniques for the mean behaviour of Gaussian processes in the detection of shifts in the covariance matrix is developed. Moreover, the proposed methodology allows the use of an estimator of the covariance matrix based on a single observation. An extensive simulation study reveals the advantages of the considered approach.

Boundaries of the risk aversion coefficient: Should we invest in the global minimum variance portfolio?

Due to estimation risk, the portfolios on the efficient frontier can be statistically indistinguishable from the global minimum variance portfolio. We provide a methodology for determining a bound on the risk aversion coefficient, which separates portfolios that are equivalent or significantly different from the global minimum variance (GMV) portfolio. We conclude that investing in the GMV portfolio is statistically justified for investors with a very wide range of the risk aversion coefficients.

On the exact and approximate distributions of the product of a Wishart matrix with a normal vector

In this paper we consider the distribution of the product of a Wishart random matrix and a Gaussian random vector. We derive a stochastic representation for the elements of the product. Using this result, the exact joint density for an arbitrary linear combination of the elements of the product is obtained. Furthermore, the derived stochastic representation allows us to simulate samples of arbitrary size by generating independently distributed chi-squared random variables and standard multivariate normal random vectors for each element of the sample. Additionally to the Monte Carlo approach, we suggest another approximation of the density function, which is based on the Gaussian integral and the third order Taylor expansion. We investigate, with a numerical study, the properties of the suggested approximations. A good performance is documented for both methods.

Behavior of EWMA type control charts for small smoothing parameters

A general family of EWMA charts is considered for monitoring an arbitrary parameter of the target process. The distribution of the run length is analysed for the case when the smoothing parameter tends to zero. The key impact on the results from the use of the exact variance of the control statistics vs. the asymptotic one and the presence of a head start. For fixed head start, the run lengths for both the exact and asymptotic monitoring procedures degenerate to a binary quantity. To guarantee a feasible monitoring procedure, the head start has to be chosen proportional to the smoothing parameter and the control statistics have to be modified when used with the asymptotic variance. This result underlines the weakness of schemes with a fixed head start and of schemes based on the asymptotic variance if the smoothing parameter is small. The assumptions on the target process are very weak, and are usually satisfied for stationary processes. In addition, the asymptotic equivalence of the EWMA schemes and of repeated significance tests is shown.