Safe Withdrawal Rates – ERN #4

Today’s post is our fourth visit to ERN’s mammoth series of articles on Safe Withdrawal Rates (SWRs).

Cash Buffers

Article number twelve in ERN’s series is about Cash Buffers, or as he calls them, Cash Cushions.

• This one should be particularly interesting for me, as I use a cash buffer, and (as usual) ERN is sceptical about them.

I realise that keeping a cash buffer of several years of living expenses is a small drag on my returns, but I like the comfort of knowing I won’t have to sell stocks during a market downturn.

• I don’t completely trust myself to do the right thing in a crash, and I want to give myself the best chance of success.

ERN is talking about a two year buffer, but I tend to think in terms of four or five years.

• Usually the stock market will have recovered during that period.

With an SWR of 3% pa, two years works out at 6% cash, and four years is 12% cash.

• I can live with those numbers (though ERN can’t).

---

The first thing ERN warns about is double counting.

• If you spend your dividends, you won’t have them in your pot, compounding away.

I haven’t come across this issue personally, but ERN says:

You’d be surprised how often we see folks making the mistake of double-counting the dividend yield.

They design the annual cash cushion size as withdrawal rate minus yield (assuming consumption of dividends) but then use the equity drawdown length of the Total Return Index (with dividends reinvested).

I won’t be doing this, but I will be keeping 6% to 12% in cash.

• I can’t remember the last time I had 5% or less in cash, and at the moment I’m way over 12% (stock market valuations are very high).

ERN’s second point is a better one.

• Drawdowns last longer when you take into account inflation.

This is true enough, but the real question is whether you would be better off coping with these longer real downturns by staying fully invested or by keeping a cash buffer.

• ERN points out that there have been four long (10+ years) real drawdowns in the S&P 500.
• In fact there have been more years in real drawdown than not over the past century.

But I would bet that the S&P is more prone to long drawdowns than a globally diversified multi-asset portfolio.

• And even if you can’t hold enough cash to completely protect against all eventualities, it might still be a good idea to protect against most drawdowns by holding a 10% cash buffer.
• Plus there is the psychological aspect to consider – a cash buffer makes you feel safer, and makes you less prone to do something stupid.

I do agree with ERN that a 40%+ cash buffer is not going to happen.

ERN’s third point is a combination of the first two and involves a look at real, price return only drawdowns.

• Obviously, the combination of the two is worse than either one on its own.

If you want to see the detail, ERN’s post is here.

ERN’s fourth point is that some recoveries don’t last long enough for you to replenish your cash buffer by selling stocks at high valuations.

• But again, the real question is whether the attempt produces better results on average than not keeping any cash.

ERN’s fifth point is that cash has essentially zero real return, and so all the work to make up your SWR has to come from what remains of your invested portfolio.

• With a 3% SWR and a 10% cash buffer, you are really spending 3.3% of your invested portfolio.
• I’m fairly comfortable with that, especially when you consider that you have a 3+ year backstop of cash.

ERN’s sixth point is that dividends are sometimes cut, particular in bear markets.

• I think this has already been covered under the double counting point, and anyway, that’s not a strategy that I will be following.

Looking at all of ERN’s points, I think that he has scored a couple of hits:

1. Real drawdowns last longer than people think.
2. You won’t keep enough cash to be safe under all scenarios.

But he’s ignored the psychological benefits of a cash buffer, and he hasn’t demonstrated that some cash is worse than no cash.

---

In fact, ERN has done some case studies on the impact of a cash cushion.

• He looked at a cushion of between 3.5% and 10.5% (one to three years for an SWR of 3.5%).

Whether or not the cash cushion came out ahead depended on which period of market returns he looked at.

• That suggests to me that it might be possible to modify the CAPE strategies we looked at last time to include a cash buffer that grows as stocks become expensive and shrinks as they fall in price.

ERN mentions two other approaches to be covered in later articles:

1. McClung’s Prime Harvesting uses bonds instead of cash, and draws down the bonds when equities are low.
2. A similar approach is to use a rising equity glidepath.

Prime Harvesting

Having looked at fixed withdrawal rates (posts 1 to 8), dynamic withdrawal rates (9 to 11) and cash buffers (12), article 13 looks at dynamic allocation strategies, and Prime Harvesting in particular.

• In the broad sense, a cash buffer strategy is also a dynamic allocation strategy.

