Cash buffer vs. rebalancing

[rayvt]

Since most money that is lost in a down market (and it can be substantial) is due to what could be called "psychological" reasons it only makes good sense to have a 2-4 year cash buffer to last through most bear markets.

In other words "studies" are great but they don't mimic the real world IMO.

If it, as you say, makes good sense then the historical data should show that. Problem is, the historical data doesn't show that at all.

The problem with a 2-4 year cash buffer is twofold:
1) On the 3'rd or 5'th year of said bear market, you are out of cash, deep in the bear, and have to sell stocks when they are way down.
2) Everybody ignores the question of how & when to refill the cash buffer after yo have drawn it down.

Here's a spreadsheet that models a portfolio with a cash buffer, using actual real-world historical data. You can changes all sorts of parameters like # of years in cash, AA, refill method, etc. https://www.dropbox.com/s/xf4ma5blug27aws/SPY_Withdraw_by_CashBucket_rules.xls

I haven't been able to find one set of parameters that come out better than just maintaining the chosen asset allocation.

 

[audreyh1]

If it, as you say, makes good sense then the historical data should show that. Problem is, the historical data doesn't show that at all.

The problem with a 2-4 year cash buffer is twofold:
1) On the 3'rd or 5'th year of said bear market, you are out of cash, deep in the bear, and have to sell stocks when they are way down.
2) Everybody ignores the question of how & when to refill the cash buffer after yo have drawn it down.

Here's a spreadsheet that models a portfolio with a cash buffer, using actual real-world historical data. You can changes all sorts of parameters like # of years in cash, AA, refill method, etc. https://www.dropbox.com/s/xf4ma5blug27aws/SPY_Withdraw_by_CashBucket_rules.xls

I haven't been able to find one set of parameters that come out better than just maintaining the chosen asset allocation.

We have historical data on how markets and investments behave. But that is not the issue. We don’t really have much info on how people actually behave in downmarkets other than investors don’t seem to do as well as the investments they hold for numerous reasons.

 

[OldShooter]

... I haven't been able to find one set of parameters that come out better than just maintaining the chosen asset allocation.

I think you have posted this link before; I recall looking at it briefly.

I didn't and don't intend to spend much time with it because the model appears to have some significant flaws, for example:

While the cash bucket is an asset your model appears to isolate it and not consider it as part of the AA. It is part of the AA as are all investable assets. This fallacy (IMO) drives your conclusions in a fairly major way, I think.

Along those lines, after x years of "cash" have been set aside, to use a 60/40 AA for the long-term money is extremely conservative. I can't see where this is an input parameter. It may be but regardless, your "Readme" conclusions are based on 60/40.

I can't tell for sure, but it appears that your cash bucket is not generating any returns. This is unrealistic. No intelligent investor would tie up years' worth of cash without any return. Actually, since the so-called "cash" is simply part of the fixed income portion of the portfolio, it will be invested according to the owner's risk tolerance and will yield accordingly.

The replenishment strategies are interesting but also unrealistic. Assuming a "bad" case where the investor needs cash before he/she feels that the market has adequately recovered, he/she will almost certainly do only a partial refill. This is hard to model, I get that, but it is real world.

I have no idea what your simplifying assumption of having investments only in the S&P 500 and in 10-year notes (not bills, FYI) might mean for the model. That is certainly a feasible portfolio but hardly a typical one. A reliable model needs to have been tested with several portfolios, including some with TIPS, to see how its results are affected, if at all.

Your time scale is 55 years. That's unrealistic, of course, makes your graphs irrelevant to the question. A model like this needs to be run and re-run for more realistic periods like 20 or 30 years with different start points. For example, run it for each 30 year period that your data covers beginning: 1960, 1961 ... 1984 -- 24 runs. Among other things, this will help tease out the effect, if any, of the high inflation around 1980 +/-.
​

Models are fun but the temptation is to stop modeling when they tell us what we wanted to hear. BTDT. But I really don't think this model, as it stands, "proves" anything.

