# Safe Withdrawal Rates – ERN #6

Today’s post is our sixth visit to the mammoth series of posts on Safe Withdrawal Rates at the ERN blog.

Contents

###### Equity glide paths

In Article number nineteen, ERN looks at equity glide paths – plans to vary the percentage allocation of your pot to equities during retirement.

We’ve already looked at these twice before, first as a way of dealing with sequencing risk, and second as a way of dealing with low future equity returns.

- Outside of the ERN series, we also looked at glide paths in our post called Eat Your Bonds First.

The basic idea is that you increase your equity allocation in the early years of retirement (by spending your bonds instead).

- This protects against a fall in equity prices (sequencing risk) that would otherwise force you to sell stocks at a low valuation.

Note that this is the opposite approach to target date funds, which start with a higher equity allocation and lower it (to “reduce risk”) as you approach retirement.

- These funds made sense when people were most likely to buy an annuity at retirement age, but they are a poor fit for extended drawdown over 30 years or more.

ERN notes that Michael Kitces and Wade Pfau have debunked the idea of reducing equity allocations in retirement.

Here’s Kitces’ preferred glidepath:

Note that the equity allocation goes down in pre-retirement, as per the Vanguard target date fund.

- If we were feeling charitable, we might say that the years just before retirement are used to build up the bond buffer that is consumed in the early years of retirement.

But for long-term drawdown success, you will need a high equity allocation.

- ERN notes that the Kitces equity allocations are too low for FIRE retirees who might be looking at a 60 year retirement.

###### ERN’s simulation

ERN uses his regular rules, with final pot value targets of 0% (depletion), 50% and 100% and a retirement horizon of 60 years.

As well as the regular 21 static equity allocations (0% to 100%, in 5% steps), ERN also looked at 24 glide paths using the following parameters:

- End points of 80% and 100% equities – the more usual 60% target isn’t enough for 60-year retirements, though it might suit most of us.
- Starting points 20%, 40% and 60% below the end point.
- “Slopes” of 0.2% and 0.3% for the 20% paths
- 0.3% and 0.4% for the 40% paths
- 0.4% and 0.5% for the 60% paths

- Passive (slavishly following the slope) vs active (increase only when the S&P 500 is below it’s all-time high).
- Active protects against selling bonds too early – you need them when equities go down.
- Active glide paths will vary with cohort (ie. with market conditions).

ERN’s first result is the failure rate for his preferred 3.5% SWR (at this stage, my favourite is still 3.25% pa).

The top of the chart covers a final equity allocation target of 80%, and the bottom has a target of 100%.

- As you would expect, the lowest failure rates are with full depletion (final value = 0%).

I’m more interested in the central block – 50% final value.

- I’m using this as a safety buffer, with the option to leave bequests if things go well.

In this area, the glidepaths from 60% or 80% to 100% equities look the safest.

- The same is true for the other final target values as well.

Once again, the message is lots of equities, and an increasing allocation through retirement.

- The best glide paths increase equity allocation by 20% to 40%.

The next chart shows the failure rates when the Shiller CAPE is high (above 20).

- It’s the same message – 60% and 80% to 100% are the safest.
- Active glidepaths are slightly better than passive.

The advantage over the static allocations is larger when CAPE is high.

The next chart shows failsafe SWRs for all scenarios, plus the 1st, 5th, 10th and 25th percentiles.

- With CAPE > 20, an active 60% to 100% glidepath gives me a 3.47% pa failsafe SWR.
- Note that these failsafes include a zero final value, not the 50% that I would prefer.

Building in a margin of safety for less than perfect execution (and possible bequests), I’m still sticking with a 3.25% pa withdrawal rate.

- The lowest failsafe is 2.58% (100% equities fixed allocation).

For fixed allocations, 75% equities works best (as previously), with a failsafe SWR of 3.25% (even when CAPE > 20).

- It’s also worth noting that when CAPE >20, the 4% rule produces a failure rate of more than 25% under all scenarios.

The final chart from the article reveals where this apparent free lunch comes from.

- A glide path gives you a higher SWR when equity markets fall early in your retirement.

The cost of this is a lower SWR if stocks rally soon after you retire.

- Of course, SWRs are a lot higher under those conditions, so you probably won’t care.

