# Actuarial Harvesting

Michael McClung’s “Prime Harvesting” is something I’ve written about many times before. (Here’s one example.) It is an attempt to provide a better answer for how you withdraw money from your portfolio during retirement.

The basic idea of Prime Harvesting is pretty straightforward:

- Only sell bonds to fund your withdrawals.
- Whenever your stocks are up an inflation-adjusted 25% from when you retired, sell some and buy bonds.

One of the arguments against Prime Harvesting is that this introduces an arbitrary “memory effect”.

Let’s assume there are twins that are identical in every way. They are both 65 years old. They have identical asset allocations. They have identical portfolios of $1,000,000. They have identical expenses. And so on. Let’s call them Adam and Bob.

Adam retires in January 1942. Bob falls victim to One More Year syndrome and decides he’ll retire in January 1943.

At the end of 1942, the market is up 11.75% in inflation-adjusted terms. Adam doesn’t do anything special because he hasn’t reached the Prime Harvest trigger of 25% yet.

Bob retires at the start of 1943. At the end of 1943, the market is up 20.32% in inflation-adjusted terms. That’s not enough for Bob to trigger Prime Harvesting so he does nothing. But Adam’s portfolio is up a total of 34.43%, which ** is** enough to trigger Prime Harvesting for him, which will result in him selling some stocks and buying some bonds. Now Adam and Bob have different portfolios.

But that seems weird. We said that Adam and Bob are identical in every way. The only difference is one retired slightly earlier than the other. That’s the “memory effect”. Even 25 years later when you retired — the value of the portfolio on that day — will continue to affect the strategy.

The situation seems even weirder when you realise that the exact day has an effect. Retiring the day before and the day after Black Friday 1987 matters. It is hard to come up with a plausible explanation for why a 2 or 3 day difference in *when* you retire should matter to your strategy.

This “memory effect” seems like a mark against Prime Harvesting. Is there a way to have something like Prime Harvesting but without that blemish?

Ken Steiner has written a lot of interesting stuff about using actuarial methods when thinking about your retirement planning & income. Here’s a recent example where he talks about an “Actuarial Budget Benchmark”. Actuarial calculations allow us to solve questions like “how much money should I have if I need $40,000 a year for the next 27 years and expect to earn 6% a year on my investments”

=pv(6%, 27, -40000, 0, 1)

=560,126.55

What if we use this to construct our “harvesting trigger”? We just need three things:

- How much we want to withdraw. We can just use the current year’s withdrawal amount, so this is easy.
- How many years are left? We can pick whatever planning period we want. For the examples below I’ll just choose “until age 100”. So if you’re 82 then you’ve got 18 years remaining in the calculation.
- What is the expected return? This is also tricky. In the examples below I’ll use the global average return on equities since 1900: 5.4%

=pv(5.4%, 100 - current_age, current_withdrawal, 0, 1)

If our current portfolio is bigger than that number, then we harvest — sell some stocks and convert to bonds.

As with everything in life, it turns out I wasn’t the first person to have an idea like this.

A few years ago Salter et al published “Standby Reverse Mortgages: A Risk Management Tool for Retirement Distributions” in the *Journal of Financial Planning*. They use a similar tool for deciding when to take a withdrawal from your portfolio versus using a reverse mortgage.

And in the book *Asset Dedication* Huxley & Barnes use a similar calculation to decide when to extend a bond ladder.

Let’s take a look at some preliminary results of our “Actuarial Harvesting” strategy. This uses the HREFF metric, which I’ve written about before.

So…that’s not great, right? Dead last.

Next time…let’s try to find out why Actuarial Harvesting fares so poorly.