Clicky

The Best Retirement Withdrawal Strategies: Digging Deeper

New Reader? Get free regular updates from Can I Retire Yet? on saving, investing, retiring, and retirement income. New articles about twice/month. Join more than 11,000 subscribers. Unsubscribe at any time:

A couple of months ago I published new research here. The topic was retirement withdrawal strategies. I was trying to answer a question that had long puzzled me. It’s one that I hadn’t seen addressed elsewhere: What is the best way to liquidate the asset classes in your retirement portfolio? For example, if you have a simple portfolio of stocks and bonds, and you need to generate cash, when do you sell the stocks and when do you sell the bonds?

I built a computer model and simulated the problem using historical data. I found that certain strategies did far better than others. “Success rates” (not running out of money) varied by more than 15%. Ending portfolio values varied by more than $4 million. This was all using the same simple portfolio, based solely on how you withdrew from the major asset classes.

When I clicked “Publish” on that original post, I was taking a chance, dabbling in financial research. But I felt the question of valuation-based retirement withdrawals was important, and hadn’t been explored fully elsewhere. I hoped my article would raise awareness of the issue, offer some useful data, and generate some questions.

And it did. The response was huge. I received more email on that single post than any other that I’ve done. Lots of positive feedback, lots of good questions, and some interesting observations. I’ll try to respond to many of those now, while leaving the rest for future posts….

Validation

For starters, how have my findings stood up, a few months after that initial publication?

Several readers, some with professional credentials in the field, said my work was news to them. Another technically savvy reader with experience in financial modeling wrote to say he had partially confirmed my results for the Equal Withdrawals and CAPE Median strategies. (Thanks Will.)

But I’m an engineer, trained to verify. And I wasn’t fully satisfied. I also had a full plate of new reader questions that I couldn’t easily answer with my original model, dedicated to a single type of simulation. So, I spent some time on a fun new software project, building a much more powerful personal finance model. It lets me research and answer these and many other related questions. You’ll be seeing the fruits on this blog going forward.

The new model is 100% new code from the ground up, carefully tested at each step of construction against other leading financial tools. And, I’m happy to say, when re-calculating the results for my previous research, it produces precisely the same median portfolio values, to the dollar. So, while there could still be questions about my methodology, or my input data, I’m highly confident that my math is correct.

Given the complexity involved, I did find one minor mistake in my earlier results: There was a small rounding error in one of my original functions that led to misclassifying a few failing scenarios as “successful.” So the success ratio for the CAPE Median strategy was overstated by 1.7% (probably not statistically significant), and the success ratios for some of the momentum-based strategies were overstated anywhere from 2-7% (those strategies were already lagging). I have corrected the numbers in my first article. And none of the corrections change my overall conclusions.

New Research

Using the new model, I proceeded to generate a batch of new results to answer some of the many questions you’ve posed. For this new research, I kept the basic scenario from my original article: a $1 million portfolio, with a 4% initial withdrawal rate ($40,000), adjusted annually for inflation, over a 30-year retirement.

Given their lagging performance I dropped the three momentum-based strategies (Last Year, 3-Year, 7-Year) from consideration. I also renamed the “Equal Withdrawals” strategy to “EqualTarget,” reflecting that this strategy is really about withdrawing in proportion to the target asset allocation, which was no longer constrained to 50/50.

Next, by reader request, I added a new withdrawal strategy for consideration: “Proportional.” This is likely the withdrawal strategy you’d be using if you weren’t thinking about this issue at all. It simply withdraws in the same proportion as your portfolio’s current asset allocation, however it has grown. So, if your portfolio is at 65/35 stocks/bonds, so is your withdrawal.

Also by reader request, I began studying rebalancing strategies, in addition to withdrawal strategies. Here is the difference between the two: A withdrawal strategy executes during the year, and specifies the logic for how much or in what proportions to withdraw money from your portfolio for living expenses. A rebalancing strategy executes at the end of the year (or years), after your portfolio has grown, and specifies the logic for transferring money between holdings solely to adjust their proportions.

