Controlling Risk in Global Multimanager Portfolios

The issue of country and sector effects is important for portfolio construction and risk management. Unlike the many studies that have addressed the issue by analyzing individual stocks, this article focuses on analyzing managed portfolios. Using global data on manager excess returns and portfolio characteristics, I examined the relative effects of country versus sector bets on excess-return variations. I then evaluated alternative portfolio construction strategies by using optimization and simulation methods. I found that active country exposure is a larger driver of variations in individual managers' excess returns than is active sector exposure. Using this information, money managers can learn current investment industry trends in country and sector selection and fund managers can construct their multimanager portfolios to reduce tracking error by weighting individual managers to limit country exposure in an overall portfolio. Portfolio construction and risk management in international equity investing is especially difficult because of the number of risk factors involved—such as country allocation, sector allocation, currency management, and stock selection. Past studies of country and sector effects have found that when country and sector tilts are relatively the same size in relation to the benchmark, country factors make the strongest impact on benchmark-relative volatility. These studies examined individual stock returns, however; similar studies involving global portfolios are lacking.The drivers of individual stock returns and risk may differ from those of portfolio returns and risk because active managers' bets vary in size and volatility. Using the return data and information on characteristics of managed portfolios from the World and World ex United States universes of Frank Russell Company from roughly 1993 to 1998, I examined the relative importance of country and sector effects on the managers' excess-return variations, their types of country and sector bets, and the performance of hypothetical alternative strategies for multimanager portfolio construction and risk control.A simple factor model was used to examine the cross-sectional regression of the managers' excess returns on the benchmark-relative country and sector weights and establish these factors' explanatory power. In these tests, country deviations explained 50 percent of managers' excess-return variations whereas sector deviations explained only 31 percent. In combination, the two factors explained 65 percent of excess-return variations.The analysis of managers' country and sector bets suggests that the following key tilts were significant in contributing to tracking error in the sample studied: For the portfolios benchmarked to both non-U.S. equities (MSCI Europe/Australasia/Far East Index as benchmark) and global equities (MSCI World Index as benchmark), active tilts away from Japanese equities contributed the most to benchmark-relative tracking error, followed by tilts toward nonbenchmark countries.The largest sector impact for non-U.S. global portfolios came from persistent underweighting of financial services, which caused a 43 basis point (bp) increase in tracking error.In global portfolios, managers' sector weightings tended to stay closer to the benchmark, on average, than they did in the non-U.S. global portfolios. Nevertheless, a persistent underweighting of utilities added 21 bps to tracking error.In the portfolio strategy simulations, 500 random draws of the managers in the database were made to evaluate three strategies—equal weighting, optimal weighting to minimize country exposure, and optimal weighting to minimize sector exposure. Compared with the equal-weighting strategy, minimizing country exposure reduced tracking error 108 bps whereas minimizing sector strategy reduced tracking error only 40 bps.This study provides a number of insights into portfolio construction and risk management for pension and endowment fund mangers. First, it provides a clear picture of the overall trends in managed global equity portfolios as to country and sector selections. It also clarifies the close link between the magnitude of individual country and sector tilts and their impact on tracking error. This study suggests that for a given size of the tilt, country tilts generate more benchmark-relative risk than sector tilts. Thus, in selecting investment managers and building international portfolios, fund managers need to assess the correlated risks of country tilts among portfolios. Finally, the study suggests that considering country and sector deviations when establishing or rebalancing manager weightings can be an effective tool for controlling benchmark-relative risk.