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; a spreadsheet with ~100 rows instead of one with over 1,000. It is easier to graph and explain the results.

For instance, Bengen’s famous Safe Withdrawal Rate study assumed each retiree chose to retire on January 1st and made their annual withdrawal on January 1st. In many cases, a simplification like that is fine. It is unlikely to have changed Bengen’s results if he assumed the person retired on June 7th. Or if he checked the numbers of every single possible day.

Prime Harvesting introduces an interesting problem because of its 20% harvesting trigger. From Daryanani’s paper on rebalancing, we know that “checking every day” on whether you need to rebalance offers superior returns. We know that balances fluctuate throughout the course of the year. And since the whole point of Prime Harvesting is to “lock in gains” by converting stocks to bonds…it seems like checking more regularly than once a year would only improve the strategy.

Schiller makes monthly S&P 500 data available, going back to 1871. That allows us to do a comparison between checking annually and checking monthly: how long does a retiree have to wait before seeing that 20% rise that triggers harvesting?

If you only check once a year — always on January 1st, then your wait times look like

Mean wait: 2,484 days (6.8 years)
Median wait: 1,461 days (4 years)
Standard deviation: 2,223 days

But if you instead check at the start of every month, you catch some of those intra-year rises

Mean wait: 1,693 days (2.1 years sooner)
Median wait: 761 days (1.9 years sooner)
Standard deviation: 1,911 days

That’s what you would naively expect. You harvest sooner because you when things rise up during June and July, you sell them off, before the they fall back in November and December.

As an extreme example of what this looks like in practice: If you retired on January 1965 and only checked annually…you wouldn’t harvest until January 1973. But if you checked monthly, you would have harvested on October 1968. Five and half full years sooner. There was actually a market peak of 106.50 on December 1968. But you didn’t check until January when it had faded away and it wouldn’t reach those levels again until March 1972. (But, again, you didn’t check in March and don’t notice the rise until 9 months later in January.)

But just because you are harvesting sooner, doesn’t mean it is pure, unadulterated good. After all, this also means you’re missing out on momentum, which is part of the secret sauce of Prime Harvesting. It seems likely that you’re going to trigger harvesting soon after reaching 20% instead of riding things up to 25% or 30%. More harvesting but with smaller results. At least, that would be my guess without having yet checked.

So the real question is, do the safety gains outweigh the momentum gains?