(I also get that there is a lot of work here. Whew! But that doesn't mean that there aren't serious flaws in the model.)

 

[FinancialDave]

If it, as you say, makes good sense then the historical data should show that. Problem is, the historical data doesn't show that at all.

The problem with a 2-4 year cash buffer is twofold:
1) On the 3'rd or 5'th year of said bear market, you are out of cash, deep in the bear, and have to sell stocks when they are way down.
2) Everybody ignores the question of how & when to refill the cash buffer after yo have drawn it down.

Here's a spreadsheet that models a portfolio with a cash buffer, using actual real-world historical data. You can changes all sorts of parameters like # of years in cash, AA, refill method, etc. https://www.dropbox.com/s/xf4ma5blug27aws/SPY_Withdraw_by_CashBucket_rules.xls

I haven't been able to find one set of parameters that come out better than just maintaining the chosen asset allocation.

I am just a little confused by your comment above UNLESS you are expecting that a person is going to have a cash buffer WHILE they are working. You only apply the cash bucket in the 5 years leading up to your retirement. There is no way a person is going to be better off with no cash bucket after 5 years of a bear market because they would be selling off assets at a loss for the whole 5 years. So in fact it doesn't matter whether you refill the cash bucket or not, you are better off not having to sell your assets for the whole 5 years.

In fact when I apply this strategy about half the cash bucket is future dividend payments that are not reinvested but go directly to the cash bucket so you are not sitting with 5 years of cash, it is more like 2.

Also, in the real world, peoples retirement budget is full of at least some expenses that can be cut in times of need. People just don't blindly spend a lot of unnecessary money if their account is down by a large amount.

I would also like to know when you last saw a 5 year bear market where the market was down greater than 30% for the whole 5 years.

 

[youbet]

Cash is king in a crunch. You cannot allocate into a crunch without cash. That is how Buffett gets his bargains. The cash drags his balance sheet for years or decades. And then it doesn't.

I think you’re mixing “trading cash” with “cash reserves for spending.” Two very different things.

 

[OldShooter]

... Also, in the real world, peoples retirement budget is full of at least some expenses that can be cut in times of need. People just don't blindly spend a lot of unnecessary money if their account is down by a large amount. ...

Of course. I deliberately didn't get into that in my post a couple above, but IMO this whole idea of a level withdrawal rate is completely fictional. Not only can spending be managed to a significant degree, the need for spending can also vary quite a bit. Need to replace a car? How about a couple of special vacations in one year? Remodel the kitchen? Capital campaign at your church? Sell the house and move to a progressive care facility? Not only do all of these things make withdrawal needs lumpy, many of the lumps are in categories that can be managed -- as you point out.

I look at our withdrawal in a year like I look at a speedometer. Worth checking but not anything to get excited about unless we seem to be going dangerously fast for no good reason.

 

[FinancialDave]

I think you have posted this link before; I recall looking at it briefly.

I didn't and don't intend to spend much time with it because the model appears to have some significant flaws, for example:

While the cash bucket is an asset your model appears to isolate it and not consider it as part of the AA. It is part of the AA as are all investable assets. This fallacy (IMO) drives your conclusions in a fairly major way, I think.

Along those lines, after x years of "cash" have been set aside, to use a 60/40 AA for the long-term money is extremely conservative. I can't see where this is an input parameter. It may be but regardless, your "Readme" conclusions are based on 60/40.

I can't tell for sure, but it appears that your cash bucket is not generating any returns. This is unrealistic. No intelligent investor would tie up years' worth of cash without any return. Actually, since the so-called "cash" is simply part of the fixed income portion of the portfolio, it will be invested according to the owner's risk tolerance and will yield accordingly.

The replenishment strategies are interesting but also unrealistic. Assuming a "bad" case where the investor needs cash before he/she feels that the market has adequately recovered, he/she will almost certainly do only a partial refill. This is hard to model, I get that, but it is real world.

​

You make a lot of good points here - better than I could.