Prime Harvesting (PH) comes from Michael McClung’s book Living Off Your Money, which I haven’t got around to reading yet.

The method has four rules:

1. Choose an initial stock allocation (say 60%).
2. Sell bonds to fund drawdown, until the equity allocation hits 120% of the initial setting (72% in this case).
3. When this happens, sell 20% of equities (14%) and buy bonds with the cash.
4. Go back to funding with bonds until you hit the guardrail again.

ERN likes PH because:

PH helps you when you need it the most: when the fixed withdrawal method only sustains a sub-4% withdrawal rate.

He points out that the idea isn’t new:

Wade Pfau and Kitces have a paper on this exact topic: Why you want a rising (!) equity glide path in retirement.

And we covered a similar paper in our post Eat Your Bonds First.

---

ERN also has a few quibbles with the method:

1. He likes monthly analysis rather than annual.
2. He thinks that no more than 20% of the original allocation should be sold (to take us back from 72% equities to 60%).
   • Selling 20% of the new allocation takes us below our original target.
   • It’s possible that this is what McClung meant – I haven’t read the book yet and ERN admits that he might have misunderstood.
3. ERN thinks that only selling stocks back to the upper guard rail might be better.
   • He calls this approach McClung Smooth.
   • I can see the potential advantage of more frequent smaller sales of stocks, but this sounds like you might be selling every other month at least.

ERN begins his analysis using the difficult 1966 cohort we have seen before.

• He uses a 30-year retirement horizon and a fixed SWR (with CPI indexation).
• He targets capital preservation rather than depletion.
• The initial allocation is 60/40 stocks to bonds.
• The guardrail is 120% of the initial stock allocation (CPI adjusted).

He compares five allocation rules:

1. Fixed 60/40
2. PH (here called simply McClung)
3. McClung Smooth
4. Eat your bonds first (known as forced bond liquidation)
5. Glidepath (increasing stocks gradually to 100%).

All of the other methods beat the fixed allocation SWR (with capital preservation) of 2.8%

• PH is actually second worst, at 3%
• The other three methods all came out at 3.1%.

ERN also looked at another couple of stock market peaks.

• He notes that the 2000 cohort benefited from outstanding bond returns that countered poor stock returns.
• ERN suggests that PH (or McClung Smooth) may not do so well when bond returns are poor.

Here are the combined results for all three cohorts:

ERN’s conclusion is:

The advantage of the PH method is small. But it is consistent – we never found a retirement cohort where the fixed SWR was below 4% and the PH method hurt you.

ERN prefers his McClung Smooth approach because:

For the original McClung, the final value is non‑monotone (!) in the withdrawal rate. 

ERN means that small changes in the withdrawal rate produce step changes (sometimes large ones) in the final value.

• That shouldn’t happen, so McClung Smooth is better.

Sequencing Risk

Article 14 is about Sequence of Returns Risk (sequencing risk), which we’ve looked at before:

• Pound Cost Ravaging
• Who’s Afraid Of Sequencing Risk?
• Who Ate Joe’s Pension?

Sequencing Risk (which we’ll shorten to SoRR) is an extension of the simulation analysis that ERN has been doing to analyse withdrawal strategies:

1. Different retirement cohorts have different outcomes with the same initial pot size and drawdown strategy.
2. That’s because investment returns are not entirely predictable.
3. But they are not entirely unpredictable, either – that’s why the CAPE-based strategies do well.

SoRR is the important additional finding that even if two cohorts were to have the same set of annual / monthly investment returns, their final outcomes would be affected by the order in which they experienced them.

• In a nutshell, a sequence of bad returns at the start of retirement will deplete your pot so badly that you can’t recover.

---

For a buy-and-hold investor, SoRR doesn’t apply.

• Geometric growth rates don’t depend on order (they are commutative).

But when you make withdrawals from the portfolio (and indeed, when you make contributions during accumulation) the order does matter.

• It’s a weight of money argument – the bigger your pot at the time, the greater the impact the prevailing return will have.

Alternatively, you could look at it as the negative effect of withdrawing money at the bottom of the market (after an early fall).

• The same logic explains why the average investors in a fund will  underperform the fund itself – more money comes in after a good run, and goes out after a bad run.

The cash flows of the saver and retiree add up to exactly the Buy and Hold strategy. When the retiree’s IRR lags Buy and Hold, the saver’s IRR must beat it. There is a zero-sum game between savers and retirees! When the retirees suffer from SRR, savers will benefit.

