buckets in FIRECalc ?

[JohnEyles]

Posting here because dory36 doesn't seem involved in the thread
"Ray Lucia...Buckets of Money" in the "FIRE and Money" forum.

In page 5 of that thread, I make a plea for adding support for
the "buckets of money" theory to FIRECalc. Seems like a good
idea, and if so, FIRECalc should show better SWRs.

Please ...

[dory36]

Can you send a summary of the process? I didn't see it in the thread, and his website wanted me to jump through more hoops whan I was willing to, just to see what he is talking about.

[Rich_in_Tampa]

Dory,

Fore firecalc purposes, if you could just allow assets to be allocated into three presumed "rate of return" categories, that would do it.

Example:

Bucket 1 - enter amount, and enter rate of return expected (e.g. 4%)
Bucket 2 - same; user might enter 6%
Bucket 3 - same; user might enter 10%

Then look at the aggregate performance of all three (weighted by their relative balances) in doing the projections and calculations, retain the inflation factor as it is now and voila.

I think that calculating the amount in each bucket is beyone the scope of FC and already is available on Lucia's site.

You can almost do pieces of this already in the "how is it allocated" tab but you can't quite combine them in the above manner.

Happy to beta this for you. I think it's a useful request.

[brewer12345]

Rich, if you start assuming things aboutthe rate of return you A) negate the very basis of FIRECalc, which uses historical simulation rather than simple assumptions about returns, and B) reduce it to something you could do pretty easily on an excel spreadsheet.

[Rich_in_Tampa]

Originally Posted by brewer12345

Rich, if you start assuming things aboutthe rate of return you A) negate the very basis of FIRECalc, which uses historical simulation rather than simple assumptions about returns, and B) reduce it to something you could do pretty easily on an excel spreadsheet.

But it's impossible to use historic data for buckest which are custom-selected by each user.

Generally, bucket 1 approaches MMF or short-term bonds (maybe 4%), bucket 2 mimics intermediate bonds or maybe a bond-oriented blend, and bucket 3 is the standard sliced and diced stock only allocation, say 10%.

No different from Dory's third choice in "How is it invested." All user-defined -- already overrides historic data.

This would just be a useful additional tool. Agree it can be replicated on a spreadsheet.

[dory36]

Originally Posted by Rich_in_Tampa

Happy to beta this for you. I think it's a useful request.

Originally Posted by Rich_in_Tampa

Agree it can be replicated on a spreadsheet.

OK, send me a basic spreadsheet and I'll look at it as an add-on.

[JohnEyles]

Originally Posted by Rich_in_Tampa

But it's impossible to use historic data for buckest which are custom-selected by each user.

Generally, bucket 1 approaches MMF or short-term bonds (maybe 4%), bucket 2 mimics intermediate bonds or maybe a bond-oriented blend, and bucket 3 is the standard sliced and diced stock only allocation, say 10%.

No different from Dory's third choice in "How is it invested." All user-defined -- already overrides historic data.

Yes, but for the "mixed portfolio" choice, it DOES use historic data for each of the
various asset classes, no ? So all that is needed is to provide this set of choices
for EACH of the 3 buckets (it could be pruned down a little, e.g. only a moron
would put treasuries in Bucket 3 or small-cap in Bucket 1).

Seems to me, the bigger problem is how does the rebalancing (i.e. funds trickling
to lower-numbered buckets over time) occur ? As I said in the Buckets thread,
I think the essential point of buckets is that the rebalancing occurs
opportunistically rather than automatically (that's the whole reason for having
Bucket 2 as a buffer). Doing it automatically would, I believe, give very similar
results (to simply doing an annual rebalance) and thus be pointless.

We must define heuristics that specify how money trickles down from 3 to 2,
and from 2 to 1. Perhaps, you sell enough of 3 to replenish 2, if and only if
the aggregate performance of 2 (since retirement age) is above some APR ?
The rule for moving 2 to 1 could be simpler I think, striking some balance
between keeping 1 funded and not depleting 2 too badly.

As usual, the devil is in the details. Dory, you ROCK for considering this.
I would be willing to try to write some code, or psuedo-code, to define
the heuristics; what programming environment do you use ?