This issue that you are alluding to is that a cash bucket is in reality nothing more than a "bond" allocation in your portfolio.

In general you hold bonds in your portfolio as a method of risk management not as an item that will generate excess returns for you, because they won't over long periods of time.

I would actually even submit that holding cash in say a Vanguard Money Market Fund paying just over 2% right now is a safer bet than holding your funds in a short-term bond fund during retirement facing a rising rate environment where the price of your bonds can drop (and just did recently as a matter of fact.)

Like you say, what many call a cash bucket, is just a larger asset allocation to fixed income (bonds and cash) and certainly if you don't get rid of the same amount of bonds, adding cash will do nothing but lower your risk and your returns long-term.

My comments about using a cash bucket are related to a bunch of work done by James Cloonan, the founder of AAII. Besides the cash bucket the rest of the asset allocation is 100% in equities.


Dave​

 

[OldShooter]

... what many call a cash bucket, is just a larger asset allocation to fixed income (bonds and cash) ...

Well, and maybe not "larger." This is all just mental accounting anyway. In our case, having been very fortunate in life, our fixed income tranche is probably more money than we will ever need. And, actually, as we did our most recent portfolio review we decided to reduce it as a % in the AA because a bigger number was unnecessary. Inside that tranche we see a bucket that we will draw from and some TIPs that may eventually be liquidated into the bucket. Alternatively, some day we might liquidate some equities into the bucket and continue to hold the TIPS. In liquidating equities we would inevitably (gasp!) be doing some mild market timing.

... My comments about using a cash bucket are related to a bunch of work done by James Cloonan, the founder of AAII. Besides the cash bucket the rest of the asset allocation is 100% in equities. ...

Yeah. Actually it seems that most of these anti-bucket rants (including Kitces IIRC) get very mixed up about AA. AA is more like an output of a bucket strategy than it is an input. And our long-term bucket is also 100% equities. Basically it is funding for our sons' trusts and for some charitable bequests so its time horizon is a long one. Longer than ours.

 

[pb4uski]

.... Here's a spreadsheet that models a portfolio with a cash buffer, using actual real-world historical data. You can changes all sorts of parameters like # of years in cash, AA, refill method, etc. https://www.dropbox.com/s/xf4ma5blug27aws/SPY_Withdraw_by_CashBucket_rules.xls

I haven't been able to find one set of parameters that come out better than just maintaining the chosen asset allocation.

I can't really follow your spreadsheet, but what about a decision rule that says rebalance only if equities are above target (60% as you define it), otherwise spend from cash or bonds.

 

[rayvt]

** what about a decision rule that says rebalance only if equities are above target (60% as you define it), otherwise spend from cash or bonds.

The spreadsheet models a portfolio where the stock/bond asset allocation is rebalanced every month.

It is quite possible to question the assumptions & methods of the spreadsheet. But keep in mind the reason I did it in the first place. I had read many articles about the benefits of having a cash bucket, but the details of how to manage it were always vague or omitted. Christine somebody at Morningstar wrote a few articles with examples and I just had a hard time figuring out how it would work---and the EXACT steps I would have to take to implement her advice. Jean-Luc Picard can wave his hands and say "Make it so.", but my bank and broker require specific instructions on moving money.

So I build a spreadsheet. Tried to do my best to model the same way that the cash bucket articles say----take your entire amount of money, put X years of living expenses into a cash account, and put the remainder in a balanced stock/bond investment portfolio.

When it comes time to make a withdrawal, take money from the investment gains first and any shortfall you pull from the cash account. In a bull market, the gains are more than your scheduled withdrawal, so you don't touch the cash account. In a bear market there are no gains, so you take the entire withdrawal from the cash account.

** your model appears to isolate it and not consider it as part of the AA.

Correct. See above. Everybody handles the cash bucket as a source of funds for withdrawals when the portfolio doesn't have adequate gains. Cash bucket is "X years worth of withdrawals" and is independent of how you handle the stock/bond investment account.

