(Note: This article has been corrected for adjusted S&P composite prices including dividends based on data from Robert Shiller.)
I retired from a 30+ year career in banking (finance and risk management) in 2013 upon turning 65. Since then, I stopped adding to my retirement accounts (deferred compensation and 401k) and started living on those savings. In the course of several casual conversations with friends at a similar stage in life, the topic often turned to which investment asset class performed best for us. My comment would always be along the lines of: "I have not quantified the returns, but I believe that my long-term investment grade corporate bonds outperformed all of my other investment classes." The usual reaction was disbelief or skepticism. This report presents the results of a formal case study analysis on this topic.
I have constructed an Excel spreadsheet to analyze the risk and return performance of monthly contributions to representative investment funds. The spreadsheet is available for download at github.com/DennisUyemura/retirefunds. It can easily be adapted to analyze a time series of the adjusted prices of any investment fund, as well as for any pattern of contribution amounts and withdrawals.
Case Study Analysis
Input data were downloaded from Yahoo Finance and Robert Shiller. They consisted of monthly adjusted prices from 1/1/1980 through 6/1/2016 for representative funds as follows:
- STGovt (MUTF:VSGBX): Short-term government bonds (Treasuries and Agencies)
- BondAgg (MUTF:VBMFX): Proxy for Lehman Bond Index (all government bonds and investment grade corporate bonds)
- LTCorp (MUTF:VWESX): Long-term investment grade corporate bonds
- SPComp: S&P composite adjusted for dividends using data from Robert Shiller website. (Adjusted prices were derived from the index values and dividends.)
Note: I selected this long time horizon starting at 1/1/1980 because it is appropriate for retirement savings for those about my age. The Yahoo Finance time series was available for LTCorp for the desired time periods. However, the STGovt and BondAgg series began in 1987 or 1988. I created proxies for the missing periods using a 12-month ladder of 12m Treasury bill yields for STGovt and 60m ladder of 5y swap rates for BondAgg. (The downloaded monthly data are as of the first business day of each month. I reset the dates to be on the first of each month.)
Monthly contributions (purchases) were made at the beginning of each month to each portfolio starting with $1,000 at, for example, 1/1/1980. Contribution amounts were grown at a constant 3% annualized rate as a simple inflation adjustment. For each month, calculations were conducted of the portfolio valuations, number of shares purchased, and total returns (based on the IRRs of the streams of growing contributions and final total valuations as of each month).
In conducting any study of total returns, the results are often more a reflection of the exact start and end dates of the analysis. To minimize the influence of this issue, I analyzed a variety of starting dates at 5-year increments starting on 1/1/80, 1/1/85, and so on until 1/1/05. Doing this should accommodate the retirement contribution start dates of workers of a wide variety of ages. The first study used 6/1/2016 as a uniform final assessment date. (A second study using a different end date will be described later to address any bias related to extraordinarily low current bond yields.) The results of the first case study are presented in exhibits 1 and 2.
Exhibit 1: Tabular risk, return, and Sharpe ratio results for all investment funds and all start dates as of 6/1/2016. (Note: Total returns and standard deviations of returns are annualized. Sharpe ratios use STGovt results as a proxy for risk-free returns. Excess returns are the difference between the monthly total returns of the other funds compared to STGovt returns.)
Exhibit 2: Risk versus return profiles for the tabular data presented in exhibit 1.
(For those who are crying foul because of record low current bond yields, please be patient. I will address that bias in the second case study.)
The patterns of results in exhibit 2 suggest that the three bond funds form a kind of "efficient frontier" with the LTCorp fund showing the highest returns and standard deviations among the bond funds. Are the higher returns of LTCorp worth the higher risk? We can examine the Sharpe ratios from exhibit 1. To my eyes, the BondAgg and LTCorp Sharpe ratios are similar, with a slight preference for LTCorp. Hence, the tradeoff of higher risk for higher return with LTCorp seems acceptable. The SPComp returns appear to be slightly higher than LTCorp, but substantially higher in risk. Using the Sharpe ratios, the LTCorp results dominate the SPComp results for all start dates except 1/1/1980. Again, many readers may be repeating their cry of "foul". Please read on.