[sgeeeee]

It's been awhile since I considered the 3 bucket approach, but it seems to me that if all you do is consider the average returns of three buckets of money with annual rebalancing, the simulation won't be very interesting. You will simply get a weighted average return of the 3 buckets and may as well simulate it that way -- using a spreadsheet.

The 3 bucket approach only differs significantly from an asset allocation & periodic rebalance plan if you use a different spending model -- reserving the right to leave your buckets unbalanced for multiple rebalance periods based on market performance.

A bond ladder approach is similarly only distinct from an asset allocation approach if you reserve the right not to rebalance or replenish the bonds in years when the market is down.

To do a meaningful simulation, you need to come up with rules that drive the modified spending model. When will you rebalance? What bucket will you spend from? How do you quantify the circumstances when the answers to these questions change?

[dory36]

Start maybe with just a list of actions each year. Or a basic spreadsheet.

[donheff]

Originally Posted by sgeeeee

The 3 bucket approach only differs significantly from an asset allocation & periodic rebalance plan if you use a different spending model -- reserving the right to leave your buckets unbalanced for multiple rebalance periods based on market performance.


To do a meaningful simulation, you need to come up with rules that drive the modified spending model. When will you rebalance? What bucket will you spend from? How do you quantify the circumstances when the answers to these questions change?

Yup, you need an algorithm. You could think along the lines of rules that would have you A) always spend from bucket 1 if funds are in it; B) spend from bucket 2 if bucket 1 is empty;

The drip algorithm gets very complicated because it will depend on the levels of buckets 2 and 1 and the condition of 3. E.g. C) drip from bucket 3 to bucket 2 when bucket 2 is not full and bucket 3 theoretical ROI (i.e. without respect to forced withdrawals) is positive historically?; D) drip from bucket 2 to bucket 1 when ?; D) pour (i.e. cover all expenses and x replenishment) from bucket 3 when 2 and 1 are empty. ARRRGH

Can this be sensibly modeled? Someone would need to do a PHD dissertation to come up with a rule set. It seems like bucket replenishment will involve an intuitive sense of touch. I am going back to Clagett's 95% rule.

[sgeeeee]

You could start with a 2 bucket model (ie bond ladder):
- bucket 1 = short-term fixed income investments (bonds, CDs etc)
- bucket 2 = remainder of balanced portfolio

Bucket 1 should have a value equivalent to 5 years of spending requirements (including some inflation consideration). You can choose 7 years or 10 years or whatever makes sense, but I'll use 5 here for illustration.

Bucket 2 should be an equity/bond mix. The total mix of Bucket 1 plus Bucket 2 should provide and equity/bond ratio consistent with your individual risk profile. Each year when rebalance occurs, it is the overall equity/bond raito that needs to be preserved.

Here's the spending rebalance model:

Each year annual cost of living comes from bucket 1 as long as bucket 1 has enough money to cover those expenses. Any earnings from bucket 2 go to replenish bucket 1 up to the amount of inflation adjusted spending needed 5 years from now.
- If bucket 2 produced more earnings than is needed for replenishing the bond ladder, that money is rebalanced within bucket 2 but it is the total portfolio equity/bond ratio that is preserved.
- If bucket 2 produced earnings, but less than is needed to replenish bucket 1, the earnings go into bucket 1, but the bucket 1 value is reduced to less than 5 years worth of living expenses. bucket 2 is rebalanced to achieve overall portfolio risk.
- If bucket 2 produces losses, no money is shifted between bucket 2 and bucket 1 and bucket 2 is rebalanced to achieve overall equity/bond ratio.

But this approach gets even more complicated. If bucket two returns are insufficient to replenish bucket 1 for several years in a row, bucket 1 becomes very small. Each year with positive bucket 2 earnings then has to consider replenishing as much of bucket 1 as possible -- multiple years worth of living expenses. If bucket 1 is reduced to less than one year's expenses, the annual spending must come out of bucket 2 until enough years of positive return allow bucket 1 to be replenished.

