Flaws in Vanguard?s Withdrawal Strategy: Income versus Total-Return Portfolios
Vanguard advertises that its mission is to simplify investors’ retirement decisions. In a recently published study, however, it oversimplified the critical choices investors and their advisors face in constructing a portfolio for the withdrawal phase of retirement.
The study in question, written by Colleen M. Jaconetti, examined the relative merits of income and total-return strategies. By her definition, an investor who uses an income approach relies only on the returns his or her principal generates; a total-return approach draws down principal as well. Drawing on extensive historical data, Vanguard found that total-return strategies are superior for investors in the withdrawal phase. Vanguard summarized its findings succinctly:
In conclusion, the total-return approach to spending is identical to the income approach for investors whose portfolios generate enough cash flow to meet their spending needs. For those investors who need more cash flow than their portfolios yield, the total-return approach is the preferred method. Compared with the income-only approach, the total-return approach is likelier to increase the longevity of the portfolio…
Not so fast. This study glosses over a number of issues, some of which have a substantial bearing on its results. In this article, I’ll explore the issue of income versus total-return strategies in a manner consistent the Vanguard study, but I find that a total-return strategy does not enjoy an advantage over an income-oriented approach.
The Vanguard study uses historical data going back to 1926 and looks at the range of possible outcomes for a 25-year retirement that starts at each year in that time period. Success in a retirement strategy is measured by longevity risk – the probability that the portfolio will be depleted prior to the expiration of the 25-year horizon.
In a slightly odd twist, for retirements starting in the last 25 years, Jaconetti reverted to data from 1926 to provide a continuous data series. For example, the retirement path starting in 2007 was modeled using returns from the 1920s and 1930s. Regardless, the approach chosen by Vanguard makes sense for the most part.
The only major possible objection to Vanguard’s overarching methodology is that the average returns on stocks for the period from 1926-2007 were higher than we expect in a future characterized by PIMCO’s “new normal” of low economic growth and muted returns. John West of Research Affiliates has recently presented the case that expected returns on stocks will be well below their long-term average. Vanguard’s data nevertheless shows average nominal returns for stocks of approximately 10% per year and nearly 6% for bonds (see their Figure 1).
An additional limitation that comes with using purely historical data is that there is no way to model income-generating asset classes such as high-yield bonds because there is insufficient long-term data on these asset classes.
In my approach to the same question Vanguard tackled, I used Monte Carlo simulation (the Quantext Portfolio Planner). The benefit of using a Monte Carlo simulation instead of historical data is that I can explicitly set the expected return for stocks (the equity risk premium). As a baseline setting, I assume that equities (the S&P500) will return approximately 8% nominally and that inflation will be 3% per year, which is its average over the past century. Another benefit of Monte Carlo simulation? I can project risk and return for any asset class, and I can also account for the levels of correlation between asset classes observed in modern times. This approach allows me to examine the role of income-generating asset classes such as corporate bonds.