SWR, terminal values, TIPS, I-bnds & comm paper

[sgeeeee]

Tonight I have irrigation scheduled. My times to switch the valves so that I get my grove watered are at 1:00 AM to 2:00 AM. So I have time on my hands. . .

A couple of years ago I was looking at the effect on SWR and on average terminal value of using TIPS and I-bonds vs commercial paper. The results made me a convert on TIPS and I-bonds. Recently, I revisited these simulations and figured out that I still like TIPS and I-bonds, but not quite as much as I did originally. If you are a glutton for tabulated punishment, you might find this interesting. It's a lot easier to understand with graphs, but I don't know how to do that on these boards.

30 year SWR Simulations: It's interesting what TIPS and I-bonds (as opposed to commercial paper) do for the SWR and terminal value expectations. I used the following assumptions with intrcst's SWR 6.1 simulator:
(1) January values for the S&P500
(2) 50% of the annual withdrawal taken at the beginning of the year, 50% at year end.
(3) inflation indexed to the CPI.
(4) fixed income series: 1st case based on 4-6 month commercial paper -- 2nd case based on TIPS @ 2% -- 3rd case based on I-bonds @ 2%.
(4) 0.20% annual expense ratio
(5) portfolio rebalanced annually
(6) Stated maximum safe withdrawal rate is that which creates a terminal value of $1 at the end of 30 years.

SWR: I ran the 30 year simulations to find SWR and average terminal value. TIPS and I-bonds dramatically affect the results of the simulations for lower stock allocation percentages. The maximum SWR is obtained at much lower stock allocation values when TIPS or I-bonds are used. In particular, maximum SWR for the commercial paper case is 4.12% with a stock allocation of 60%. For the TIPS case, the SWR increases to 4.3% with a stock allocation of 30%. And the I-bond simulations resulted in a SWR of 4.32% with a stock allocation of 25%. See table below for abreviated data summary.

Stock ---------max SWR----------
Alloc. TIPS comm paper I-bnd
100 3.74% 3.74% 3.74%
90 3.87% 3.85% 3.87%
80 3.98% 3.95% 3.98%
70 4.07% 4.04% 4.07%
60 4.16% 4.12% 4.16%
50 4.22% 3.91% 4.21%
40 4.28% 3.64% 4.28%
30 4.30% 3.35% 4.31%
20 4.10% 3.05% 4.31%
10 3.89% 2.70% 4.29%

Average Terminal Value: Average terminal value increased monotonically with increasing stock allocation for all three cases. I've tabulated the inflation adjusted terminal values as a multiple of the initial balance in the table below. TIPS and I-bonds reduce the average terminal values for stock allocations of 50% or less. In fact, the average terminal values become alarmingly low in order at stock allocation levels that achieve "optimum" SWR. If you are looking only at SWR, you might be missing another kind of risk to the portfolio because overall balances become so low.

Stock ----ave terminal value----
Alloc. TIPS comm paper I-bnd
100 3.208 3.208 3.208
90 2.671 2.641 2.744
80 2.176 2.246 2.362
70 1.763 1.861 1.979
60 1.372 1.53 1.669
50 1.042 1.422 1.377
40 0.751 1.356 1.089
30 0.51 1.299 0.848
20 0.407 1.248 0.658
10 0.316 1.215 0.484

Terminal Values with equivalent Withdrawal Rates: Part of the effect noted above (ie. reduction of average terminal value using TIPS and I-bonds) is due to the use of increased SWR for these cases. So I re-ran the simulations capping the SWR for the TIPS simulations at the maximum SWR for commercial paper. (I leave the I-bond simulation for someone else). The TIPS case still results in a reduction of terminal value, but not as significantly.

