Asset Allocation Strategy Horse Race

Do sophisticated asset allocation strategies beat simple ones? In the December 2010 version of their paper entitled “Risk Parity Portfolio vs. Other Asset Allocation Heuristic Portfolios”, Denis Chaves, Jason Hsu Feifei Li and Omid Shakernia conduct a horse race among six asset allocation weighting strategies applied to nine asset classes:

1. 60/40: 60% S&P 500 Index/40% Barclays Capital U.S. Aggregate Bond Index.
2. U.S. Pension Model: 55% stocks (80% U.S. and 20% international); 35% bonds (60% U.S. Long Treasury, 20% investment-grade corporate and 20% global); and, 10% alternative investments (2.5% each commodities, REITs, emerging market equities and high-yield bonds).
3. Equal: simple equal weight for all nine asset classes.
4. Risk Parity: equal volatility contribution for all nine asset classes (asset classes contribute equally to expected portfolio fluctuations).
5. Mean-Variance: mean-variance optimization based on a five-year rolling return history.
6. Minimum-Variance: like mean-variance optimization but using only variances and ignoring returns, based on a five-year rolling return history.

The nine asset classes proxies are: Barclay’s Capital U.S Long Treasury Index; Barclay’s Capital U.S. Investment Grade Corporate Bond Index; J.P. Morgan Global Goernment Bond Index; Barclay’s Capital U.S. High Yield Corporate Bond Index; S&P 500 Index; MSCI EAFE Index; MSCI Emerging Market Index; Dow Jones UBS Commodity Index; and, FTSE NAREIT US Real Estate Index. Using annual and monthly data as available for these indexes over the period January 1980 through June 2010, they find that:

• While the Risk Parity strategy does not consistently outperform on a risk-adjusted (Sharpe ratio) basis the Equal or U.S. Pension Model strategies, it does significantly and consistently outperform Minimum-Variance and Mean-Variance strategies (see the chart below).
• Although Risk Parity performs on par with equal weighting, it does provide better diversification in terms of risk allocation and, thus, warrants further consideration as an asset allocation strategy.
• However, we show that the Risk Parity strategy’s performance can be highly dependent on the investment universe. Thus, to execute on Risk Parity successfully, careful selection of asset classes is very critical and, for the time being, remains an art rather than a science.

The following chart, constructed from data in the paper, compares overall and subperiod gross annual Sharpe ratios for six asset allocation strategies applied to the nine indexes selected as asset class proxies. Results indicate that the very simple Equal strategy and the Risk Parity strategy produce the most consistent risk-adjusted performance over the last 30 years.

In summary, evidence from a 30-year horse race indicates that a very simple equal weighting asset allocation strategy is about as good as any other simple or complex strategy.

Cautions regarding findings include:

• The sample period is not long relative to potential secular socioeconomic trends (such as disinflation since 1980).
• Return calculations are gross, not net. Including reasonable costs for constructing annual portfolios tracking these indexes would reduce reported returns. Costs may be substantial for illiquid asset classes.
• Statistical methods generally assume tame return distributions. To the extent distributions are wild, interpretation of resulting statistics breaks down.

Frequently Asked Questions

What is a risk parity portfolio in asset allocation?

A risk parity portfolio weights asset classes so each contributes equally to expected portfolio volatility, rather than allocating equal dollar amounts. In the study, the strategy was applied to nine asset classes, including U.S. Treasuries, corporate bonds, high-yield bonds, U.S. and international stocks, emerging market equities, commodities and REITs.

Do sophisticated asset allocation strategies beat simple ones like equal weighting?

Sophisticated asset allocation strategies did not beat simple ones in a 30-year test (January 1980 to June 2010). Equal weight and Risk Parity produced the most consistent Sharpe ratios across subperiods, and both significantly outperformed Mean-Variance and Minimum-Variance strategies. Equal weighting was about as good as any other simple or complex strategy.

How are mean-variance and minimum-variance portfolios built in an asset allocation comparison?

In the study, mean-variance and minimum-variance portfolios were built by optimization using a five-year rolling return history, with annual rebalancing across nine asset classes. Mean-variance optimization uses both returns and variances, while minimum-variance uses only variances and ignores returns. Performance was compared using gross annual Sharpe ratios over the full period and subperiods.

How does risk parity compare to equal weight and the 60/40 portfolio?

Risk parity performed on par with equal weighting and did not consistently beat Equal or the U.S. Pension Model on Sharpe ratio. It did, however, provide better diversification in terms of risk allocation than equal weighting. Both Risk Parity and Equal produced more consistent risk-adjusted results than the Mean-Variance and Minimum-Variance strategies.

What are the limitations and risks of relying on risk parity backtest results?

Risk parity performance can be highly dependent on the investment universe, so careful selection of asset classes is critical and remains an art rather than a science. The 30-year sample is short relative to secular trends such as disinflation since 1980, and returns are gross, so portfolio construction costs (potentially substantial for illiquid assets) would reduce them. Statistical methods also generally assume tame return distributions, so interpretation breaks down if distributions are wild.