Safe Withdrawal Rates: A Do-It-Yourself Approach
January 10, 2012
by Wade Pfau
Reconciling the assumptions that underpin safe withdrawal rate studies with one’s own capital market expectations and constraints is a daunting task, since those studies rarely reflect the practical realities of an advisory practice. But new research now provides a generalized framework for determining a safe withdrawal rate for a given retirement duration, acceptable failure probability, asset allocation and capital market expectations. Advisors no longer must be constrained by the assumptions and choices of others.
I presented this research in my article, Capital Market Expectations, Asset Allocation, and Safe Withdrawal Rates, which was published in the January 2012 Journal of Financial Planning.
The do-it-yourself approach
Most researchers use one set of capital market expectations to study aspects of safe withdrawal rates. Often, they decide to base their analyses on US historical data since 1926. But that is far from a perfect solution. When determining a forward-looking safe withdrawal rate, what matters are our expectations about future market returns as driven by their underlying components: income, growth, and changes in valuation multiples. For instance, the average historical bond return may be of little relevance if current bond yields are at historic lows.
Not all research is based on the historical averages. Sometimes researchers use different assumptions, either because they wish to illustrate a particular concept with basic assumptions, or because they have incorporated their own capital market expectations.
But a general problem remains: Safe withdrawal rates research leaves unanswered questions about what would happen if different asset classes are included, or if different assumptions are made about returns, volatilities, and correlations. My research generalized the framework for safe withdrawal rates, presenting a structure that incorporates user-specified capital market and retirement assumptions.
Accompanying the article, I created a series of graphs at my blog. I used Monte Carlo simulations to calculate the combinations of real portfolio arithmetic returns and volatilities that would support different withdrawal rates for various retirement durations and acceptable failure probabilities.
From here, it is a matter of determining the portfolio’s expected return and volatility based on one’s assumptions about asset classes and their expected returns, volatilities, and correlations. In order to find the optimal asset allocation, these assumptions can be transformed into an efficient frontier. Overlaying the efficient frontier onto my figures lets you find the point corresponding to the highest maximum sustainable withdrawal rate. The asset allocation for the optimal point can then be determined separately from the underlying efficient frontier characteristics.
You have now found your recommended withdrawal rate and recommended asset allocation for your own specifications.
The article also describes a secondary finding of my research: Often there are a wide range of asset allocations that support withdrawal rates nearly as high as the optimal asset allocation. Many retirees will be able to support a withdrawal rate within 0.1 percentage points of the optimum with a markedly lower stock allocation. Conservative retirees need not be pressured into uncomfortably aggressive asset allocations, such as the recommendations for 50-75% stocks found in prominent research articles by William Bengen and others.