Stock -----SWR----- -ave terminal value-
Alloc. TIPS comm paper TIPS comm paper
100 3.74% 3.74% 3.208 3.208
90 3.87% 3.85% 2.671 2.641
80 3.98% 3.95% 2.176 2.246
70 4.12% 4.04% 1.763 1.861
60 4.12% 4.12% 1.400 1.53
50 4.12% 3.91% 1.110 1.422
40 4.12% 3.64% 0.850 1.356
30 4.30% 3.35% 0.600 1.299
20 4.10% 3.05% 0.407 1.248


Summary: So for all three bond types, there is a tradeoff between maximizing safe withdrawal rate (SWR) and increasing your odds of experiencing portfolio growth that outstrips inflation (thereby increasing your SWR). Use of TIPS or I-bonds increases SWR over commercial paper, but only by moving the optimum stock allocation to values that reduce odds of large portfolio growth. If you use TIPS or I-bonds with equivalent stock allocations, you make marginal increases in SWR with small reductions in average terminal value.

[wabmester]

These simulations don't include tax consequences, right? *And the main difference between TIPS and i-bonds is the tax treatment, right? *Ergo, I'm confused about why you'd get different results for TIPS and i-bonds at the same real yield. * I must be ignorant of a difference in the interest compounding.

(Edited to say nevermind, I just read this: http://www.retireearlyhomepage.com/tipsbond.html)

Also, isn't it the case that if TIPS ever have a real yield of 4% again, we can go 100% TIPS, safely withdraw 4%, and our terminal portfolio value will be our initial value (not adjusted for inflation)? *That's what I want -- a CPI-indexed pension sponsored by the US Treasury.

[Ted]

These results seem incorrect for another reason.

The yield on commercial paper is close to the yield on Treasury Bills. *The data by Ibbotson Associates indicates that the real return on Treasury Bills over the past 65 years has barely exceeded inflation, so the same must be true of commercial paper. *But the yield on I Bonds would have exceeded inflation by perhaps 1.5% to 2%, and the yield on TIPs would have exceeded it by perhaps 2.5% to 3%, if these had been available during the same period. *So, how could a person accumulate more wealth holding commercial paper?

[sgeeeee]

Quote:

. . . *So, how could a person accumulate more wealth holding commercial paper?

Ted,

I went back and checked the results, plus ran checks with FIRECALC. Although FIRECALC always gives me slightly different numbers than intrcst's SWR calculator, they are usually very close and they were for this simulation too. Both models show the same qualitative trends and give 4-6 month commercial paper a slight edge over TIPS @ 2% for wealth accumulation -- even at equivalent initial withdrawal amounts.

Maybe intrcst and Dory can provide an answer.

[sgeeeee]

I just found this site that provides a thorough essay to explain these results. Just type "TIPS" in the dialogue box and read the results.

This sight also seems to have excellent essays on several other financial topics.

http://radioworldwide.gospelcom.net/essaygenerator/

[cute fuzzy bunny]

Here's a question: is there a way to rework firecalc or interpret its data such that for a given investment amount, it can TELL you what the optimal combination of investments is to produce the best combination of higher withdrawal rate and the largest terminal portfolio size? Or at least produce a table showing all 100 combinations of stock/bond percentages.

Would seem to take a lot of the guessing out of it.

[cute fuzzy bunny]

Quote:

I just found this site that provides a thorough essay to explain these results. Just type "TIPS" in the dialogue box and read the results. This sight also seems to have excellent essays on several other financial topics. http://radioworldwide.gospelcom.net/essaygenerator/

A better read than some of the reports I read while in the working world. Certainly more detailed.

[sgeeeee]

Quote:

Here's a question: is there a way to rework firecalc or interpret its data such that for a given investment amount, it can TELL you what the optimal combination of investments is to produce the best combination of higher withdrawal rate and the largest terminal portfolio size? *Or at least produce a table showing all 100 combinations of stock/bond percentages. Would seem to take a lot of the guessing out of it.

I did this study in 5% increments using SWR 6.1 instead of FIRECALC. My understanding is that these programs share a common origin and underlying data base. But SWR 6.1 made it easier to do this particular study. For other types of analysis FIRECALC has big advantages over SWR. I posted the graphs of SWR and average terminal values over on the MSN boards several months ago and they still appear:

http://groups.msn.com/REHPDiscussion...pgmarket=en-us

Look along the right hand side of the screen and click on each of the graphs. The assumptions used to build those graphs are as stated in the post above. I think that is the information you are asking about. But it still doesn't address Ted's question.