For starters I added the options to rebalance annually, or not at all. (In future research I expect to look at rebalancing at longer intervals, or based on a “percent band.”)

Lastly, again by reader request, I generated data for a range of starting portfolio asset allocations. Because, not everybody is served by a 50/50 portfolio. So, this time around, I looked at starting portfolio asset allocations of 80/20, 60/40, 50/50, 40/60, and 20/80 stocks/bonds. (One of those should be a fairly close match to your desired allocation.)

I’ve also calculated and reported the average asset allocations over the length of the simulation. This is so you can get a sense for how closely the portfolio’s risk/return profile stays to its starting allocation, depending on the different strategies.

Overview of Results

Fair warning, a bunch of numbers follow below. For those who aren’t inclined or interested, I’m summarizing the main points here, so you can skip the data….

The CAPE Median strategy — choosing to liquidate stocks or bonds based on Robert Shiller’s Cyclically Adjusted Price-to-Earnings ratio (CAPE) — continues to stand out. Variations on this strategy occupy the top three slots in my results for success rate, and the two top slots for ending portfolio value.

The data reminds us of a general principle when progressing from higher to lower stock allocations: holding more stocks generally increases your success rate, and your ending portfolio value. But it also increases volatility. Likewise, the unmodified CAPE Median strategy produce higher average stock allocations, along with higher success rates and ending portfolio values. The odds are for coming out ahead, but it may be a bumpier ride!

One strategy for smoothing out that ride is to add “rebalancing.” Notably, combining CAPE with Annual rebalancing produces impressive success rates and ending values, while reducing volatility and risk.

Contrary to conventional wisdom, rebalancing is not about juicing performance. (Helping you to buy low and sell high.) Rather, rebalancing reduces risk. In my simulations, adding rebalancing always reduces volatility (stock allocation), at the price of reducing your ending value, while having relatively little impact on success rates.

Annual rebalancing is like the Last Year strategy I studied before. It assumes that last year’s outperformance should be liquidated, but in fact we know that most stock market trends last far longer than one year. So it’s inefficient. It becomes obvious from my simulations, as well as a recent article from Michael Kitces, that rebalancing is actually suboptimal for long-term returns. But you might prefer the smoother ride….

Implementation

My original article discussed the simple steps for using the CAPE Median strategy. According to my research, if you apply this strategy consistently, you’ll come out ahead, possibly way ahead, of other withdrawal strategies.

But, if you intend to pursue a CAPE strategy, there is one practical limitation for locating your assets: Much as I love and advocate and own balanced funds, holding separate stock and bond index funds would be necessary to enable selling one or the other asset class at a time.

That said, you can simplify in one area. My research implies that cash/bond “buckets” could be overrated. Bucket strategies are often just alternative views of your asset allocation. Refilling buckets may reduce to the same problem my research tries to address: When and how to sell assets. Rather than stockpiling one asset class, it may be perfectly safe to simply sell whichever one (stocks/bonds), is in favor. Though, personally, I will probably always keep at least a year of cash living expenses on hand in retirement.

My ongoing research into retirement withdrawal strategies continues to underline the importance of consistency. Reliably following any of my top withdrawal strategies will beat the momentum-inspired moving average strategies, and will almost surely trounce any emotion-driven attempts at market timing. Starting from the same portfolio, the top strategies can put you years, and millions, ahead of the lesser alternatives….

