One of the most crucial questions for income investors is how much yield can realistically be achieved in real-world portfolios at different levels of risk. What is the most yield that you can get with a portfolio which has one third of the total risk level of the S&P500, for example? In answering this question, there is a key variable that must also be considered: exposure to changes in bond yields. Some asset classes will tend to rise when bond yields are rising, and others will tend to fall. I quantify this sensitivity using the correlation between a portfolio’s total return with the 10-year Treasury bond yield. A positive correlation means that the portfolio’s returns tend to increase with rising yields, while a negative correlation means the opposite.
The methodology that I have developed compares the yields of a wide range of asset classes to their risk levels. One of the first articles that introduced this approach appeared in Financial Planning in 2010. I further developed the approach under the title ‘The Ultimate Income Portfolio’ at Advisor Perspectives–also in 2010–and I have followed up on the performance of the strategy and how the portfolios have performed in each year since. The key theme is that we want to focus on portfolio characteristics that are measurable. Yield is directly observable. Risk requires some analysis but it is also measurable and past risk levels tend to be a good guide to future relative risk levels between asset classes. I look at historical risk, as well as at projected risk using a portfolio simulation model. The projected risk uses the relationships between asset classes to assist in coming up with expected future risk. In addition, the emergence of options on ETFs provides an objective measure of expected future risk. The prices of options depend on belief’s about future risk. The implied volatilities of options give us an excellent check on risk levels. We have three ways to estimate future risk, then: historical risk, simulated risk, and option implied volatility. When these show consistent results, we have a solid estimate of risk. The critical test is that the simulated risk matches the option implied risk. The agreement is typically very good, which provides some confidence.
After coming up with expected risk levels and yields for each asset class, I then run an optimizer to identify the asset allocation that provides the highest level of yield for a given risk level and constraint on interest rate exposure (as measured by correlation to Treasury bond yield). This is similar to the process used to create the ‘efficient frontier’ except that the traditional efficient frontier is optimized to identify portfolios with the highest return at each risk level rather than yield. The efficient frontier of return vs. risk is problematic for decision making because expected return for each asset class is highly uncertain. Past yield for portfolios is an excellent predictor of future yield (as opposed to total return, for which the past is not a robust predictor), so the yield vs. risk frontier is a far more useful analysis for making decisions.
In the tables below, I provide the yield and interest rate exposure vs. expected risk for each of a series of income asset classes, followed by the yields, risk levels, and interest rate exposures for the optimal yield portfolios. I do not provide the actual allocations for the optimal yield portfolio here–this is a premium service–but an example of such portfolios analyzed in this framework is provided here. Even without knowing the makeup of these optimal portfolios, this table is exceedingly useful as a benchmark that shows the income levels that are achievable in the current market conditions.
Individual Asset Classes
The starting point for analyzing the optimal yield portfolios is to assess the universe of possible asset classes for inclusion in an income portfolio. A summary of the yield, risk, and correlation to 10-year Treasury bond yield for individual asset class ETFs is shown below. This table is sorted on the basis of yield, with the highest-yield asset classes at the top.
The volatility and expected return are projected values from Quantext Portfolio Planner (QPP). The yield is the current yield and the correlation to 10-year Treasury bonds uses three years of historical data. The projected values use the most recent three years of historical data as inputs to QPP.
In general, bonds have negative correlations to Treasury bond yield and stocks have positive correlations to Treasury bond yield.
|ETF||Ticker||Volatility||Yield||Correl to 10-Year Treasury Bond Yield||Expected Return|
|iShares Mortgage REIT||REM||18.1%||15.3%||-3.5%||9.8%|
|Wisdom Tree Global Real Estate||DRW||23.6%||10.1%||17.0%||12.5%|
|SPDR S&P International Dividend||DWX||23.9%||6.5%||37.7%||12.7%|
|iShares High Yield Bond||HYG||10.3%||6.2%||19.0%||5.9%|
|Market Vectors High-Yield Munis||HYD||9.5%||5.7%||-48.5%||5.6%|
|iShares S&P U.S. Preferred Stock||PFF||9.0%||5.5%||20.3%||5.3%|
|PowerShares Build America Bond||BAB||9.3%||5.1%||-78.2%||5.4%|
|iShares Emerging Market Bond||EMB||10.6%||4.7%||-24.9%||6.1%|
|PowerShares Insured National Muni||PZA||8.4%||4.2%||-73.8%||5.0%|
|iShares Corporate Bond||LQD||7.2%||3.8%||-52.0%||4.4%|
|iShares U.S. Utilities||IDU||12.9%||3.1%||-20.8%||7.2%|
|iShares Select Dividend||DVY||10.0%||3.1%||25.8%||5.8%|
|iShares Cohen and Steers REIT||ICF||20.3%||3.1%||11.3%||10.9%|
|iShares 20+ Year Treasury Bond||TLT||17.6%||3.1%||-90.0%||9.5%|
|iShares National AMT-Free Muni||MUB||7.6%||2.9%||-72.8%||4.6%|
|iShares Intermediate Credit Bond||CIU||3.9%||2.8%||-42.8%||2.8%|
|iShares MSCI EAFE Index||EFA||18.2%||2.7%||42.7%||9.9%|
|iShares Aggregate U.S. Bond||AGG||3.7%||2.4%||-86.1%||2.6%|
|Vanguard Total Stock Market Index||VTI||15.8%||1.9%||55.1%||8.6%|
|iShares MSCI Emerging Mkts Index||EEM||25.8%||1.8%||36.1%||13.6%|
|iShares 7-10 Year Treasury Bond||IEF||7.5%||1.7%||-95.8%||4.5%|
|iShares Short-Term Bond||SHY||0.7%||0.2%||-58.9%||1.2%|
|SPDR Barclays 1-3 Month T-Bill||BIL||0.1%||0.0%||-4.0%||0.9%|
This table was updated in November 2013
Optimal Yield Portfolios
The process of creating optimal yield portfolios is quite straightforward. Using the yields of individual ETFs (above), the risk levels of individual ETFs, and the correlations between them, I use an optimizer to determine the portfolios with the maximum yield, given constraints on total risk and correlation to 10-year Treasury bond yield. I have limited the allocation to any single ETF to no more than 20% of the total portfolio. There are two exceptions to this constraint: T-Bills (the proxy for cash, BIL) and the total stock market index (VTI).
The optimal portfolio outcomes are arranged by volatility and correlation to 10-year Treasury bond yield. There are tables showing both yield and expected return. If you are willing to take on a portfolio with volatility of 7.5% and you want zero correlation to changes in the 10-year Treasury yield, you can create a portfolio with a 6.5% yield, for example.
|All ETF Portfolio Yield|
|Correlation to 10-Year Treasury Bond Yield|
|All ETF Portfolio Expected Return|
|Correlation to 10-Year Treasury Bond Yield|
Optimal portfolios as of November 2013
The actual portfolio allocations that correspond to these results are available for a fee.