Gathering this kind of information with FIRECALC is difficult (unless I've missed something). For each case you have to run the simulator twice -- once to determine the maximum SWR, and again using that SWR to determine the average terminal value. I intend to do this for my own personal case eventually so that I can include the effect of my pensions and social security in a meaningful way. That's one advantage of FIRECALC. I just haven't gotten around to it and am not sure the results would be of general interest anyway.

By the way, another nice feature of SWR is that it gives you not only the average terminal value, but the highest and lowest for each simulation. The highest terminal value gives you reason to dream large.

[dory36]

Quote:

Here's a question: is there a way to rework firecalc or interpret its data such that for a given investment amount, it can TELL you what the optimal combination of investments is to produce the best combination of higher withdrawal rate and the largest terminal portfolio size? *Or at least produce a table showing all 100 combinations of stock/bond percentages. Would seem to take a lot of the guessing out of it.

A few years ago, before I retired, I built a PC-based program that did something like that, as I recall. It was on the REHP for a while, but it's not there anymore.

I stopped working on it, for a philosophical reason, as follows:

Building a safe rate calculator based on historical results is valid, if you accept the premise that the future is unlikely to be any worse than the worst we have ever seen.

Building an optimal allocation calculator based on historical results is valid only if you accept the premise that the future returns of bonds vs stocks will be sufficiently similar to historical returns that you're willing to bet your retirement on it.

The first seems reasonable to me. The second seems like it would be hard to support.

Dory36

Hi Dory! Worst case planning has worked for me.

BTW, congratulations on your good fortune. I am certain you will not be spoiled by your sudden wealth.

John Galt

[wzd]

Quote:

Building an optimal allocation calculator based on historical results is valid only if you accept the premise that the future returns of bonds vs stocks will be sufficiently similar to historical returns that you're willing to bet your retirement on it.

But you already have one. As SG and TH indicate, you have to input the ratio into firecalc, and can make multiple runs. If firecalc indicates a better SWR at 85% equities compared to 75% equities, then it did asset allocation decisions for you. If you decide historical data is good for SWR, then I am not so sure you are placing much more trust in it when checking for asset allocation optimization.

However, I agree with the sentiment behind the comment. Also, I think there may be better tools around that look at risk/reward of different portfolios, and those might be better for portfolio allocation decisions. The MPT based asset allocation calculators are the ones I am thinking of.

Also, I think the decisions might be better to be kept seperate, although there is some cross coupling between the results. The asset allocation decision should be made first, to set the point on the risk/reward curve you are comfortable with. Then run Firecalc with that percentage. Note that both MPT and SWR agree on avoiding 100% domestic equity allocations.

edited to add: after looking at SG's chart, I'm not sure how seperate the decisions really need to be. But I'm still a little uncomfortable in combining risk/return decisions completely with SWR decisions. Let me think on this a little more and I'll jump back in with an opinion later.

Wayne

[wzd]

SG -

You can include an image, if you first post it elsewhere on the web. Many ISP's provide home page space you can use, or even other BB's. Then place the url inside of image tags (second row, 4th from left on YABBC tag list)

For example, is this the right image from msn?



and the other one
[img]http://groups.msn.com/_Secure/0UAAAAFUZCB!hqIt3FnBSqBuqV32VnU!JdXQHR2KiC6hxLs6A4 RZfBW*k5AMwLKPb8V7zMmxfSXuxkCAPNowG*ncKRk7Q8a*JGAk 7JdbPYkqNZwnvH3pmRQAAAAAAAAAA/TermValues.tif?dc=4675444219909799355[/img]

From the long urls, I suspect that msn may do something to make the image unaccessable from a different ip... let me know and I will edit this.

Wayne
[edited to add second picture]

[cute fuzzy bunny]

I had the same thoughts on the tool...if its good enough to tell you whether your plan should work, based on historical data, then the premise should be good enough to tell you an optimal asset allocation based on your criteria. But the chart above helps.

Unless i'm mistaken, this says that with commercial paper, the right balance should be a 60/40 stock/bond...astounding since thats the standard balanced portfolio setup. With tips or i-bonds, it looks like 80% tips/20% bonds...yes?