Another way of looking at this is that a cohort that retires around a market peak will have accumulated larger than expected pots.

• They started saving under a low return regime, then when their pot was bigger, they benefited from higher returns.

But then in retirement they can expect to do less well, with low returns in the early years impacting their final pots.

• This is another reason why millennials shouldn’t be so jealous of boomers retiring at the moment – yes they have bigger pension pots, but they will probably need them.

As Solon put it, a long time ago:

Call no man happy until he is dead.

Low Returns

Article fifteen continues ERN’s look at SoRR.

• He says that low average returns are less of a problem in retirement than SoRR.

The table above looks at three case studies, using SWRs that exhaust the pot over 30 years:

1. 1968 has average returns of 6.16%, yet an SWR of only 3.8%
2. 1979 has returns of 6.03%, but an SWR or 9.12%
3. 1955 has returns of only 3.45%, but an SWR of 5.72%

The correlation between average returns and 30-year SWR is not great.

• In fact, 30-year average returns only predict 31% of the variance in the SWR.

But if you know the average returns over the six 5-year periods within the 30 years, you can account for 96% of the variation in the SWR.

• Sequencing is key, explaining another 64% of the SWR variance.

---

ERN notes that the worst SoRR scenario is to retire half-way through a recession / market correction.

• Here you will be hit for the last few (and biggest pot) years of accumulation.
• And then you will be hit again in the first few years of retirement.

The 1959 cohort is a good example:

• Sequencing reduces their final pot size by almost 50%.

---

ERN next looks at a couple of ways to deal with SoRR.

VPW doesn’t work:

If you try to equalize the final portfolio value through VPW then SRR hits you through the withdrawal amounts! If you try to equalize the withdrawal amounts then SRR hits you in the final portfolio value. Pick your poison! 

One thing that does work is “lifecycle investing”:

Borrow against your future retirement account contributions and invest the whole loot as one big lump-sum payment without further contributions in the future (i.e., use your future retirement contributions to pay down the margin loan).

This is an interesting (and radical) concept that I will return to in a future post (outside the ERN series).

It’s too risky for ERN, but he recommends a mitigation strategy:

• Hold a 100% equity portfolio early on..
• Don’t bother about paying down low-interest debt when young (mortgage, low-interest student loans). Put every last dollar you can scrape together into the stock market.
• Once you have a critical mass of equity investments, then tackle your low-interest debt.

That’s a kind of watered down version of lifecycle investing:

You gain from the higher expected return in equities over fixed income investments and  you alleviate the SRR by spreading around the equity risk more evenly across time.

Conclusions

It’s been another great set of posts from ERN, and we’ve managed to squeeze four in this time.

Let’s recap:

1. ERN doesn’t like cash buffers
   • In real terms, drawdowns last a long time, and you can’t afford to keep a large enough cash buffer to ride them out completely.
   • I think that cash helps psychologically in a market crash, even though the buffer is as likely to impact overall returns as to enhance them.
   • I also think that drawdowns in a globally-diversified multi-asset portfolio will be shorter than those in the S&P 500.
   • I plan to keep a 5% to 10% cash buffer under all foreseeable circumstances.
   • I would be interested to see any CAPE-based strategies for a variable-sized cash buffer.
2. ERN does like the Prime Harvesting (PH) dynamic asset allocation strategy.
   • But he prefers his own variation – McClung Smooth.
   • He also likes a rising equity glidepath and the Eat Your Bonds First strategy.
3. Sequencing risk (SoRR) means that cohorts with equal average returns will have different outcomes.
4. SoRR is more important than low average returns.
   • Average returns explain 31% of the variation in 30-year depletion SWRs.
   • Sequencing explains another 64%.
5. If you are unlucky, you can have have a SoRR double whammy – poor returns in the last years of accumulation and the first years of retirement.
6. Those retiring with larger than expected pots (boomers) will need them, as SoRR is more likely to affect them in retirement.
7. You can protect against SoRR by using lifecycle investing.
   • This is risky, and ERN prefers a watered-down version where young people run their low-cost debt (mortgages and student loans) in order to put more into equities.
   • When you have enough equities, start paying off the loans.

That’s it for today.

Until next time.

Share this with Twitter, Google+, Pinterest, LinkedIn, Tumblr, Reddit and StumbleUpon.