Tax-Managed Factor Strategies

Abstract

We examine the tax efficiency of an indexing strategy and six factor tilts. Between June 1995 and March 2018, average value added by tax management exceeded 1.50% per year at a 10-year horizon for all the strategies we considered. Tax-managed factor tilts that are beta 1 to the market generated average tax alpha between 1.59% and 1.89% per year, while average tax alpha for the tax-managed indexing strategy was 2.26% per year. These remarkable results depend on the availability of short-term capital gains to offset. To a great extent, they can be attributed to loss harvesting and the tax rate differential.

Disclosure: Aperio Group manages equity for taxable investors.

Editor’s Note

Submitted 22 March 2018

Accepted 13 November 2018 by Stephen J. Brown

Acknowledgments

The authors are grateful to Financial Analysts Journal editors Stephen J. Brown, Luis Garcia-Feijóo, and Daniel Giamouridis, as well as three anonymous referees, for their insightful comments on early drafts of this article.

Notes

¹ The “Is Your Alpha Big Enough?” series includes Jeffrey and Arnott (1993); Arnott, Berkin, and Bouchey (2011); and Arnott, Kalesnik, and Schuesler (2018).

² Dammon, Spatt, and Zhang (2004); Wilcox, Horvitz, and diBartolomeo (2006); Reichenstein (2006); Fabozzi and Wilcox (2013); and Geddes, Goldberg, and Bianchi (2015) are among the growing collection of works that summarize principles of tax-managed investing.

³ The value of timing the realization of capital gains and losses is discussed in Constantinides (1983, 1984).

⁴ Managing the tax rate differential by taking long gains to raise the cost basis and hence the likelihood of harvesting short-term losses is sometimes known as tax arbitrage. A simulation-based study of the impact of managing the tax rate differential is presented in Stein, Vadlamudi, and Bouchey (2008).

⁵ A number of simulation-based studies address the efficacy of loss harvesting. Berkin and Ye (2003) use simulation to conclude that loss harvesting based on highest in, first out (HIFO) accounting is elevated in a market with high stock-specific risk, low return, and high dividend yield. Geddes (2011) compares the tax benefits of indexed exchange-traded funds (ETFs) with those of tax-managed separately managed accounts (SMAs). Berkin and Luck (2010) analyze the after-tax performance of “extended mandate” equity portfolios, which allow short positions.

⁶ Sialm and Sosner (2018) find that including short positions in a tax-managed equity portfolio can improve performance.

⁷ The start date of our investment horizon is dictated by data constraints. Specifically, we chose the earliest date that would allow us to run backtests in both the US and global markets. Bouts of market turbulence between 2000 and 2010 may have tended to elevate estimates of tax alpha.

⁸ The importance of focusing on an investment horizon of practical relevance is emphasized in Goldberg and Leshem (2014) and Goldberg, Hand, and Cummings (2017). Unlike tax-exempt strategies, otherwise identical tax-managed strategies launched a month apart can have materially different return–risk profiles.

⁹ We scale the one-step portfolio optimization developed in Goldberg and Leshem (2014).

¹⁰ In effect, we are looking at historical after-tax returns through the lens of the current tax environment.

¹¹ State taxes are 0.0% in Texas, and they are 13.3% for short- and long-term gains in California.

¹² The injection of fresh cash into a tax-managed strategy elevates the cost basis and, as a consequence, increases the benefits of future loss harvesting.

¹³ The mechanisms of redemption and creation of units in a large, diversified ETF make the distribution of capital gains relatively rare.

¹⁴ Due to the right skew, conditional value at risk (CVaR) may be a better risk measure for after-tax returns than volatility.

¹⁵ Strategy goals may involve targeting factor exposures or satisfying sector constraints and position limits.

¹⁶ In this study, we use Barra’s US Total Market Equity Model for Long-Term Investors (USSLOWL), which is documented in Bayraktar, Mashtaler, Meng, and Radchenko (2014).

¹⁷ In the liquidation disposition, it is important to be sure that all tax-related effects are incorporated in tax alpha. These effects include the intraperiod taxes on dividends and capital gains as well as the lump-sum payment from the realization of all capital gains at the end of the investment horizon. This consideration applies to both the portfolio and its benchmark.

¹⁸ For example, five-year tax alpha for a Russell 1000–tracking ATBAT (after-tax backtesting analysis tool) run started in July 2008 was 4.61%, whereas it was 0.55% for an analogous run started in March 2009.

¹⁹ Berkin and Luck (2010) find implicit tax rate differential management (referred to as “tax arbitrage” in the reference) in simulated tax-managed extended equity mandates, such as 130/30 portfolios.

²⁰ All strategies are benchmarked against diversified indexes. Active returns to lower-risk strategies are not beta adjusted.

²¹ The impact of rebalancing frequency on performance is an important consideration not addressed in this article. A preliminary investigation of this issue on an index-tracking strategy against the S&P 500 Index showed that less frequent rebalancing lowered tax alpha, but the impact diminished as the investment horizon increased. For example, over the period June 1995–March 2018, rebalancing every other month instead of every month diminished median tax alpha by 0.21% at a 5-year horizon and by 0.14% at a 10-year horizon.

²² Strategy settings used for simulation and live performance differ because the latter are customized to individual investors.

²³ The association between volatility and market decline is known as the leverage effect, and it was first documented in Black (1976).

²⁴ Tracking error is the standard deviation of the difference in return between a portfolio and its benchmark before taking account of taxes. Incorporation of tax alpha introduces a substantial right skew, which calls for a nuanced risk assessment.

²⁵ The decomposition in Figure 7 was obtained by compounding returns to a synthetic strategy consisting of the pre-tax return plus the value of short-term losses.