When you’re retired and need to withdraw money from your portfolio you have to make a decision about how exactly you withdraw that money. In his book Living Off Your Money, Michael McClung calls that “income harvesting” and he provides his own recommended way of doing that.
A few months ago I compared McClung’s Prime Harvesting with Annual Rebalancing…
Prime Harvesting vs Rebalancing (graphs!)
The recent book Living Off Your Money by Michael McClung is excellent, if you’re a retirement research junkie. He…
In that I primarily relied on “lifetime total income” when talking about why Prime Harvesting outperforms Annual Rebalancing.
I’ve also written about concerns with the percentage of bonds held by Prime Harvesting…
Prime Harvesting bond levels: the problem with averages
The last time I wrote about Prime Harvesting (from Michael McClung’s book Living Off Your Money: The Modern Mechanics…
…and an incomplete investigation of how Prime Harvesting performs when using monthly instead of annual returns data.
Prime Harvesting with Monthly vs. Annual Returns
A lot of retirement research uses annual returns. The data is more widely available. The programming is more tractable…
I eventually did a fuller comparison of Prime Harvesting and Annual Rebalancing in a three-part series…
Prime Harvesting: Sleep Well At Night (1 of 3)
[You can now read the second and third parts.]
…that eventually used mortality-weighted certainty equivalent withdrawals with a variable withdrawal strategy to decide that Prime Harvesting does appear to (somewhat) outperform Annual Rebalancing, though the performance gap doesn’t seem to quite as stark as McClung’s research suggests.
But it recently occurred to me that I never looked any of the other income harvesting strategies in McClung’s book. I should apply a similar kind of test to more income harvesting strategies.
McClung used the Maximum Safe Withdrawal Rate as his fitness metric to decide what income harvesting strategies were best. That’s basically using the traditional “fixed (inflation-adjusted) dollar” withdrawal rate strategy, the one we all say isn’t very good, since you should be using a variable withdrawal rate strategy of some sort.
When we do use a variable withdrawal rate strategy, does that change our results?
I’ll look at the following income harvesting strategies:
- Bonds First: Make withdrawals from the bond part of your portfolio. Never rebalance. Eventually you’ll run out of bonds. Start making withdrawals from the stock part of your portfolio.
- Omega Not: If stocks have appreciated above their (inflation-adjusted) starting value then sell some of the excess to cover withdrawals. If that’s not enough then sell bonds. If that’s also not enough, then sell more stocks. Again, there is no rebalancing.
- Weiss: Sell from bonds first. If that’s not enough, then also sell from stocks. Only rebalance when a rebalancing condition is met: you’ve run out of bonds, your stock percentage has dropped below it’s starting point (i.e. you started with 60% stocks and now have 55% stocks), or your stocks have grown by more than 5% annualised.
- 100 minus age in bonds. Every year rebalance but your bond percentage increases slightly every year. At age 65 you would have (100–65) = 35% bonds.
- 110-age in bonds and…
- 120-age in bonds. No one can agree on what number to subtract their age from, so we’ll try a few varieties.
- Prime Harvesting and…
- AltPrime Harvesting. McClung’s two strategies from his book Living Off Your Money. AltPrime Harvesting is the more aggressive (but riskier) one.
- 60% stocks, annually rebalanced
- 35% stocks, for those that want something much more conservative
- 100% stocks
- A glidepath, which is similar to the “number minus age” strategies but doesn’t change linearly. There are lots of possible glidepaths; this is one among many with no pretensions that I’ve chosen the “best” one.
US Data 1871–2015
Here are the results using US data from 1871–2015.
(At the bottom of the post is a link to a Google Sheet that contains all of the tables. That is probably easier to read than this.)
In all those stats, what are the ones that we probably care about most? In retirement, the real risk is not having enough money. So we want to look at metrics that try to paint a picture of that.
Well, the minimum is interesting, since that’s the lowest income we can expect. Omega Not does best, with a minimum of $21,973. But the difference between best (Omega Not) and #8 (110 minus age) is less than 5%. The Glidepath performs poorly but everything else is clustered together, so this doesn’t seem like a determinant.
Omega Not also comes out on top here, with a 5th percentile income of $33,218. Again, the spread for the top 7 strategies is under 4%.
Worst 10% and Mean of Lowest 25%
The same story here. Omega Not on top but not a huge difference among the top 5 or 6 strategies.
In fact, if we mark a cell as green if it is within 5% of the winner we can see that many strategies are pretty close on many metrics for the US dataset.
Finally we mix things up a bit! 100% Stocks and Bonds First take the top two spots with median incomes of $68,521 and $65,026. There’s another cluster of AltPrime Harvesting, Omega Not, and Prime Harvesting around $62,000 in median income.
(In everything we’ve looked at so far, the aged based and glidepath strategies consistently are the worst.)
Since I just said standard deviation isn’t a great metric…
Semideviation instead of Standard Deviation
A lot of investing is based on using standard deviations as a measure of risk. It is used in things like the Sharpe…
…we know better than to put too much weight on it.
We can define a variant of semideviation to measure downside volatility. Normally semideviation is measured against the average. If we just did that then we’d be unfairly penalising strategies that had a higher average. What we really want is to calculate the semidevation relative to some minimum income level. We will use $40,000 (i.e. 4% of your initial portfolio). Due to the widespread knowledge of the 4% Rule, I think many people use that as their idea of a “worst case scenario”. With semideviation-4 we have a measure of how much they miss that goal by. (But it doesn’t tell us how often they miss it.)
If we make a summary of who did best on each metric, Omega Not racks up the most wins. The only time it doesn’t rank on the top 3 (median income) it places 4th.
Other strong strategies are AltPrime Harvesting and Bonds First, though Bonds First doesn’t do well on the semideviation-4 metric.
We also have data available from the UK for over a hundred years. How do these fare on the UK data?
(Yes, there really is a scenario where you end up being able to withdraw almost $500,000 in a single year.)
With UK data, 100% stocks is the clear top performer. It even does well on our semideviation-4 metric that tries to measure downside risk. Omega Not still does well but falls slightly. Bonds First does even better than it did on the US data.
Just as with the US dataset, most of the best performers are actually pretty close to one another, which an absolute ranking obscures. The best minimum belongs to Omega Not with $13,551 but fifth place Prime Harvest is only $88 behind with $13,463. This is easier to see if, again, we mark something green if it is within 5% of the best mark.
We can see that the top five strategies all perform somewhat similarly.
I have less data for Japan than the US or UK. But it still provides a very different picture than those two Anglo-countries.
The best performer is Prime Harvesting, followed by AltPrime Harvesting. 100% Stocks has a decidedly mixed showing. Everything else other than 60% Stocks has at least one questionable metric.
When you look closer, Prime Harvesting’s lead is even stronger. Not only does it come in first, it is more than 5% better than in most metrics. As an example, the 5th percentile income for Prime Harvesting is $41,712 which is 11% better than the runner-up.
I’ve also uploaded everything to a Google Sheet (if you don’t want to try reading the tables in this post).