# Safe Withdrawal Rates – ERN #3

Today’s post is our third visit to ERN’s excellent series of articles on Safe Withdrawal Rates.

Contents

###### Google Sheets Toolbox

The seventh post in the series is about a Google Sheets Workbook (ERN call is a Toolbox) that lets you play around with the variables in his calculations.

There are six assets that you can allocate your portfolio to:

- Stocks
- Bonds
- Cash
- Gold
- Small Stocks
- Value Stocks

It would be nice to have Property as an option, and as I am based in the UK, it would also be nice to have UK flavours of these assets.

- Also, since a couple of Fama-French style factors have been included, it would have been nice to have the rest (Momentum and Low Vol in particular).

You can go some way towards a UK tilt by adjusting the projected future return rates.

- There are two settings here: “next 10 years” and “after that”.

You can also set your retirement horizon, your expense ratio and your final value target.

- And you can account for the State Pension using an “additional cash flows” section.

Go and have a play round with it – you can find it here.

ERN explains how his tool is different from the popular cFIREsim:

- We use monthly data, while cFIREsim uses annual
- We project forward return forecasts beyond 2016 year-end so we can calculate SWR for more starting dates.

- For example, the January 2000 cohort is already far underwater.
- Even aggressive return assumptions will still wipe out the portfolio before too long and we like to count those cohorts as 4% SWR failures even before the utter failure is actually confirmed.
- cFIREsim asks you for a specific withdrawal rate and then simulates how that rate would have performed over time for each of the different starting dates.

- We go the opposite route: We specify a final value target and our spreadsheet calculates the exact initial withdrawal rate that would have precisely matched the final value target for 1,739retirement cohorts
- We can then easily calculate the failure rates of different initial SWRs.

The tool calculates:

- The failsafe SWR for both the entire sample and post 1950.
- The 1%, 5% and 10% fail SWRs for the entire sample and post 1950.
- The failure probabilities for SWRs from 3% to 5% (in 0.25% increments) for the entire sample, post 1950 and for three CAPE values (<20, 20 to 30 and >30).
- There’s also chart of these five failure rates by SWR.

There’s also a tab that shows the distribution of final values for a given SWR.

- Using the 4% rule and ERN’s default settings, the median final value after 60 years is 8 times the initial value.
- The largest final value is 62 times the initial value.
- But 10% of cohorts ran out of money.

Another tab shows a time series of SWRs, with a chart.

###### Technical Appendix

Post number eight is a Technical Appendix which looks at the maths behind the simulations.

- I don’t propose to go through it in detail because I don’t think that most readers will be interested in it.

Here is the formula ERN uses:

If you are interested in the detail behind it, ERN’s post is here.

- The method he uses is a variation on the mortgage payment calculation, but using variable returns rather than a fixed payment / return.

###### Dynamic Spending

Post number nine in the series is about Dynamic Withdrawals.

- We came across these at SORC 18, in the presentation by Vanguard.

ERN looks at a form of dynamic spending called Guyton-Klinger (GK) rules, which I hadn’t come across before.

- Essentially, GK provides a pair of guard rails above and below the central SWR (which is adjusted each year by the CPI rate).

If the proposed withdrawal breaks through the upper guardrail, it is reduced.

- If it falls through the lower guard rail it is increased.

GK has four rules:

- Don’t apply the CPI increase after a year where portfolio returns are negative.
- A sub-rule states that the CPI increase should in any case be capped at 6%.

- When the withdrawal rate (percentage) is more than 120% of the initial withdrawal rate, reduce it by 10%.
- When the withdrawal rate (percentage) is less than 80% of the initial withdrawal rate, increase it by 10%.
- Drawdown first from the assets that have had the highest returns, as they are the most overweight.
- This basically means use the cash flows out of the portfolio to rebalance to the target asset allocation.

ERN is sceptical about GK, since it has been used to support SWRs above 4%.

His testing of GK is designed to be as close as possible to his original work on static SWRs:

- monthly withdrawals and revaluations (and rebalancing)
- two assets – US large cap stocks and 10-year Treasuries (in an 80/20 portfolio)
- CPI is skipped in any month where the 12-month trailing (real) return is negative

ERN doesn’t use the CPI cap of 6% as he says it eroded purchasing power in the 1970s

The good thing about GK is that it’s almost impossible to run out of money.

- Using the standard GK 20% rails with 10% adjustments, no cohorts ran out of money.

This required the GK adjustments to be tested for throughout retirement.

- The original GK paper stopped applying the rules 15 years from the end of the planned retirement period.

The bad thing about GK is the flip side of the good thing.

- Sometimes the amount of money you can withdraw is significantly lower (in real terms) than the year 1 withdrawal.

ERN tests GK on a 1966 cohort, the last time that the 4% rule failed over 30 years (its clear that the 2000 cohort will also fail, but it hasn’t happened yet).

Holy Mackerel!!! GK beats the 4% rule and it’s not even close. The GK-4% has surpassed the initial $100 (adjusted for CPI!) after 26 years while the old 4% has gone bankrupt after 28 years.

The 5% rule is almost back to normal and the 6% rule is hanging in there pretty well, too.

The problem lies with the inflation adjusted dollars that are withdrawn each year – the purchasing power of your pension.

- Each of the GK strategies drops below 2% in year 17.
- That’s a halving of purchasing power for GK-4, and even worse for GK-5 and GK-6.