So we have confirmed that glide paths (especially from 60% to 100%) can guard against the sequencing risk that comes with a high equity allocation (which in turn is needed for pension pots to last more than 30 years).

- Glide paths cost money on average, but they deliver just when you need them.

With a static allocation, the 3.25% SWR from 75% equities is the best.

###### Why glide paths work

Article 21 in the series is also about glide paths.

- First ERN explains in more detail why glide paths work.

He uses the following example:

- 10 years of withdrawals of $35K from a $1M portfolio.
- Withdrawals indexed by 2% pa to match inflation.
- Rebalancing to target weights at the time of withdrawal.
- In one scenario, equities fall 30% in year 1, then 5% in year two, followed by 8 years of gains.
- In an alternate scenario, the two bad years are at the end.

- Bond returns are negatively correlated with equities.

The top half of the table uses a glide path from 70% to 90% equities.

- The bottom half uses a constant 80% equity allocation.

The glide path wins easily.

- 60% of glide path withdrawals are from bonds, especially in the first few years.
- 85% of fixed allocation withdrawals are from equities.
- Almost half of the glide path advantage is created during the bull market.

When the bull market comes first, the constant allocation wins.

But of course, the final values are much higher when the bull market comes first.

- The glide path protects you when you need it, and is much less susceptible to sequencing risk (there is much less variation in the final value).

###### Thirty two glide paths

Next, Ern added some more glide paths:

- A Kitces / Pfau glidepath of 30% equities to 70% equities, over 30 years.
- This was chosen to optimise the success of the 4% Rule using historical average returns.
- This has a 0.111% per month passive slope, but ERN added steeper slopes (0.2%, 0.3%, 0.4%).

- A lower glide path of 20% to 60% equities, with the same slopes.

He also switched from high CAPE vs all CAPE to CAPE > 20 and CAPE < 20.

The new paths don’t do at all well (note that they were chosen to cover a 30-year retirement).

- The failsafe winner is still the 60% to 100% equities glidepath, 0.4% per month active (with an SWR of 3.47% pa).

When the CAPE is less than 20, this SWR rises to 3.51% pa.

- With CAPE < 20, 80% to 100% has a higher SWR or 3.52%.

The best fixed allocation for CAPE > 20 is 75% equities (SWR = 3.25%).

- When CAPE is less than 20, 100% equities produces a failsafe SWR of 3.49% pa.

So you don’t need a glide path over 60 years when CAPE < 20.

The next chart is a repeat, but over 30 years.

- Now the failsafe SWR for 60% to 100% equities is 3.9% (CAPE > 20).
- 40% to 100% does even better, with 4.0%.

The best fixed allocation (CAPE > 20) is 70% equities, at 3.83%.

Once again, the Kitces / Pfau glidepaths do badly.

- And once again, glide paths are not needed when CAPE < 20.

The next chart shows the failure rates for various withdrawal rates when CAPE is more than 20.

- ERN shows results for 30 and 60 year retirements.

The highest failsafe SWR for 30 years is 3.75%.

- A fixed allocation of 60% to 75% will give you this.
- So will many of the glidepaths.

Over 60 years, 3.25% is the highest failsafe SWR.

- This comes from a fxed allocation of 75% equities.
- Glide paths of 60 to 80, 80 to 100, 60 to 100 and 40 to 100 (equity percentages) also achieve this.

The 4% Rule does not perform well.

The next table shows failsafe SWRs for different final value targets (0%, 50% and 100%).

- It’s for starting CAPE > 20 and a 60-tear retirement.

The best failsafe for my 50% target is 3.42% pa, with the 60% to 100% glide path.

- The best fixed allocation is 75% equities, but this only gives a failsafe SWR of 3.15% pa.

The advantage of glide paths over fixed allocations is greater when you want to preserve some capital.

###### Monte Carlo

In the final section of the article, ERN looks at why he gets different results to Kitces and Pfau, even over 30 years.

- The major difference is that ERN uses historical returns and Kitces / Pfau use Monte Carlo simulations.

ERN has three issues with Monte Carlo:

- After a major crash, there is always a strong recovery.
- ERN calls this short-term mean reversion.
- In contrast, the random walk of a Monte Carlo has no memory.
- This means that glide paths do better with real data.

