Editor’s note: Seeking Alpha is proud to welcome Alexander Steinberg as a new contributor. It’s easy to become a Seeking Alpha contributor and earn money for your best investment ideas. Active contributors also get free access to the SA PRO archive. Click here to find out more »
How much of your investment portfolio can you withdraw and spend annually? Lots of retirees and even people with jobs have to answer this question.
The most well-known approach is a 4% rule proposed and substantiated by William P. Bengen in 1994 known as the “Bengen rule.” It suggests that you can withdraw 4% of your investment portfolio the first year and then the same amount adjusted for inflation each succeeding year. A concise review of this rule was published recently at Rational Walk blog, in which the author proposes a more conservative 3% rule in line with current low interest rates and relatively high stock valuations. How applicable are these simple rules in the long run and is there a better approach?
The Bengen rule and similar rules are based exclusively on historical return and inflation data and nobody knows whether the future will resemble the past. Secondly, the withdrawal rate is expected to be somehow dependent on the specifics of your portfolio, such as asset allocation, investment income, tax rates, and investment real estate that many people hold in their portfolios. Perhaps a simple “one size fits all” rule can be nothing more than ballpark. Finally, the most significant limitation, in my opinion, is the static nature of these rules. Normally people make decisions based on short-term feedback, resulting in outcomes that are more dynamic than what these rules can account for. That said, why would financial decisions in retirement stage be different in this regard?
Here I suggest a practical model that I initially developed for myself, but perhaps can be useful to others. Contrary to static fixed percentage rules, this model is dynamic and accounts for changes in one’s personal circumstances as well as in the financial environment -i.e., interest rates, inflation, taxation, risk asset valuations, etc.
Let us illustrate the model using a hypothetical “Joe,” who is a retiree with an investment portfolio in the table below.
Source: Author’s calculations
Joe splits his assets and sources of income into four baskets:
- High Quality Liquid Assets (HQLA): cash, short-term Treasuries, and short-term liquid bonds from high-quality issuers.
- Fixed Income: primarily investment grade bonds and high-quality preferred stocks.
- Stocks: Joe uses this generic name for all types of risky and volatile assets that also include MLPs, stock mutual funds, ETFs, etc. Risky preferreds and deep junk bonds more likely than not belong to this basket as well.
- Income from Non-Liquid Sources include Social Security, pensions, net cash flow from rental properties, fixed annuities and other similar sources of income.
Please note that Joe did not include income from jobs since he considers it temporary and therefore not reliable in the longer run. Otherwise you can include it in the 4th basket. For simplicity, Joe decided to assign rental income in the 4th basket as well (perhaps Joe does not have this income at all), but many investors may benefit by assigning their rental income in a separate basket.
For each basket, Joe enters its current asset value and current income generated by this basket in 2 columns and, after that, he calculates Reduced Income as income remaining after some of the baskets are liquidated: first HQLA, then Fixed Income assets and then Stocks. This approach implies that we at first spend HQLA assets, then use funds from maturing or sold Fixed Income assets, and finally liquidate Stocks. In reality, we do not have to exhaust assets in this particular order nor are we stating that this order is optimal. Rather we use these assumptions only for modeling purposes since they allow easy calculations.
Now Joe is ready to evaluate his levels of withdrawal as illustrated in the next table:
Source: Author’s calculations
The top row indicates possible range of expenses. Since Joe identified his income from Non-Liquid Sources in Table 1, he can calculate what he needs to withdraw from investments as listed in the second row to cover projected expenses. For example, to cover $60,000 of expenses he needs to withdraw $41,908 from investments since his Income from Non-Liquid Sources is $18,092. Purely for analytical purposes, in the next 3 rows for each level of expenses, Joe calculates withdrawal rate, investment yield, and reliance on capital appreciation, i.e. percentage of Stocks needed to be sold to cover the gap between existing investment income and projected expenses.