The Strategies

For those who want to dig deeper into my results, or implement their own withdrawal strategy, here are more detailed explanations of the different approaches I studied:

Withdrawal strategies:

EqualTarget — When I was studying only 50/50 portfolios, this strategy was dead simple: Whatever the total required withdrawal in a given year, just divide it equally (50/50) and withdraw the same amount each from the current balances of stocks and bonds. In retrospect that “Equal Withdrawals” strategy seems to make less sense now. Most people aren’t going to withdraw according to their target asset allocation. Rather they will withdraw in proportion to their current allocation, possibly after rebalancing. So I’ve renamed that initial strategy to “EqualTarget.” It seemed to make sense with the 50/50 allocation, assuming the actual allocation would stay in that range. But now the idea of withdrawing in a constant allocation, regardless of what the underlying portfolio allocation is doing, seems wacky. And, compared to the current crop of strategies, the results are mediocre.

Rebalancing — This strategy uses the withdrawal as the means for bringing the asset classes back to the starting, target allocation, if possible. So, if the difference between the current holding of stocks and bonds is less than the withdrawal amount, the strategy brings them back to precisely the target allocation. If not, it gets as close as possible. This withdrawal strategy differs from actual rebalancing in that it moves no more money than required for the withdrawal, and may fail to always bring the portfolio back to target. It seeks to minimize transactions in your account.

CAPE Median— This is the simple strategy based on Robert Shiller’s Cyclically Adjusted Price-to-Earnings ratio (CAPE). If the CAPE is greater than its long-term median, then we assume that stocks are highly valued, and the strategy withdraws entirely from stocks. If the CAPE is below its long-term median, the strategy withdraws entirely from bonds. Despite the apparent simplicity, it remains the leading strategy in many scenarios.

Proportional — This strategy simply withdraws in the proportion of the portfolio’s current asset allocations. If your portfolio started at 50/50 but has drifted to 55/45, then this strategy makes an annual withdrawal of 55% stocks/45% bonds. This is what most people would probably do naturally, in the absence of another strategy.

Rebalancing strategy:

Annual — Here we are talking about a rebalancing strategy, not a withdrawal strategy. This one is probably the simplest of all rebalancing strategies. At the end of each year, you identify the asset class that has drifted higher and transfer enough into the lagging asset class to bring them both back to your target allocation. So, if your portfolio target is 60/40 but it has drifted to 62/38, then, at the end of the year, you liquidate 2% of your stocks and use the proceeds to buy bonds.

Technical Discussion

One reader pointed out that some of my reported success rates may be a few percent lower than those reported in some of the traditional safe withdrawal rate research. I haven’t compared methodologies in depth, but explanations could include the timing of events in the model. (Bengen apparently deducted withdrawals at the end of the year, whereas I deduct them at the start of the year.) Also my data is for a different time period (1928-2014) than most of the traditional studies, and includes data points that didn’t exist when earlier research was performed. For what it’s worth, the Trinity study reported a 96% success rate for a 4% withdrawal from a 50-50 portfolio over 30 years, which is not far from my results. At any rate I’m not trying to establish safe withdrawal rates, but just to compare the relative performance of various withdrawal strategies.

Have I in fact demonstrated some merit to a valuation-based withdrawal strategy, or do my strategies simply reduce to a rising equity glide path? In other words, since most of my portfolios shift to contain more stocks over time, are they simply benefiting from the historical outperformance of equities? (And the subsequent downside of increasing risk/volatility late in retirement.) We can get some insight by comparing the CAPE and Proportional strategies with Annual rebalancing. That rebalancing holds the asset allocations constant at the starting, target allocation. Yet, even when allocations are held constant, CAPE always beats Proportional for Ending Value, and beats or ties it (in a single case) for Success Rate. So there seems to be value to specific withdrawal strategies beyond just increasing the allocation to stocks.

There remain strong arguments against CAPE as a useful, long term metric. User beware. I’d be out of my league trying to address these points in mathematical detail. But they seem to boil down to “things change,” especially over decades. I’m open to that argument, though I think the bar is higher to argue that averages change as well. Also, I note that this is an “optimistic” argument, right now: If stocks aren’t overvalued, and the market has room to run, that would mean that Cape Median is sub-optimal, but the consequences would be softened, because more economic growth is ahead.

