The asset allocation will depend on a number of factors like risk aversion, income during retirement (Social Security and pensions), and essential expenses.Ben,
Thank you for all your work on this. I have been playing around with the TPAW planner and my results are nearly identical to the Fidelity retirement planner. This is except for asset allocation. For retiring early at 49, Fidelity would have optimal AA now at 70% stocks/30 bonds. TPAW has me at 35% stock/ 65 bonds. It seems like a pretty big difference. I thought for long retirements you need higher stocks (at least 60%?). Would you help me understand the asset allocation? What am I missing?
Thank you!
DD
But taking an example of a 49 year old retiree planning to age 95, with a $1 million portfolio and $3,000 per month of Social Security starting age 70, and risk tolerance set to the default 12, and other settings also left at their defaults, we get a median glidepath of:
- starting AA of 46/54
- rising to 83/17 by the start of Social Security
- declining to 62/38 by max age
Link to this plan: https://tpawplanner.com/link?params=jYd ... wj9vw2hvGw
You may be getting a lower stock allocation than this due to
- lower risk aversion
- higher ratio of portfolio to Social Security
- extra essential expenses (which will be funded by 100% bonds)
But generally speaking, a long retirement does not call for a higher stock allocation. Stocks are risky, and they don't become less risky over longer horizons. See Risk and Time by John Norstad about the myth of time diversification. Both the risk and the reward of holding stocks increases over time. If returns are independent, the risk and reward increase proportionally and the optimal asset allocation remains the same over all horizons. If returns are mean reverting, risk increases slower than the reward and longer horizons would merit a higher stock allocation. But even if stock returns are mean reverting, we can expect stock risk to be meaningfully higher than bonds even over long horizons. Otherwise there should be no equity risk premium over long horizons. What happens is that over long horizons, stocks become less likely to underperform bonds, but more likely to underperform more severely. For example, over the course of a day, stocks have a high probability of underperforming bonds, but a low probability of underperforming by more than 20%. Over the course of a year, stocks are less likely to underperform bonds, but more likely to underperform by more than 20%. The same effect holds over longer horizons.
The typical SWR analysis using historical returns is prone to arriving at an inappropriately stock heavy recommendation because
(1) SWR analysis grades pass/fail and ignores the severity of failure. Since stocks are less prone to underperform bonds over long horizons, but more prone to underperform severely, ignoring severity underestimates the risk of stocks.
(2) We don't have a lot of historical returns. Even 150 years of US historical returns is only three 50 year runs in a single country that has done very well during that time. Stocks are unlikely to underperform bonds in long horizons, so a small sample of returns may not contain an instance of stocks underperforming bonds over long horizons. But we can reason based on an appeal to efficient markets (and perhaps international data) that this is a rare but—importantly—not an impossible outcome that we can ignore. I wrote more about that here.
In short, we can hold a stock-heavy portfolio over long horizons if we are willing to take both the higher risk and higher return. We should not hold a stock heavy portfolio over long horizons just because we need the higher return to make the portfolio last a long time. Because with that higher return comes higher risk, which over long horizons takes the form of a small probability of severe underperformance. To make the portfolio last over a longer timeframe to support early retirement, you'd have to either save more during working years or spend less during retirement years. Stock-heavy portfolios have frequently been held up as a solution for this problem, but they are not the right solution.
Statistics: Posted by Ben Mathew — Mon Sep 16, 2024 12:55 am — Replies 1000 — Views 278192