** Models are fun but the temptation is to stop modeling when they tell us what we wanted to hear.

I didn't do the model to "prove" anything. I did it to see how the cash bucket strategy would have worked---using actual historical data. Kitces says cash bucket is a mirage. Christine says cash bucket is great. Both are fuzzy on details. I wanted to write down the rules for how to manage the cash bucket method and to see how it worked.


** to use a 60/40 AA for the long-term money is extremely conservative.

I agree. The standard AA that is always talked about is 60/40, which is why I used it as the default.
You can change it in the spreadsheet. Main tab, L7 thru L8.

** I can't tell for sure, but it appears that your cash bucket is not generating any returns.
No, it earns the 1 yr T-bill rate.

** The replenishment strategies are interesting but also unrealistic.

They are all strategies that people have proposed. The most common suggestion is "When the market recovers". Easy to say, but see the Picard quote above.

** Assuming a "bad" case where the investor needs cash before he/she feels that the market has adequately recovered, he/she will almost certainly do only a partial refill. This is hard to model,

Yeah, I can't figure out how to model "Investor does whatever he feels like doing at the moment."

** I have no idea what your simplifying assumption of having investments only in the S&P 500 and in 10-year notes (not bills, FYI) might mean for the model. That is certainly a feasible portfolio but hardly a typical one. A reliable model needs to have been tested with several portfolios,

There are an infinite number of feasible portfolio investments and allocations. Can't test them all. Makes most sense to test with a standard go-to AA. You don't need to inspect every board in your house for termites once you have found them in several joists.

And, really, the point is not to try to find some arcane portfolio that a cash bucket works for. The point is to see if a cash bucket saves your bacon in bear markets. If it doesn't do so with a standard plain-vanilla portfolio then it doesn't do what it is supposed to do.

** A model like this needs to be run and re-run for more realistic periods like 20 or 30 years with different start points.

Easy enough to do. Pick your start year and look only at the first 20 or 30 years.
Ex: Start 1980, $100K, 4% SWR, 3 years in cash bucket. Col Z is value without cash bucket, col C is with cash bucket (total value is investments value + cash bucket).

Better yet
Ex: start in 1973, stepping right into a bear market.
This is EXACTLY the scenario where cash bucket is supposed to save you.
And it fails miserably.
Total value at 30 years using 3 year cash bucket is $156K.
Total value at 30 years without cash bucket is $230K.

In most cases I looked at, the cash bucket method fell behind. I didn't find one case where the cash bucket was ahead.

Opposite example, starting off into a bull market.
Ex: Start in 1975.
You didn't need to be saved.
Total 30 year value: $1.1M with cash bucket. $1.7M without.

 

[rayvt]

Oh, one thing.

If you have another refill rule that you'd like to propose, and if it clear enough that it can be written down with a mechanical rule, let me know and I'll add it to the spreadsheet.

I tried to cover the most common suggested methods, but who knows, maybe I missed something.

 

[pb4uski]

Oh, one thing.

If you have another refill rule that you'd like to propose, and if it clear enough that it can be written down with a mechanical rule, let me know and I'll add it to the spreadsheet. ...

At the end of each year if stocks exceed target then sell stocks down to target... if stocks are less than target, then stand pat... let cash absorb any withdrawal and if cash is depleted then take from bonds.

IOW, if stocks are bearish, then use cash first and then bonds if cash is depleted.... when stocks recover, rebalance and add proceeds to bonds up to target and then to cash. So cash and then bonds are a buffer for any stock underperformance.

I "think" this is the conventional buckets approach as I understand it (though how buckets get refilled has never been well described in anything that I have read on buckets).

 

[pb4uski]

Oh, one thing.

If you have another refill rule that you'd like to propose, and if it clear enough that it can be written down with a mechanical rule, let me know and I'll add it to the spreadsheet. ...

At the end of each year if stocks exceed target then sell stocks down to target... with proceeds going to bonds, which in turn are rebalanced if they exceed target with proceeds going to cash.

If stocks are less than target, then stand pat... withdrawals come from cash first and then bonds if cash is depleted.