Relative Asset Values
To remove any "bias" of the current interest rate environment, we should select a different evaluation date. My intuition is to select the middle of 2006, since that was the endpoint of the last concerted period of rising interest rates and avoids the disturbances of the financial crisis. In selecting an alternative evaluation date, it is helpful to chart the relative asset values of the investment funds after setting all of their adjusted prices to a value of 1.0 at 1/1/1980. The resulting relative asset prices are presented in exhibit 3.
Exhibit 3: Relative asset values for the investment funds after indexing all values at 1/1/1980 to a value of 1.0.
The two funds of major interest are LTCorp and SPComp. Notice that the two lines nearly merged in the depths of the financial crisis in late 2008. In mid-2006, when interest rates were peaking, stocks were near a peak. Hence, mid-2006 is a time that should, if anything, favor stocks over bonds. We will conduct a second case study using 6/1/2006 as the endpoint.
Second Case Study
We repeat the case study and analyze the results as of 6/1/2006. The same two exhibits of results will be presented.
Exhibit 4: Tabular results for all start dates ending at 6/1/2006.
Exhibit 5: Risk and return profiles of tabular results from exhibit 4. (Note: The Jan-2005 start point is omitted from this exhibit. It would be too punitive to the bond funds in the sense that Jan-2005 to Jun-2006 was a period of continuously rising rates and the time interval is too short to be representative.)
First, consider the results in exhibit 4. (I will ignore the last row where contributions start on 1/1/2005. There are losses in the bond funds only because the evaluation date of 6/1/2006 was selected to be especially punitive to bond performance. Also, the investment period of 17 months is too short to be representative.) The Sharpe ratios here favor LTCorp over BondAgg in that for four of the five start dates being considered (excluding 1/1/05), LTCorp had the best performance. For the earliest two start dates (1/1/80 and 1/1/85), SPComp outperforms the bond funds. This is attributable to strong relative performance of equities over fixed-income items during the 1980s. However, from 1/1/90 on (excluding 1/1/05), the risk-adjusted returns of bonds outperformed equities by a considerable margin.
In comparing the graphical patterns in exhibit 2 compared to exhibit 5, there is much greater dispersion of LTCorp and SPComp returns in exhibit 5. With exhibit 5, it is important to view the results across funds by individual start date. For example, considering the four points starting at 1/1/80 or 1/1/85, equities outperformed. For the four points starting at any of the remaining start dates, bonds outperform with a modest preference for LTCorp over BondAgg. For the 1/1/2000 start point, there is a considerable outperformance by LTCorp.
Sources of LTCorp Outperformance: Excess Returns of Corporate Bonds?
During my banking career, I developed a strong intuition that long-term investment grade corporate bonds should outperform many alternative investments over long investment periods. My rationale is that there are two sources of potential excess return associated with long-term corporate bonds:
1. The term (liquidity) premium in the yield curve. The implied forward rates embedded within term rates are far higher than actual rollover patterns of short-term rates. This is quantified in detail in my book: Bank Risk Models: The Effects of the Financial Crisis.
2. The credit (default) risk premium for investment grade corporate bonds is far larger than historical loss rates over long investment periods.
Anecdotally, if one believes that the term premium is an excess return, then a corollary should be that one should never accept a fixed-rate mortgage when purchasing a home. Indeed, I have never held a fixed rate mortgage and have never found any benefit in refinancing any mortgage in my life. My mortgage is indexed to one-month Libor.
In conclusion, I believe that this study shows that over many long investment horizons, long-term investment grade corporate bonds have outperformed many other asset classes on a risk-adjusted basis. What about the future? I have no idea when rates will start to rise. However, I do believe that the excess term premium and excess credit spread embedded in long-term corporate bonds will persist even when rates do rise.
Disclosure: I am/we are long VWESX.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.