Now . . . add another bucket, develop rules that apply to all three buckets, and . . . here's the good part . . . develop a program that follows these rules.

[JohnEyles]

Originally Posted by dory36

Start maybe with just a list of actions each year. Or a basic spreadsheet.

Ok...

For the inputs, I think the only change needed is in the "How Is It Invested"
tab, where you add a new choice called "Buckets", subdivided into 1/2/3, with a
dollar figure for 1 and 2 (and the remainder in 3). Then, for each bucket, there
is a list of asset classes like those in the "mixed portfolio" choice. (To save space,
omit stock classes from Bucket 1, omit fixed income from Bucket 3, etc). It's the
user's job to decide the starting dollar figure for 1 and 2, based on number of years
income you want to have in each, what growth you want to assume for 2, etc.

For annual actions, first, the growth. For each bucket, compute the growth for
each asset class using the historical data for that asset class. Do NOT rebalance
(within) the bucket at this point.

Next, the spending. Spend from 1, 2, and 3 in that order, depleting each before
moving to the next. Withdraw according to the asset allocation AFTER growth,
before any rebalancing. When they're all empty, it's a "failure".

Finally, the tough part, the rebalancing between buckets. The goal is to make
1 and 2 "full" again, but this may not always be optimal (see below). But what
does "full" mean - perhaps the original dollar value grown by inflation ?

First, you completely replenish-to-"full" 1 from 2 (withdrawing from 2 in
proportion to its allocation AFTER its growth computed earlier in the annual
actions, before any rebalancing within 2). This may empty 2 more than
halfway, we'll deal with that later.

Then you have to decide how much of 2 to replenish from 3 - striking some
balance between the level of depletion of 2 (easy to quantify) and the perceived
value of 3 (a little trickier, perhaps the overall return-since-retirement and/or
the previous year's return). Maybe you bring 2 up to 50% of "full", if 3 has
simply beat inflation since the previous year. Then you bring 2 up to 75%
if 3 has had overall return since retirement which beats inflation. And you fill
2 completely if the overall return-since-retirement of 3 has been higher than
X%. Again, you withdraw from 3 in proportion to its allocation AFTER growth.

Finally, if 2 is STILL below 50% (after replenishing 1 and rebalancing from 3).
move enough money from 1 back to 2 to either bring 2 to 50% or to equalize
their amounts, whichever comes first.

Again, all moves OUT of buckets were done in proportion to the asset mix AFTER
growth was computed (so the more appreciated asset classes are harvested more),
and moves INTO buckets were done into cash positions. Finally, after all inter-bucket
movement is done, you rebalance WITHIN each bucket to its original asset allocation.

Hope this helps. Chime in folks. Sounds like Dory will do this if we give him
some help ! Be nice to get it close to correct the first time through, since
what we're really doing is writing the "rules" of buckets, and the performance
will depend on how good those rules are.

[sgeeeee]

Originally Posted by donheff

. . .Can this be sensibly modeled? Someone would need to do a PHD dissertation to come up with a rule set. It seems like bucket replenishment will involve an intuitive sense of touch.. .

Yup. You need several rules for each bucket. Transfer rules to cover transfers from that bucket to each of the other buckets, and rules to cover spending from that bucket. Each rule would have to depend on the value in every bucket, the recent performance of every bucket as well as on the most recent inflation. To do this right, you would also need to establish relative weighting functions for each of the above stated variables for each rule. The weighting functions could dramatically change the results.

It would make a great PhD dissertation. Of course the results of such a study might well be that such an approach does not provide a consistent advantage over straightforward asset allocation.

[youbet]

Originally Posted by JohnEyles

what we're really doing is writing the "rules" of buckets, and the performance
will depend on how good those rules are.

I think what would really, really be neat would be if Ray Lucia would write the "rules" of buckets!

[Nords]

Originally Posted by youbet

I think what would really, really be neat would be if Ray Lucia would write the "rules" of buckets!

Bucket #1: Write a book.
Bucket #2: Go on the road to explain the book.
Bucket #3: License the concept to others who'll write books & explain them.
Bucket #4: Update Bucket #1...