IOW, cash and bonds are a buffer to stock underperformance and get refilled only when stocks exceed target... and within the buffer withdrawals are first from cash and secondarily from bonds.

I "think" this is the conventional buckets approach as I understand it (though how buckets get refilled has never been well described in anything that I have read on buckets).

 

[OldShooter]

... And it fails miserably. ...

@rayvt, let me make my point a little more concisely:

Your model is nowhere near robust enough to draw any general conclusions about a bucket strategy. It has baked-in assumptions that pretty much guarantee that the spacific bucket strategy that you are modeling will fail, specifically the failure to invest the near bucket sensibly and the ultraconservative strategy for the far bucket. Probably you are just concluding that a more conservative AA will underperform an aggressive one.

Even with an adequately robust model, it is not enough to poke around and look at various 20- and 30-year runs. It is necessary to run all fifty of them and then look at the big picture. (Even this would amuse the hard-core statisticians because there is only 50 years of data here. But that is another thread.)

You have a lot of work in this model. I have done that kind of thing and I understand. But to build an adequately robust model, one that would adequately compare the strategies, is probably only within the scope of a graduate student in economics or an organization that can come up with funding for a man-year or more of effort. So ... While I admire your effort, I don't think that the result is adequate to reach any useful conclusions.

 

[Huston55]

An Interesting (But Incomplete) Study

Some thoughts about this paper:

1. No behavioral component to the paper:. For me, this is a fatal flaw. If a “cash” (narrowly defined in this paper as MMF) buffer positively impacts investing behavior (which BTW, although anecdotal, it seems to for many posters here), then it’s achieved it’s goal. Two items of support for that line of thought:

Morgan Housel’s paper on the Psychology of Money: excerpt, “...investing is not the study of finance. It’s the study of how people behave with money.”

http://www.early-retirement.org/forums/f28/the-psychology-of-money-93769.html

This Gallup poll seems to clearly show that the behavioral side of investing is very important.

https://news.gallup.com/poll/211052/stock-ownership-down-among-older-higher-income.aspx

2. What’s “cash?” Does anyone really put 4yrs in MMF?:. As discussed in many posts above, intermediate CDs (which I’ve called ‘cash’) are actually FI IAW the OP paper. If I count CD & ST Bonds as “cash”, I have 4 yrs. If I count only MMF (the paper’s definition), I have 4 mos. Big difference.

The most useful thing I got from this paper is that I now have another way to think about/define “cash.”

 

[rayvt]

** specifically the failure to invest the near bucket sensibly

What would define "sensibly"? All the articles on cash bucket method say this needs to be safe and liquid and the yield is less important. What's wrong with 1-yr T-bills?
I also tried using the 3-yr T-bill rate. Hardly any difference. Just to go crazy, I also tried 10-yr treasuries. Ditto.


** and the ultraconservative strategy for the far bucket.

The 60/40 AA is not generally considered to be conservative, let alone ultraconservative. But that's just the default value in the spreadsheet. It lets you go all the way to 100/0.


** Even with an adequately robust model, it is not enough to poke around and look at various 20- and 30-year runs. It is necessary to run all fifty of them and then look at the big picture.

Actually, no it isn't necessary. Remember, I wasn't trying to prove anything. I was trying to see how this method worked in the circumstance where it was supposed to save your bacon. Just like disproving a mathematical theory, all that is needed to disprove it is to find ONE case where it fails.

See, you don't need to look at the big picture or look at fifty 20- & 30- years runs. All you need to do is find ONE run that starts in a bad year and the strategy failed at its stated purpose. You don't care how it works when it doesn't need to work---you care how it works when it _needs_ to work.

