Academic journal article Journal of Real Estate Portfolio Management

Ex-Ante and Ex-Post Performance of Optimal REIT Portfolios

Academic journal article Journal of Real Estate Portfolio Management

Ex-Ante and Ex-Post Performance of Optimal REIT Portfolios

Article excerpt

Executive Summary. This study examines the out-of-- sample performance of equity real estate investment trust portfolios based on the NAREIT sector indices. The article examines the use of alternative techniques to reduce estimation error and this improves out-of-sample performance. The findings reveal that unlike previous studies of the capital markets, the tangency portfolios tend to out-perform out-of-sample, despite the instability in the weights and the presence of corner solutions. The minimum-variance portfolio continues to under-perform despite the reduction in estimation error.

Introduction

The majority of conventional portfolio/asset allocation studies tend not to robustly address the issue of estimation error or the out-of-sample performance of optimally estimated portfolios. The majority of empirical studies have found that optimal portfolios tend to perform badly out-ofsample, and much of this underperformance can be attributed to estimation error and problems in the estimation of the portfolio allocations. Portfolio optimizers are extremely sensitive to the inputted parameters and further more if not constrained have a tendency to arrive at corner solutions. The fact that optimal portfolios are therefore often not diversified and contain large allocations in a small number of assets goes against the principles of diversification and can lead to poor ex-post performance. These problems are further exacerbated when the portfolios are examined on a rolling basis due to instability in the portfolio weights due to the optimizer's sensitivity to the inputs.

This study examines the performance of optimal real estate investment trust (REIT) portfolios outof-sample through the use of NAREIT sector indices. The use of such a dataset also allows an examination of potential trading strategies available to REIT mutual funds. Such funds have grown considerably over the last decade together with the general growth in REITs. Gallo, Lockwood and Rutherford (2000) note that in 1990 there were only two real estate-based mutual funds in the United States. In 2002, NAREIT lists sixty-three such funds. These funds invest the majority of their portfolios in REITs; therefore the predicable nature of the trusts is of interest to the funds.

The cyclical nature of real estate may induce a degree of predictability into REIT returns that can therefore be exploited by real estate mutual fund managers.

A number of articles have examined alternative methods to improve performance by attempting to reduce estimation error. One of the simplest means of reducing estimation error is to use constraints. The primary advantage to such an approach is that it overcomes the problem of corner solutions, with studies such as Chopra, Hensel and Turner (1993) examining the use of constraints in the capital markets. However, one of the major problems with this technique is that the choice of constraints is at best arbitrary, leading to the results being hard to generalize. Two of the most common alternatives methods used to reduce estimation error are Bayes-Stein estimators and the analysis of the minimum-variance portfolio. Bayes-Stein estimators are used to reduce estimation error and the tendency to arrive at corner solutions. Such an approach reduces the differences between extreme observations in the sample means by effectively shrinking them towards a specified global mean. The approach used here is that proposed by Jorion (1985, 1986), which provides an empirical means of estimating the shrinkage weight. The use of the minimum variance portfolio weights has the advantage in that the portfolio weights are purely determined by the two risk parameters. A number of studies, such as Chopra and Kiemba (1993) and Stevenson (2001b) have found that portfolio optimizers are particularly sensitive to variations in the means. Therefore, the use of the minimum variance portfolio eliminates the largest cause of estimation error from the estimation process. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.