Long-short extension strategies, such as 130-30, allow ne of the major innovations in portfolio construction during the past several years has been the adoption of long-short extension strategies that allow managers to fully exploit the cross-sectional variation in forecasted security returns. Generalizations of the Grinold and Kahn (1994) theory of active management by Clarke, De Silva, and Thorley (2002) and by others focused on the role of formal constraints in portfolio construction, particularly the negative impact of the long-only constraint. At the same time, innovations in prime brokerage practices and the acceptance of shorting by institutional fiduciaries led to a proliferation of long-short strategies and products. Because long-short extensions are new to many market participants, Jacobs and Levy (2007) addressed misconceptions about the strategies. The analytical model we develop here will further improve investors' conceptual understanding of the factors that determine the size of the short (and equivalent long) extension in long-short strategies.The short extension model is based on the concept of the expected short weight for individual securities in the benchmark, similar to Sorensen, Hua, and Qian (2006). We also use the assumption of a constant correlation matrix and other modeling techniques used in an early analytical treatment of long-short strategies by Jacobs, Levy, and Starer (1998). In this article, we describe how the expected short weight for a security depends on the relative size of the security's benchmark weight and the active weight assigned to that security by the portfolio management process. The formal mathematical model and approximations enhance perspectives from previous studies that depended on time period-specific numerical examples or on insights from simulations.The derivation of the long-short extension model rests on the assumption of an unconstrained portfolio optimization and, therefore, gives an upper bound on possible long-short ratios in practice. In the language of the fundamental law of active management, we assume a transfer coefficient of 1 and thus the maximum possible expected information ratio. 1 As discretionary constraints are imposed, the long-short ratio declines from the upper bound suggested by the model, with a corresponding decline in information ratio. As a result, empirical illustrations that use the S&P 500 Index and other common equity benchmarks have longshort ratios that are generally higher than applied strategies, in which a variety of additional constraints are often used. In addition to being difficult Roger Clarke is chairman and Harindra de Silva, CFA, is president