-------------

Ah, Benz. Christine Benz was the name I couldn't remember. I found this recent paper of hers. https://www.aaii.com/journal/article/for-bucket-portfolios-the-devil-is-in-the-details Also this: https://www.kiplinger.com/article/r...mplement-the-bucket-system-in-retirement.html

You know what I don't see? A _specific_ list of steps to take to implement the strategy.
What I do see is a bunch of handwaving. Surely if this method is so good, somebody would have made the effort to set up a backtest that showed how it worked, particularly at a period with sequence risk, beginning right before a bear market. Yet I am not aware of such a backtest or even a multi-year example. It's not really that hard. My first cut at it took less than a day. The refinements took longer, of course, but didn't change the basic conclusion.

Christine Benz has been writing about it for years. She is "director of personal finance for Morningstar", so she has access to resources that could backtest it. Certainly stronger resources than one guy on the internet whacking up Excel spreadsheets.

The other thing that I don't see in their discussions: They talk about 3 buckets, soon (1-3 years), later (3-10 yrs) and long-term (11+ years). But they never talk about the passage of time. They never talk about money flow from the longer buckets to the nearer buckets as the years pass. Are the timespans of these buckets fixed at the initial date? Or are they for the N years from now, whenever "now" happens to be? "Today is the first day of the rest of your life." In 11 years, you will still have to manage the money you'll need in the next 1-3 years, the next 3-10 years, and the 11+ years.

I really would like one of these people to show the buckets as of today and as of 15 years from today. And then I'd like to see them to show how they got from here to there.

 

[OldShooter]

... I wasn't trying to prove anything. ...

Got it.

 

[indiajust]

A paraphrase from A Wealth of Common Sense: Consider how many folks save little or nothing. Just having enough for retirement is well above average. Just having an allocation across your net worth is another leap ahead of the pack. And having a process for reallocation is an additional step ahead


Pareto's law, 80/20, haggling over the specifics of the allocation is deep in the weeds of sub 1% of the humans. Interesting, but not likely to move the performance needle.

 

[TwoByFour]

To me, a cash bucket is an asset pool that is outside of one's AA, and cash as part of one's fixed-income portion is inside of one's AA. If you use a cash bucket and it is small, like 1%, it probably makes no difference all in all. But if that bucket is big, like > 5%, it does make a difference. But sleeping well at night trumps everything, and if a cash bucket does that, so be it.

My model for management of cash flow is a waterfall, in which equities are at the top, then bonds, and cash at the bottom. Everything flows down hill:
* If equities are > 60%, sell the excess and buy bonds.
* If cash is < 3%, then sell bonds to bring it up to 3%.
* All expenditures - planned withdrawals and unplanned expenses, come out of the 3% cash portion of fixed income.

Notice that there is no step to buy equities so this is not true rebalancing. Most of the bonds are a bond-ladder but there is a small bond fund I use to buffer the ladder.

 

[jimbee]

A cash bucket outside of your asset allocation is fooling yourself, in my opinion.

 

[audreyh1]

A cash bucket outside of your asset allocation is fooling yourself, in my opinion.

It might be helpful to explain what you think the fallacy is.

 

[Dtail]

A cash bucket outside of your asset allocation is fooling yourself, in my opinion.

As @audreyh1 and others have explained enough times in the past, not everyone including myself wish to subject certain cash funds or portions of their emergency funds to market fluctuations. Thus they are not part of the AA.

This doesn't mean that this concept isn't taken into account when incorporating one's AA%.

 

[jimbee]

Just because it is cash doesn't mean it isn't part of your asset allocation. You don't have to subject it to market fluctuations, it's cash. Not including it is ignoring math. It's similar to not including house equity as part of net worth.

 

[Fedup]

I think I fail on both counts.

 

[Dtail]

Just because it is cash doesn't mean it isn't part of your asset allocation. You don't have to subject it to market fluctuations, it's cash. Not including it is ignoring math. It's similar to not including house equity as part of net worth.

Not saying that. I also have cash which is part of my investment portfolio and thus part of my AA%.
However, some of us for example take a set % of the remaining portfolio for the year. Let's say it is 3%, but we only use 2.5%, the remaining 0.5% not used goes into a cash type account which is not subject to future stock market movements.
If there is a bad year or let's say a one time large expense, it might come out of this account.
Just a different way of accounting for assets.