- Long-term mean reversion is also missed by Monte Carlo.
- The chart above looks at the S&P-500 return over neighbouring 15-year periods.
- There’s a strong negative correlation.
- This is why glide paths need to be steep – they have to change allocations within 15 years.
- It also why high equity allocations work best.

- Stock / bond correlations.
- Monte Carlo simulations choose a single correlation number whereas in the real world, the correlation is highly variable.
- In the 1970s / early 1980s it was positive (stocks and bonds both lost money).
- From 2000, the correlation has been negative and bonds have been a great diversifier.

ERN has convinced me that the Kitces / Pfau approach (using Monte Carlo) is inferior to his own (using real historical data).

- I’m also glad to see data for 30 years, and for a 50% final value target.

And ERN has explained that the protection of a glide path (when starting CAPE is above 20) comes almost as much during the bull recovery as in the initial bear marrket.

- Glide paths also need to be steep, so that you reach the maximum equity allocation within one bear plus one bull market.

###### Debt in retirement

Post number 21 in ERN’s series is about whether he should have a mortgage in retirement.

- I’m going to widen this out to include debt of any kind, though usually a mortgage is the cheapest debt available to a private investor.

Note that we won’t be looking at the use of borrowings during the accumulation phase.

- There’s a school of thought that says that you should use debt during accumulation.

It’s called lifcyle investing, and we’ll cover it in a future post.

The basic argument is that you can borrow at a rate that is lower than the likely future return on equities, so why not do that to increase your returns.

- This is also known as leverage, and as well as boosting your returns when this go well, it will juice up your losses when things go wrong.

In the specific case of retirement, paying off a debt will front-load your withdrawals and therefore increase your sequencing risk.

- Mortgage interest is tax-deductible in the US, which helps a little, but it isn’t here in the UK.

The only way to avoid this would be to have a re-payment period as long as (or longer than) your retirement.

- Unfortunately, arranging such a loan requires you to know the date of your own death.

And if you did know that, calculating a safe withdrawal rate would be a lot simpler.

Further, the debt acts like a short bond position.

- Repaying it in effect shifts money from stocks into bonds.

Which is the opposite of what you need from your equity glidepath.

- And this again increases sequencing risk.

A debt also makes it more difficult to use a dynamic withdrawal rate (such as the CAPE-based rules that ERN likes).

- The mortgage repayment is a fixed expense that can’t be dodged.

As ERN puts it:

The mortgage will magnify the consumption impact of market volatility.

###### Mortgage simulations

ERN doesn’t have data for 30-year mortgage rates going back to 1871.

- And any simulation would involve lots of assumptions about borrower behaviour.

So instead, ERN says:

I don’t simulate how a retiree in 1929 with a mortgage in 1929 would have experienced the 1929-1989 equity and bond returns.

I calculate how a retiree today with today’s mortgage parameters would fare if we hit him/her with the 1929-1989 real, CPI-adjusted stock/bond returns.

And the 1928-1988 returns and the 1927-1987 returns, and so on.

He looks at eight models, including two baseline with no mortgage:

The variables are mortgage term (30 or 15 years), interest rate (3.25% or 3.875%) and the equity allocation (80% or 100%).

The results show that the best minimum SWR is from the no mortgage, 80/20 model (3.14% pa).

- The mortgaged models (with 100% equities) do well when you don’t need them to – when returns (and SWRs) are high.

The same is true when the CAPE is high at the start of retirement.

- The best minimum withdrawals are from no mortgage, 80% equities.

Mortgages (with 100% equities) only do well in the best 75% of scenarios.

- By adding a debt to your portfolio you increase your risk of failure.

The last chart in the article explains why a mortgage needs to be paired with a high equity allocation, even in the favourable scenarios.

- The returns on bonds are likely to be lower than the cost of the debt.

That’s it for today.

- It’s been a bit of a slog, but only because we’ve looked at some dense (and high quality) posts.

We know everything there is to know about glide paths.

- And we know not to use debt in decumulation.

We have four more ERN articles (and two topics) to go.

- I’ll be back in a couple of weeks with an attempt to get through them all in a single post.

Until next time.

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