The Great SWR Investigation-Part 2

[[KenM]]

Don't really trust averages - I never seem to get the average of anything

It'd be nice to see the range of 10, 20, 30, 40 year withdrawals - if that's easy to do????

[WiseNLucky]

gummy:

Here's the distribution of final 40-year portolios (for our 10,000 investors):

I'm not sure how to read the graph. Is it saying that at 6% of the original portfolios failed, that 3% of them had $10 million and that 1% of them had $20 million? Sorry for my ignorance.

Is the assumption of a mean 9% return too high for today's valuations? What would happen if the mean were dropped to the 6 or 7% we've been hearing about lately?

[gummy]

KenM:
My (current) sandbox is here:
http://home.golden.net/~pjponzo/SWR-3.htm
At the bottom is a link to a file you can download. It unZIPs to two spreadsheets. One gives the distribution of portfolios (after 40 years). This "distribution" spreadsheet will do the earlier years too if you just copy/paste the portfolios from the SWR-MC spreadsheet.

WiseNLucky:
In the spreadsheet you can stick in any numbers you like. (example: 6% or 7%).

... 6% of the original portfolios failed, that 3% of them had $10 million and that 1% of them had $20 million?

I'm not sure I understand the question.
The chart says that, at year 20 (for example), those investors that survived that far had an average current withdrawal rate of (about) 4.5%.

By current I mean:
(current withdrawal amount) as a percentage of (current portfolio amount)

[gummy]

... the range of 10, 20, 30, 40 year withdrawals ...

Aha! The distribution of withdrawals, NOT portfolios.
Good idea! I'll work on it.

[[KenM]]

gummy
You seem to still have a lot of sand left in your sandbox.
Wish I'd still got so much left in mine - and I'm about 10 years younger than you

[gummy]

KenM:
Okay, now there's a spreadsheet which generates the portfolios and withdrawal rates for umpteen investors, at four points in time (example: 10, 20, 30 and 40 years). That's eight columns worth.

Now you copy any of the eight columns, and there's another spreadsheet where you paste the column and this second spreadsheet plots the distribution.

The second spreadsheet looks like so:
http://home.golden.net/~pjponzo/SWR-10.gif

P.S. That second spreadsheet is also useful for plotting the distribution of any set of numbers you care to paste (like S&P 500 returns for the last umpteen days/weeks/months/years).

[WiseNLucky]

Gummy wrote:

I'm not sure I understand the question. The chart says that, at year 20 (for example), those investors that survived that far had an average current withdrawal rate of (about) 4.5%.

I'm sorry. I understand that graph (with 4 columns representing each decade). The graph I don't understand is the one that has Average=$14.07 and SD=$21.92 at the top. A distribution of final portfolios. We don't have a way to identify specific posts here so I used your verbage from the top of the post.

Does my question make sense in reference to this other graph?

[wanderer]

gummy -

how do we explain the outlier of 15% w/drawal rates occuring so frequently and how can i sign up?

[BenSolar]

We don't have a way to identify specific posts here so I used your verbage from the top of the post.

How about this for gummy's post with the chart you referenced:

http://nofeeboards.com/boards/viewtopic ... 7026#p7026

If you hover over the 'quote' button on a given post, then you can see the post number. Use the format above, but replace the two '7026's with the desired post number.

It worked when I previewed this post ... 8)

BenHacker

[ataloss]

It worked when I previewed this post ...


*************

I like it!

[WiseNLucky]

Bensolar wrote:

It worked when I previewed this post ...

Beauty

Clicking on your link took me directly to the post in question. You should post instructions on how to do this as a sticky at the NFB board.

[WiseNLucky]

I wrote:

Is the assumption of a mean 9% return too high for today's valuations? What would happen if the mean were dropped to the 6 or 7% we've been hearing about lately?

I downloaded gummy's spreadsheets and played with them. I used 7% mean returns and 15% SD with a 4% withdrawal rate and 2% inflation.

Results were awful. 33.65% of the original portfolios failed in 40 years.

This will take some time to play with and understand. But my first thought is that expected returns when starting valuations are high will render the 4% "rule of thumb" tragically dangerous.

[[KenM]]

Results were awful. 33.65% of the original portfolios failed in 40 years

Which is why, after all the helpful information from others in recent posts, my conclusion is that, for me, trying to fix an SWR that I would have any confidence in would largely be a waste of time. My estimate may well be too high or too low - and that doesn't suit the way I think. So, for me, deciding upon a withdrawal strategy would be far more important than thinking about a fixed withdrawal rate.

I'm probably going to consider either gummy's mildly pessimistic approach in his sensible withdrawal strategy (i.e. annual withdrawal, say, 3%+inflation for basic living and then take out more in good years for luxuries) or a mildly optimistic approach (i.e. annual withdrawal, say, 4%+inflation for comfortable living but in bad years reduce to 3% for basic living and in very bad years to 2% for survival living). Only an outline - and I'll spend more time thinking about it in the traffic jams on the way to work.

[gummy]

WiseNLucky:

I'm not sure how to read the graph. Is it saying that at 6% of the original portfolios failed, that 3% of them had $10 million and that 1% of them had $20 million?

Aah, now I understand! (It's that creeping senility ag'in. Sometimes I don't even understand the missus.)

Anyway, it's my fault. I stopped the vertical axis short.
I redid the 10,000 investor analysis and the chart now looks like this:

and it's generated by that "distribution" spreadsheet which I mentioned somewhere ... earlier, I think.