The key findings are in the last 3 rows. Knowing his Reduced Income without HQLA contribution in Table 1 Joe can evaluate how many years he can live on HQLA assets alone. For example, for projected expenses of $60,000 and an income (without HQLA contribution) of $53,752, the number of years he can survive on $176,500 of HQLA alone is $176,500/($60,000-$53,752) = 28 (in whole years). Similarly, he can calculate all other remaining cells in the table with one additional yet important assumption.
Since stocks are quite volatile, when calculating the last row (Years on All Liquid Assets or the so-called portfolio longevity) Joe assumes that future Stocks value will be 2 times lower than its current value. Why 2 times? This is where we utilize historical data: from 1928 to 2018 the worst 1-, 2-, 3-, 4-, 5-, and 6-year returns from stocks are correspondingly negative 44%, 58%, 62%, 65%, 49%, and 48% (7-year and longer stock returns are significantly better).
Certainly, the model is quite simple. Before moving further, let us evaluate both the favorable (i.e., making the model more conservative and unfavorable (making it less conservative) assumptions we made together with Joe.
- We ignore stock dividend growth (on average about 6% per year).
- We assume stocks will depreciate 50% from current levels after holding them for many years
- We completely ignore income from assets while these assets are being sold. For example, we assume in the table above that we can survive on HQLA assets for 28 years and during these years we will receive no income from HQLA assets at all.
- Income from illiquid sources can grow with time – rent appreciation, Social Security inflation adjustments, etc.
- Inflation, especially in the long run.
- Taxation – generally investment income and gains are taxed mildly but certain transactions can trigger significant taxes such as IRA withdrawals, sales of highly appreciated stocks and so on.
- Defaults of Fixed Income assets – low-risk event if the assets are of high quality.
- Dividend reductions and interruptions are always possible, but to a significant extent, its effect can be reduced by holding well diversified portfolio of high-quality stocks. In recent history, dividends from S&P 500 went down 21% in 2009, but caught up very quickly.
- Vicissitudes of life – this is a very big, practical factor we can account for only by requiring a generous margin of safety in our model.
Overall, I tend to think that favorables overweigh unfavorables and that the model is quite conservative. This model does not strive to be precise, but rather generates a possible range of outcomes that dynamically adjust to changes along the road.
The most important results are in the last row of Table 2, which I shaded in yellow. Years on All Liquid Funds (portfolio longevity) should be compared with your expected life longevity. If it is significantly higher, then you are generally in good shape. If Joe in our example is 60+, he can definitely spend $60,000 with a significant margin of safety and can probably spend closer to $75,000 at expense of margin of safety. However, he definitely cannot spend $90,000.
Interestingly enough, $60,000-75,000 expense levels roughly correspond to 3-4% withdrawal rates that may seem to confirm their usefulness at least as a starting point. However, I would not overestimate this coincidence. Our model is sensitive to investment income while the Bengen model is not. Joe could increase or decrease his dividend income by, say, moving funds from Berkshire Hathaway (BRK.A) (BRK.B) (that does not pay dividends) to high-dividend stocks or vice versa. These investment actions may cause noticeable divergence between our model and 3-4% percent rules even at a starting point.
Besides portfolio longevity, other calculated numbers in Table 2 are important as well. For example, Joe would be uncomfortable if he cannot live at least 5 years on HQLA alone since it may compel him to sell bonds before maturity or stocks when they trade low. He also likes the ability to enjoy many years before he has to sell the stocks, thus maximizing chances for their significant capital appreciation. There are many ways Joe is actively using data in the last table in making practical financial decisions, but in this article we will not discuss it.
The biggest advantage of the proposed approach is its dynamic nature that allows to generate a forecast at the current point in time and dynamically adjust it with time. It specifically accounts for your asset allocation and the current asset valuation, which can support your financial decision-making in various situations. For example, suppose stocks recently went significantly down. Can you buy more of them and how much additionally can you invest in stocks? Or alternatively: do you have too much of your assets in stocks? Or still: will you benefit and how upon selling your investment rental property? Evaluation of these and similar scenarios is relatively straightforward within the model, but impossible to do within fixed percent rules.
The model does not prescribe any asset allocations or specific investment decisions. They should be made outside of the model, but the model will help to assess risks involved.
Disclosure: I am/we are long BRK.B. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.