SORC 18 Part 3 – Morningstar and the 4% Rule

by Mike Rawson · Published April 30, 2018 · Updated September 4, 2018

Today’s post is our third visit to the the Science of Retirement Conference 2018, which was held in London recently. This time we’ll be looking at Morningstar’s take on the 4% rule.

Contents

Morningstar and the 4% rule
- Preamble
Retirement planning in the UK
SWRs for UK retirees
A few more slides
Conclusions

Morningstar and the 4% rule

The second presentation of the day was by David Blanchett of Morningstar.

The talk was titled “Successful Withdrawal Strategies for UK Retirees” and was based on two Morningstar white papers that he co-authored.

In practice, David looked at whether the 4% rule was still valid in the light of current market and economic conditions.

I’m going to try to bring together the two papers (written in 2016 and 2017 respectively) with David’s slides from the conference.
Since I’ll be using charts from three sources, the numbering could get a little confusing, so bear with me.

Preamble

Before I get to the underlying papers, let’s look at a few slides from David’s presentation that were not directly related to the 4% rule.

He actually put most of these at the end of the presentation, but I think that they sit better at the start.

This one shows the classic approach to how money should be used through your life:

you accumulate wealth through saving while you are working, and
you spend money (dissaving or decumulation) in retirement.

This one shows the possible outcomes at death:

Few people will get things exactly right.
The main choice is between running out of money too soon, or leaving some money behind.

But only one of these is really failure:

The only true failure is still being alive, but having no money left.

In this chart, David compares success with utility.

The success rate of a strategy (as defined in retirement planning) is the percentage change of a pot lasting for a predefined length of time.

But the utility curve is asymmetric:

More money makes us happier at a decreasing rate.
But less money than expected (or none at all) gives us a much bigger problem.

This chart contrasts the theory of decision making in retirement with the reality:

Rather than make one big decision at the start, it’s an iterative (usually annual) process, as we gradually find the best path through a complicated decision tree.

The next chart shows that rebalancing process in more detail.

This chart shows that if you ask people whether there is a crisis of retirement outcomes, many or most of them will say yes.

But if you ask them whether they have a personal crisis, they will say no.
This is particularly true of people who have already retired.

Retirement planning in the UK

On to the first of the papers – the more recent one.

Retirement planning in the UK is from October 2017.

The paper looks at five topics:

Why the SWR is lower than most believe
The impact of strong historical returns (and weaker projected future returns)
The importance of low fees
The benefits of delaying retirement, and
Whether flexible withdrawals can help.

The 4% rule

Here’s what the 4% rule looks like:

You take 4% of your initial pot at retirement as cash in the first year.
In subsequent years you raise this amount only by inflation.

The 4% rule is easy to remember, and implies a nice round number – 25 times your required income – for working out when your portfolio is large enough for you to retire.

But does it work?

The 4% rule is a backward-looking rule.

For most of the twentieth century (in the US), 4% was a safe withdrawal rate (SWR) for a 50% US stocks / 50% US bonds portfolio, with a 30 year time horizon and a 95% target certainty.

Even in the US, 4% was not an SWR during the 1960s (shaded portion of chart above).

Here in the UK, 4% was never really a safe withdrawal rate.

If the initial research had been carried out here, it might be known as the 2.5% rule.
This in turn implies that a pot of 40 times income is needed to safely retire.

The variables

When we talk of SWRs, we always need to be clear on:

The target retirement period
The confidence (security) level required
The asset allocation of the retirement portfolio

Other variables that can be included are:

Portfolio management fees
The extent to which the retiree has a secure retirement income (a state pension, or an annuity).
Or a cash buffer to ensure that market downturns can be waited out.
And sequencing risk, or the problems faced by a retiree whose portfolio falls in value in the first years of retirement.

Past vs future

We also need to be clear on whether any particular analysis is based on numbers from the past, or projections for the future.

Things that have changed since SWR analysis began in the 1990s include:

longer life expectancies
lower interest rates and bond yields
higher stock valuations and therefore lower projected future returns
potentially lower management fees

Life expectancy

UK life expectancy has clearly increased significantly over the past century.

With couples, there is a significant chance that one of them (usually a woman) will survive for longer than the other:

Valuations and returns

The 2017 paper uses the above projected return rates, which are compared to the normal (historic) returns for each asset class.

These numbers had fallen since the earlier (May 2016) paper.

This chart shows the deterioration for a 40% equities portfolio.

And here are the implied SWRs as of September 2017.

They are low (2.1% for 95% confidence of lasting for 30 years) but that reflects the low equity content of the portfolio (40%).

Here’s the actual slide of projected returns that David used, which is a bit harder to read:

Fees

The paper says that:

Fee minimisation could be the single most important variable to increase the SWR for retirees.

The chart below shows the impact of fees of 0.5% pa, 1.5% pa and 2.5% pa on the success rate of the 40% equities portfolio used above:

Delaying withdrawals

Delaying withdrawals (increasing the retirement age) has two effects:

The portfolio has longer to grow.
Fewer years of withdrawals will be needed.