The most worrisome argument to me might be that CAPE hasn’t mean reverted much over two decades. On the other hand, if CAPE has ceased to be useful in the retirement withdrawals context, I’d expect to see it stop working toward the end of my data set. But, my original study showed only a single portfolio failure using CAPE in the last two decades!

Note that I used the long-term median currently being reported for CAPE. One reader pointed out that this requires future knowledge. It’s a valid concern. I’ve put on my list to investigate a simulation that looks only backwards. But how much it matters may boil down to whether we consider the CAPE median to be constant or variable over retirement-length time spans. I don’t know. There are distinguished experts on both sides of that question.

The bottom line is that if you can sell based on valuation you’ll probably come out ahead. Right now, CAPE is the best valuation metric I’ve tested. There could be other useful valuation metrics yet to be discovered.

The Data*

For those who want to dig deeper, here are my detailed numerical results. I suggest sorting the table by clicking on the various columns to get views of the results from different perspectives:

Starting Stock / Bond AllocationWithdrawal StrategyRebalancing StrategySuccess RateMedian Ending Portfolio ValueAverage Stock / Bond Allocation Over Simulation
80% / 20%EqualTargetnone89.7%$8,125,65289.7% / 10.3%
80% / 20%Rebalancingnone93.1%$7,410,60786.2% / 13.8%
80% / 20%CAPEnone98.3%$9,809,41395.8% / 4.2%
80% / 20%CAPEAnnual96.6%$7,122,43580.0% / 20.0%
80% / 20%Proportionalnone91.4%$7,838,57588.5% / 11.5%
80% / 20%ProportionalAnnual93.1%$6,723,29780.0% / 20.0%
60% / 40%EqualTargetnone91.4%$5,881,46879.2% / 20.8%
60% / 40%Rebalancingnone93.1%$4,515,28468.8% / 31.2%
60% / 40%CAPEnone96.6%$8,142,53785.5% / 14.5%
60% / 40%CAPEAnnual93.1%$3,988,18160.0% / 40.0%
60% / 40%Proportionalnone93.1%$5,246,54275.5% / 24.5%
60% / 40%ProportionalAnnual93.1%$3,634,29260.0% / 40.0%
50% / 50%EqualTargetnone89.7%$4,723,10373.3% / 26.7%
50% / 50%Rebalancingnone87.9%$2,741,74858.1% / 41.9%
50% / 50%CAPEnone89.7%$6,771,52177.8% / 22.2%
50% / 50%CAPEAnnual93.1%$2,772,01650.0% / 50.0%
50% / 50%Proportionalnone87.9%$4,156,37068.0% / 32.0%
50% / 50%ProportionalAnnual91.4%$2,549,65050.0% / 50.0%
40% / 60%EqualTargetnone86.2%$3,564,73766.6% / 33.4%
40% / 60%Rebalancingnone86.2%$1,496,78945.6% / 54.4%
40% / 60%CAPEnone81.0%$5,385,55966.5% / 33.5%
40% / 60%CAPEAnnual87.9%$1,794,09940.0% / 60.0%
40% / 60%Proportionalnone87.9%$3,154,68959.6% / 40.4%
40% / 60%ProportionalAnnual87.9%$1,547,49840.0% / 60.0%
20% / 80%EqualTargetnone74.1%$1,531,89949.3% / 50.7%
20% / 80%Rebalancingnone55.2%$117,04720.8% / 79.2%
20% / 80%CAPEnone72.4%$1,682,91334.8% / 65.2%
20% / 80%CAPEAnnual56.9%$240,92320.0% / 80.0%
20% / 80%Proportionalnone72.4%$944,54138.3% / 61.7%
20% / 80%ProportionalAnnual55.2%$149,03320.0% / 80.0%

*Input data is a mix of Aswath Damodaran and Robert Shiller’s data for the S&P 500 stocks, 10-year Treasuries, CPI-U, and CAPE from 1928-2014.