I'm still missing something regarding tips. They return 2% plus inflation adjustments at this time, and inflation is roughly zero at this time, right? So unless inflation takes off and stays high (a popular opinion) you'd be eating it big time with a tips heavy portfolio. That would make for a moderately risky bet on inflation, in my humble opinion.

[wabmester]

Quote:

I'm still missing something regarding tips. *They return 2% plus inflation adjustments at this time, and inflation is roughly zero at this time, right? *So unless inflation takes off and stays high (a popular opinion) you'd be eating it big time with a tips heavy portfolio. *That would make for a moderately risky bet on inflation, in my humble opinion.

Depends on your mindset. To me, inflation is the enemy of retirement. At this stage, I'm not very concerned about building a bigger nest egg, living a more extravagant lifestyle, or leaving my kid a gigantic trust fund. Give me a guaranteed inflation-adjusted 4% of my portfolio each year, and I will be on financial cruise control.

Since I was TIPS-ignorant the last time the treasury made that offer, I have to take more risk. I must say, I like Ted's approach to risk management: basically do the vanilla MPT thing and spread your risk around, but skew things a bit when loud macro signals tell you to.

Loud macro signals are that a loose monetary policy will eventually lead to inflation and a weak dollar, baby boomer demographics will lead to a bigger burden / higher spending on health care as well as drive inflation if a small workforce implies higher salaries, and depletion of limited resources like oil, clean water, and young virgins will make these things more valuable.

[wzd]

Ok, thought a little. There is a real difference between portfolio survival, which SWR and Firecalc deal with, and investment growth. As you move toward lower equity and more TIPS you decrease the variability of the portfolio returns, and the likelyhood of maintaining a withdrawal is higher, but investment growth is lower. SG pretty much said this in the first post. So the issue is what are you optimizing the asset allocation for? There is a spectrum of choices, not a optimum.

Terminal value reflects some of this difference, but in an abstract way. A more understandable, and more useful to the living retiree, number would be "average new SWR after 5 years (or 10....)". To explain this better, consider a firecalc run for 30 years, which has a 95% safe SWR of $xxx per year. Now repeat the run with exactly the same parameters, but change it to 5 years, uncheck the 95% box, and note the median terminal value (I hope average is close for this case). This number is the 50% likely number for the portfolio balance in 5 years. Now repeat the Firecalc run with this balance but for a period of 25 years, check the 95% box, and note the new SWR. This is the SWR that you have a 50% chance of increasing to after 5 years. If one does the above exercise with a TIP heavy portfolio, there will not be much increase beyond inflation. With a portfolio that has more equities, there should be some increase.

Some data for a 1mil portfolio: with a 75% equity, 25% tips@2.5 the initial SWR is 46700, and the 5yr SWR is 60782. With 25% equity, 75% tips @2.5 the initial SWR is 46400 and the 5yr SWR is 54853. For some reason the delta between 75/25 and 25/75 seems different than SG's graph. I used CPI, and the rest were defaults.

So what this says, (and is somewhat obvious in retrospect) is that the max SWR for the 80% TIP portfolio comes with a price. You have a very high likelyhood of maintaining withdrawals, but will not be very likely to see a future increase.

The 80% equity portfolio has a slightly lower initial SWR, but is much more likely to allow for a future increase in SWR. This is one issue that I have not seen explicitly addressed here. Gummy does sort of address it with his "sensible withdrawal rate" dialoge.

So, I guess the process is to first assertain how comfortable you are with a 4% withdrawal for life. If the cash flow is well above your comfort level and allows whatever luxuries you want, you are best off with the TIP rich portfolio. If you consider the 4% barely adequate, and really want a little more for travel or whatever, then the equity rich portfolio would be a better match. Also inheritance may play a role, depending on your viewpoint.

Note that the dollars in the above examples are affected by inflation, but as far as I know, there is no way in Firecalc to see how much. The best I can think of is to use a 3 to 3.5% factor when looking at the numbers. (Dory - yet another idea for firecalc - provide an option to see all data in current dollars!)