The table below shows the effect of delaying for up to five years on the 40% equity portfolio at a variety of initial withdrawal rates:

There are clear benefits, but with a 40% equity portfolio, none of the numbers look good:

95% confidence (30 years) requires 2.5% pa withdrawals and a 5 year delay.
Or 2.0% pa withdrawals with no delay.

Asset allocation

The next section of the paper looks at varying the equity allocation in the portfolio:

You might expect the higher equity portfolios to offer higher SWRs, but this is only the case at lower levels of certainty:

The effect is muted or non-existent at higher levels of certainty:

You will recall that the the SWR for 40% equities (30 years, 95% confidence) was close to 2.0% pa (actually 2.1% pa).
Increasing equities to 60% leaves this SWR at 2.1% pa.

Worse still, increasing the equity allocation to 80% or 100% actually reduces the SWR.

And for 0% or 20% equities, this SWR rises to 2.2% pa.

I am surprised by this result and it warrants further investigation.

Dynamic withdrawals

By dynamic withdrawals, Morningstar don’t mean the floor and ceiling approach we discussed when looking at the Vanguard lecture from SORC 18.

Rather, they mean a constant SWR in percentage of portfolio terms – where the cash amount withdrawn is proportionate to the portfolio balance at the start of the year.

The chart below shows the effect of this policy by income percentiles (for the bottom half of income percentiles only):

Here the 80th percentile is the worst 1 in 5 runs of the model, and the 95th percentile is the worst 1 in 20 runs.

A 3% pa initial withdrawal rate was chosen, but only the best 50% of runs stay above that level..

At the 95th percentile, a large part of retirement is spend at less than 2% pa withdrawals.

But they don’t run out of money completely.

Since the withdrawal is a fixed portion of what remains each year, that is impossible.

This means that the best measure of “success” for this approach is the percentage of the initial target income that is actually withdrawn.

Even at the 95th percentile, a retiree receives more than 76% of the target income over 10 years.

The table above shows the income results for the 40% equity portfolio, over 30 years, with a 1% annual fee.

Again, for confidence at the 95% level (95th percentile) we need to stick with a 2% pa SWR.

The dynamic approach degrades more gracefully than the fixed SWR (you can’t run out of money).

But it doesn’t boost the initial safe rates.

Dynamic delay

The next approach combines dynamic withdrawals with a delay to retirement:

This table is suppose to cover the same 40% equity portfolio over 30 years with a 1% annual fee.

I’m obviously missing something, because at the 95% percentile, the result from zero years of delay and a 2.0% withdrawal rate is now 170%, rather than 99% as in the previous table.

This table makes it appear that dynamic withdrawals support a 3.5% SWR at the 95th percentile.

And that’s it for the first (latest) paper.

SWRs for UK retirees

On we go to the second, older paper, which is from May 2016.

This begins by covering some of the same ground as the later paper:

where the 4% rule comes from
the historical applicability to the UK
increasing UK life expectancy
return expectations

There’s a nice table of SWRs at various confidence levels for many countries (30 years, 50% stocks, 1% pa fees).

US SWRs are the highest across 20 countries.
The UK SWR at 95% confidence is 2.8% pa.

There’s also a table of inflation-adjusted returns by country.

Here’s a table breaking down how Morningstar forecast returns for stocks and fixed income.

This chart shows the success rate for various withdrawal rates and confidence levels over 30 years.

This shows more clearly how higher levels of equities have little impact at higher confidence levels.

This chart shows the results for a 50% equities portfolio over various retirement periods, confidence levels and initial withdrawal rates.

Once again, a 3% pa withdrawal for 30 years at 95% confidence is not available.

A few more slides

There are a few more slides from David’s presentation that aren’t covered in the papers.

This one shows the high variability of SWRs by country and year of retirement.

The slide shows the cohort effect of life expectancy, where future improvements in mortality during a retiree’s remaining years are taken into account.

This chart looks at sequencing risk by imagining the impact of a single year with a 20% loss.

If the loss happens in year 1 it has a major effect.
By year 30 it has no impact.

The final slide introduces the concept of guaranteed income (from state pensions or annuities) and its effect on SWRs.

For a given degree of flexibility in spending, more guaranteed income means a higher SWR.

What I don’t understand is why increasing spending flexibility implies a lower SWR for a fixed level of guaranteed income.

Conclusions

David’s presentation was very interesting, and more upbeat than the published papers.

Going back to the original papers has in many ways increased my understanding of the issues, but I don’t understand why higher equity allocations don’t raise the SWR.

And I can’t reconcile the “zero years of delay” dynamic option with the standard “no delay” dynamic option.

I’ll try to get in touch with David to clear up these points, but it’s clear from the papers than a 4% SWR is not appropriate for the UK today (indeed, it never was).

Options to remedy this include: