Massaging Mean-Variance Inputs: Returns from Alternative Global Investment Strategies in the 1980s

Abstract

This paper explores the impact of adjustments to the inputs on total returns, terminal wealth, and portfolio turnover in an unconstrained monthly mean-variance (MV) asset allocation over time. It is well known that MV allocations are very sensitive to small forecast errors in the means and covariances. This sensitivity is especially pronounced for errors in means. One way to control this sensitivity to forecast errors is to use Stein estimation. We examined three naive applications of Stein estimation for six individual country stock indexes, five country bond indexes and five cash indexes. This study has two major conclusions. First, any of the suggested adjustments to inputs dominate the results of an unadjusted-input MV optimization. Adjusted-input portfolios have higher mean return, less variance and greater terminal wealth than unadjusted-input portfolios. Second, these improvements become even greater with transaction costs.mean-variance optimization, stein estimation, risk tolerance, optimal portfolio, portfolio performance

How should individual investors diversify? An empirical evaluation of alternative asset allocation policies

“…7 In addition to models that try to maximize the Sharpe ratio, we employ several models that aim at constructing minimum variance portfolios, thereby ignoring information about sample mean returns. The superior performance of minimum variance optimization has been demonstrated in various studies, which mostly concentrate on stock markets (e.g., Haugen and Baker, 1991;Chopra et al, 1993;and Jagannathan and Ma, 2003).…”

Risk–reward optimisation for long-run investors: an empirical analysis

“…These estimation difficulties are by now well documented; early studies are Cohen and Pogue (1967); Frankfurter et al (1971), see also Jobson and Korkie (1980); Jorion (1985Jorion ( , 1986; Best and Grauer (1991); Chopra et al (1993); Board and Sutcliffe (1994). Brandt (forthcoming) gives a very good overview.…”

Portfolio Selection Using Tikhonov Filtering to Estimate the Covariance Matrix

“…Let μ be a vector of expected returns for each of N stocks, and let Σ be an N × N covariance matrix for the returns. The problem can be written as Even though these optimization problems play a central role in a modern portfolio theory, it has been observed that the solutions are very sensitive to their input parameters [3,5,6,7]. Thus, in order to construct a good portfolio using these formulations, the covariance matrix Σ must be well estimated.…”