THE JOURNAL OF INDEX INVESTING 83 FALL 2013 Using Index ETFs for Multi-Asset-Class Investing: Shifting the Efficient Frontier Up P ANKAJ AGRRAWAL PANKAJ AGRRAWAL is an associate professor of finance at the University of Maine in Orono, ME, and founding president of Cloud Epsilon LLC. pankaj.agrrawal@maine.edu T his article seeks to extend classic investing, which is often limited to only domestic equities or to a mix of equities and bonds, to a wider array of lower-correlation non-equity assets. The ready availability of highly liquid index exchange-traded funds (ETFs) on assets such as domestic equities, international equities, treasury/sovereign bonds, real estate, gold bul- lion, and foreign currencies has the potential to extend our availability set and the resulting risk–return ( σ-μ) efficient frontier beyond what is possible with only equity-only port- folios. This article shows that. This could take us a step closer toward meeting the all-inclu- sive, but elusive, “true market portfolio” (Roll [1977]) and thinking “out of the equity box” that we seem to be perpetually trapped in. Investors stand to gain from the additional diversification made feasible by extending into a multi-asset-class (MAC) portfolio- based covariance matrix. The shrinkage of the asset covariance structure (Choueifaty, Froidure, and Reynier [2013]) and an overall reduction in the cross-correlations of the con- stituent assets (Willenbrock [2011]) have the potential to produce efficient frontiers that would not be possible with pure-equity-based portfolios. Eventually, that would be an effi- ciency gain for the investors. The article uses the Markowitz [1956] mean–variance optimization (MVO) process to show that the efficient frontier of a MAC portfolio dominates not only the capitalization- weighted Russell 1000 Index but an all-equity mean–variance optimization (MVO) efficient frontier as well. The risk-adjusted Sharpe ratios (Sharpe [1987]) and the beta of the MAC port- folio are developed and reported as well. Fur- thermore, in addition to the MVO portfolios, a 1/N equal-weighted portfolio is created from the constituents of the MAC portfolio and plotted in the σ-μ space. The alternate port- folios are then tested for relative portfolio effi- ciency by the application of the exact Gibbons, Ross, and Shanken [1989] GRS W-test. LITERATURE AND DATA Besides risk–return efficiency, a trad- able implementation of the diversified MAC portfolio approach was also a consideration in the design of this study. To enable that, a set of highly liquid ETFs were selected (Agr- rawal and Clark [2009]) that represented six major asset classes. Roll [2013] wrote that “across asset classes, ETF heterogeneity might be acceptable… though it is not that impressive within each class” and discussed the equity, bond, commodities, and currency asset classes. The Blake, Lehmann, and Tim- mermann [1999] strategic allocation study on U.K. pension funds included real estate as an additional asset class. Black and Litterman [1992] showed how quantitative asset allo- cation models could significantly improve The Journal of Index Investing 2013.4.2:83-94. Downloaded from www.iijournals.com by Pankaj Agrrawal on 10/10/13. It is illegal to make unauthorized copies of this article, forward to an unauthorized user or to post electronically without Publisher permission.