The average withdrawals for the three GK strategies are 2.74%, 3.02% and 3.22%.

- The larger final values reflect the lower withdrawal rates.

And the advertised withdrawal rates will often not be achieved.

###### Debunking GK

ERN had so much material on GK that it spilled over in to post number 10.

He began with a couple more flaws that didn’t make it into post 9:

- Wade Pfau has shown that GK has a 10% chance of reducing withdrawals by 84% within 30 years.
- Results from GK are worse if you begin with an elevated CAPE (above 20, as we are today).

ERN’s next step was to look at the January 2000 cohort, which looks doomed to failure under the classic 4% rule.

- The results are similar to the 1966 cohort that we looked at above.

All of the GK scenarios resulted in a greater than 50% drop in their real dollar withdrawals, and the withdrawals didn’t recover to their original value within 20 years.

Next, ERN looked at all 1700 cohorts from 1981 to 2015.

The median GK-4 retiree does very nicely, with a gradually increasing income.

- But the 25th percentiles spends more time below 4% pa than above it.
- And the 10th percentile never gets above 4% pa, hitting a low of 2.3%.

Although the withdrawal rates (percentages, bottom half of the chart above) stay within the guardrails, the real monthly dollar amounts do not.

- 15% are below the guardrails
- 26% are below the month 1 amount
- 37% are within the guardrails
- and 48% are above the rails.

Next, ERN brings CAPE valuations into the analysis.

When the CAPE ie between 20 and 30, only the median retiree and above do well out of the GK-4 strategy.

- 50% of real dollar monthly withdrawals are below the month 1 amount
- and 24% are now below the guardrails.

With GK-5, even the median retiree spends more time below the initial withdrawal level.

- two thirds of monthly withdrawals are now below the advertised rate
- and a third of monthly withdrawals are below the guardrails.

ERN prefers GK to the plain 4% rule, since you have less chance of running out of money.

- But that small (in frequency, not consequences) risk is replaced by a much larger risk of significant spending cuts through many years of your retirement.

###### Criteria for grading rules

That’s it for GK, almost.

- In post eleven, ERN looks at how to evaluate (mostly) dynamic withdrawal rules (which include GK).

ERN looks at 8 rules:

- Classic fixed initial 4% (with CPI adjustments).
- GK-4 with 20% guardrails and 10% adjustments.
- GK-5
- Constant percentage (4% pa).
- Variable percentage (VPW)
- This is a variation of the RMD (required minimum distribution) strategy based on US government rules designed to maximise tax revenue by ensuring you spend most of your pension pot.
- The annual percentage increases as your remaining life expectancy decreases, so this rule should really be called escalating percentages.
- I looked at RMDs in a post I have written but not yet published on the LTA and how to approach the run-off of my SIPPs.
- As well as your age, your asset allocation will also modify the withdrawal percentage.
- ERN is using the BogleHeads version of the percentage table, rather than RMDs.
- He’s also capping the withdrawal rate at 8%, to ensure the final value is above zero (for bequests).

- A CAPE rule
- ERN uses the adjusted yield (1/CAPE) as a proxy for future stock returns.
- The withdrawal percentage is 1% + 0.5%*(1/CAPE) – this comes from cFIREsim.

- The CAPE rule, but using 1.5% + 0.5%*(1/CAPE).
- 2% + 0.5%*(1/CAPE)

ERN has six criteria to judge the rules by:

- Respond to changing fundamentals in financial markets and the economy
- Provide guidance on the appropriate initial withdrawal rate
- Withdrawal amounts should display low short-term volatility.
- Withdrawal amounts don’t suffer significant and long-lasting drawdowns.
- Withdrawals have the potential to maintain purchasing power over at least 30 years.
- The initial withdrawal amount is not unnecessarily low.

His simulation uses an 80/20 stock bond portfolio and returns data from 1957 to 2016 (60 years).

At first glance, the CAPE rules look more consistent (though over the last 60 years, the 4% rule worked, so it looks fine, too).

- Note that the chart has a log-scale vertical axis to highlight percentage changes.

Looking at the underlying data, the non-CAPE rules show a surprising amount of green, but also a lot of red.

When we look at the report card (ERN’s six criteria) the CAPE rules come out on top, with VPW next.

- If you want to know why ERN marked the scorecard the way he did, the original post is here.

1.5% + 0.5%*(1/CAPE) is probably the winner.

- And at the time ERN wrote the post, that rule would produce a 3.25% withdrawal rate.

###### Conclusions

It’s been a bit of a mixed bag today.

- The posts have been a bit lighter than the first six, which means that we’ve managed to get through five instead of the usual three.
- But as a consequence, we’ve jumped around quite a bit.

Posts 7 and 8 were the conclusion of ERN’s initial look at SWRs:

- 7 provided a workbook for readers to experiment with on their own
- 8 looked at the maths behind ERN’s analysis

In response to reader pressure ERN looked at (or demolished, more like) the GK flavour of dynamic withdrawals in posts 9 and 10.

Then in post 11, he introduced two new flavours of dynamic withdrawals – VPW and CAPE – which turned out to be better than what we had seen before.

- He also introduced six criteria with which to compare all the rules.

CAPE 1.5% base looks like the best approach overall.

- But interestingly, that would produce today an SWR in the range of 3% to 3.25% pa.

Which is where we came in.

Until next time.

*Share this with Twitter, Pinterest, LinkedIn, Tumblr, Reddit, StumbleUpon and WhatsApp.*