Anyway, you pick a range ($0 to $100) and the spreadsheet divides it into 100 slots ($0, $1, $2, etc.) and counts the number in each slot. So (using the new/improved picture) there are about 13% at $0 (they didn't survive the 40 years), about 6% in the $0-$1 range, about 1% in the $19-$20 range ... and so on.

Because the percentages depend upon the range you select, the spreadsheet also gives the percentage lying outside your range. Since I chose $0-$100, there were 437 lucky gals with final portfolios greater than $100; that's 4.37%

That "distribution" sheet looks like this:
http://home.golden.net/~pjponzo/SWR-10.gif
and shows the failed portfolios and those above the range you selected ... in different colours. If'n y'all want ALL the investors identified, choose some huge range, like $0-$500. (The spreadsheet tells you , before you choose, the maximum portfolio, say $456.78 ... so choose $0 - $460.)

[gummy]

Results were awful. 33.65% of the original portfolios failed in 40 years.

Interesting, eh?
I ran the "sensible withdrawals" spreadsheet and found that, choosing 4% (initial) withdrawal rate, actual inflation rates and random annual returns from Small Cap Growth returns (for the years 1928-2000) ... 33% failed to reach 40 years. And the Mean Return for the SmCapGwt (over that time period) was 14%!!

BUT, the Standard Deviation was almost 33% which gives an "annualized" return of (about):
Mean - (1/2)SD^2 = 0.14 - 0.33^2/2 = 0.086 or 8.6%

On the other hand, for the S&P500, only 16% failed to survive 40 years.

Moral? When you retire, don't trust some mathematically generated SWR.

[hocus]

Moral? When you retire, don't trust some mathematically generated SWR.

Or--

Include in the mathematical calculations data on all the factors that determine the answer to the question you are trying to answer.

[wanderer]

Moral? When you retire, don't trust some mathematically generated SWR.

a significant statement from a former math prof.

here's one from a former auditor: don't trust the financial statements without corroborating evidence.

[gummy]

hocus:

Include in the mathematical calculations data on all the factors that determine the answer to the Question you are trying to answer.

I must be slow, but, so that I might (more clearly) understand the point you're making (regarding mathematically generated SWRs), maybe we could agree that any mathematical prescription requires numbers (as inputs) and spits out a SWR.

Like so:


Could you suggest some "typical" numbers (as inputs) and a "typical" Question you're trying to answer ... in order to clarify your stance on SWRs?

For me, I'd suggest that if theQuestion is:
"How can I estimate a SWR, umpteen years from now, when I retire, in order to determine how much I should be investing right now?"

The inputs could be:
Number of years (after retirement), N, before I drop dead
Expected Mean Annual Return
Expected Standard Deviation
Expected Inflation Rate
A Probability of Survival (for N years)
Some distribution of returns (lognormal, normal, whatever)

The Black Box could be a jillion Monte Carlo simulations.

[hocus]

I must be slow, but, so that I might (more clearly) understand the point you're making (regarding mathematically generated SWRs), maybe we could agree that any mathematical prescription requires numbers (as inputs) and spits out a SWR.

You're not slow at all. You get it exactly. It's about inputs and outputs. If you use the right inputs, you will get a good output. If you don't, you won't.

I'll illustrate with an extreme scenario to reveal clearly the effect of changes in valuation levels. Once we all see that the effect is real, all we need to do is to be sure that the effect is accounted for both in extreme scenarios and in not-so-extreme scenarios. The effect is always present, it is just not always so obvious as in the scenerio I describe below.

John and Mary are both 50 years old. Both would like to retire as early as possible. Both plan to take 4 percent withdrawals from their portfolios and to live on $40,000 per year.

As of July 1, 2003, John has saved $1,000,000, and Mary has saved $500,000.

John retires on July 2, 2003. The Dow is at 8,000.

Over the course of the next 12 months, the DOW skyrockets to 16,000. Mary's Portfolio doubles in value to a value of $1,000,000 on July 1, 2004. She retires on July 2, 2004.

Which retirement is more safe, or are they equally safe?

Intercst says that both retirements are 100 percent safe (let's call it 95 percent if you count in the chance of the future being unlike the past). He says the two retirements are equally safe.

I say that is impossible. I know from reading Bernstein that valuation levels affect return expectations. The valuation level applicable at commencement of Mary's retirement is much higher than the valuation level applicable at commencement of John's retirement. Mary cannot reasonably expect the same returns going forward from the commencement of her retirement date as John can reasonably expect going forward from the commencement of his. So any reasonable assessment of the safety of Mary's retirement is that it is much less safe than John's.

If John's retirement is 95 percent safe, then Mary's retirement is much less than 95 percent safe.

Any tool used for assessment of the safety of a retirement portfolio that does not take valuation levels at the start of retirement into account is fundamentally flawed. This factor always affects the result. A tool that does not take it into account cannot produce the correct result. That sort of study has not taken into account all the inputs that matter. Without taking into account all the inputs, it is not reasonable to expect the analysis to produce the correct output.

[galagan]

I must be slow, but, so that I might (more clearly) understand the point you're making (regarding mathematically generated SWRs), maybe we could agree that any mathematical prescription requires numbers (as inputs) and spits out a SWR.

To put it another way, I would always want to understand what's involved in the operation of the "black box."

dan