SG - Thanks for the work that went into the graphs!

Wayne

[wzd]

TH, Wabmaster - My spreadsheet says that a 2% return with 0% inflation, and a 4% withdrawal rate will last 36 years. A return of 4% with 2% inflation will last 36 years. a return of 10% with 8% inflation will last 38 years. So, I think it is not as much a bet on inflation as a bet on 30 year payouts.

There is little difference between 30, 40 and 50 year runs for portfolios heavy in equities; there is more of a difference for portfolios light in equities.

SG: Did you do anything other than 30 year runs?

Wayne

[wabmester]

Hmm, some new versions of FIREcalc I'd like to see:

1) Best-case outcome. How much could I withdraw if we repeat the best N-year performance in US history?

2) How about using stock market data from other countries? I always find it sobering to compare the S&P500 to the Nikkei, for example:
http://finance.yahoo.com/q/bc?s=^N22...z=m&q=l&c=^SPX

3) The Great Depression and the 70's stagflation were the results of bad monetary policy. We've learned our lessons, and it's unlikely we'll make mistakes like these again. Throw out the singularities and see how you do.

4) What future returns will the stock market yield if we fail to invent cool stuff like distributed electricity, automobiles, mass production, computers, and internet porn in the next 100 years. Yeah, this would be tough to simulate, but this would be a stock investor's worst nightmare.

[dory36]

Quote:

I had the same thoughts on the tool...if its good enough to tell you whether your plan should work, based on historical data, then the premise should be good enough to tell you an optimal asset allocation based on your criteria.

I'm not describing my case very well, sorry.

The worst-case philosophy underpinning the safe withdrawal rate calculators assumes nothing about history, except that it provides a sufficiently all-encompassing set of possible scenarios.

In effect, we are looking at a whole lot of random numbers, and looking at where they fall, and then setting our strategy somewhere just beyond where the worst numbers fall. The various events and policies of history have caused these numbers to move around all over the place. All we have done is find the edges of the pattern, and positioned ourselves just beyond.
The closer we can get, the more money we can safely withdraw from our portfolio without getting hit - busting the portfolio.

Note that we are making no assumptions about how these random numbers are distributed -- just that they are feasible. We know that only because they have, in fact, occured.

I'd liken this to raindrops, as an analogy. We want shelter, but we want to be as close as we can get without getting wet.

On the other hand, trying to optimize means we have to assume that the policies and events of the future are going to be sufficiently similar to the past that the distribution can be predicted.

Or, following the raindrop analogy, we're trying to predict from past rainstorms just where in an open field we can place a bucket and catch the most water.

I don't believe it can't be done - I just don't believe that there is enough data embedded in FireCalc and its cousins to answer the necessary questions. (And I don't even know what questions they are!)

Quote:

3) The Great Depression and the 70's stagflation were the results of bad monetary policy. We've learned our lessons, and it's unlikely we'll make mistakes like these again. Throw out the singularities and see how you do.

Yep - that's why many shoot for 90% or 95% safe rates, instead of 100%.

Dory36

[unclemick]

And then, And then - there is throw out all the data and live on what you make - i.e. the portfolio yield. That was the 'old fashioned way,BC (before calculators)'. Of course that truly sucks in today's environment so 4% SWR and FIREcalc for 'our' situation is 'the best availible solution' - until/when something convinces me otherwise.

[sgeeeee]

Quote:

TH, SG: *Did you do anything other than 30 year runs? Wayne

Not for the general case that I posted. One of the problems with looking at different time periods is that means you are looking at different numbers of cases. Especially if you start looking at time periods of over 30 years, the error bars on the results get pretty large. And yet, it is for people who probably will have to fund retirement for far longer than 30 years that these results might be most important.

Give me some specific definition of what you might like to see in terms of other simulations and I will probably get around to it eventually. Then we can talk about what it does or doesn't mean.

Regarding seeing some differences in your simulation from the ones I posted, remember that I did not use FIRECALC, so I did not use FIRECALC defaults (although the I used values close to those defaults).

I believe the analysis you've posted so far is an excellent summary of what I believe these simulations might tell us. Thanks.