Smart Indexing Under Regime-Switching Economic States

ABSTRACT

Index funds that track a benchmark, such as the market cap-weighted S&P 500 index, tend to have portfolio holdings biased towards slower-growth large-cap equities that result in the fund’s under-performance, especially in economic downturns. We develop a rigorous quantitative framework that allows dynamic-rebalancing of the allocations such that portfolio exposure in a market segment can change periodically based on economic activity, measured via a set of macro-economic and financial indicators. The method incorporates potential shifts in the economic state, and the likelihood thereof, to determine the fund’s risk orientation optimally in tracking or not tracking the benchmark index. That is, the greater the likelihood of a stronger economic state, the higher the degree of tracking the market index; however, a lack of confidence in the economic state results in a more index-neutral portfolio composition. The proposed smart indexing optimal strategy generates superior risk-adjusted returns consistently in out-of-sample testing, relative to (pure) index tracking. We test several variants and present sensitivity analyses that support our actively-managed smart indexing approach.

KEYWORDS:

• Index-tracking
• smart indexing
• portfolio optimization
• risk functions
• regime-switching uncertainty
• Hidden Markov Models
• actively-managed funds

Acknowledgments

We sincerely thank the anonymous referees, and the editorial team, for their thorough and insightful suggestions to improve the paper. Any remaining errors in the paper are ours. We also thank the participants at World Finance and Banking Symposium and Multinational Finance Society meeting for their many comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. http://www.ftserussell.com/files/research/smart-beta-2017-global-survey-findings-asset-owners

2. The mutual funds in the U.S. had total net assets of $16.3 trillion in 2016, of which $2.6 trillion was under the management of 421 index mutual funds (a growth of nearly five-fold since 2004), and investors added $197 billion in net new cash flow to these funds in 2016 alone, see Investment Company Institute ICI (2017). A hefty 81% of these assets were invested in stock index funds, with S&P 500-based index funds representing one-third of it.

3. For instance, a 11% exposure to technology in the S&P 500 index in 1998 became almost 40% of the market capitalization of the index by March 2000, but by the end of 2002, the technology portion of the S&P 500 had lost more than half of its value from the high in March 2000.

4. According to the data from Invesco PowerShares, there were about 350 smart-beta ETFs available to U.S. investors in 2014, with a total value of more than $230 billion, an increase from 212 smart-beta funds with less than $65 billion in assets in 2010. Smart Beta strategies are growing nearly two-times faster (based on 3 YR CAGR) than the overall ETF market.

5. While we have chosen BIC as the information criterion here, there are other possibilities such as Akaike’s AIC or Bozdogan’s ICOMP. The main difference of AIC or BIC with ICOMP is that the latter penalizes the covariance complexity of the model instead of penalizing the number of free parameters, see Bozdogan and Esra (2016). These may lead to a different number of optimal economic regimes on the same data set.

6. One may generalize with trinomial market states by imposing thresholds ${\mu }_2<0<{\mu }_1$ so that specific portfolio strategies are specified when ${\mu }_{b{k}_t^{\ast }}>{\mu }_1$, ${\mu }_{b{k}_t^{\ast }}<{\mu }_2$, or ${\mu }_{b{k}_t^{\ast }}\in [{\mu }_2,{\mu }_1]$ hold with certainty. Such an approach would result in a new smart indexing approach, and it may be further enhanced by identifying additional portfolio strategies on even a finer partition for multinomial market states.

7. This cost rate is consistent with the Interactive Brokers published rates for an average ETF share price of $50-100.

8. Factor data for the period 2008–2019 is downloaded from the website:https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html