FIREcalc rewrite - suggestions?

[JWR1945]

Sorry Ted but I like Telly's idea. Let me suggest the following refinements:

1) Put a special button called Sensitivity Studies on the main page. Make no changes to the basic FIRECalc entries.
2) In the sensitivity study section list two sets of entries corresponding to the calendar years that are covered by data. One set of entries would cover 1871 through 2002 and that would refer to the data that already exists in FIRECalc. The other section would be next to it and it would be the what if entries. It would start with default values equal to the current data set. If someone wanted to replace years 1940-1945 with data from 1960-1965 he would replace the 1940 default value with 1960. He would replace the 1941 default value with 1961 and so forth.
3) To satisfy Telly's wishes, include a set of entries through 2042 (or some other future date) with additional, dummy default values such as the 1871-1911 that Telly mentioned. But be sure that the user could also change these projections aka guesses.
4) From a programming standpoint, you would simply be replacing the table that has your historical data with data selected from different dates.
5) Limit changes to replacing data for one year with existing data from another year. This makes the programming task manageable.
6) Be sure that any calculations remain in the special section. Keep it as idiot proof as possible. (Understanding, of course, that nature is always able to supply a superior idiot.)

Have fun.

John R.

 

[Ted]

I'm so argumentative that I have to disagree somewhat with my own analysis that I posted earlier today, although I stick by my prior recommendation to leave the data set in FIRECalc as it is.

I said that "tacking on" the returns from the start of the data period to the returns at the end of the data period would cause the early returns to be "double-counted." While this is true, it is also true that the returns from years in the middle of the data set are counted numerous times. For example, if the period of analysis were 20 years, then 1870 would be counted once, 1871 would be counted twice, and the years from 1890 to 1982 would each be counted 20 times. So FIRECalc inherently provides the heaviest "weighting" to the returns from the "middle" years.

To the extent that this causes a distortion in the results, one would expect it to be greatest for the longest periods of analysis. I don't think there is any way around that problem, but I also don't think that it is a major problem from a practical standpoint, because the "middle years" certainly include a number of years that had horrible market returns.

From a practical standpoint, the major uncertainties faced by any person in planning retirement financing are (1) not knowing how long they will live and (2) not knowing what the actual future returns on various alternate investments will be (especially after taxes and inflation). While a professional statistician might be able to slightly improve the theoretical validity of FIRECalc, a practitioner of real world economics like myself would say that the benefit of doing so to any individual planning their retirement financing would be nil.

I have enough confidence in my own intelligence to admit that JWR1945's suggestions for changes to FIRECalc are over my head . They may be brilliant, but they are definitely not simple and "idiot-proof!"

 

[JWR1945]

Thank you, Ted.

My term idiot proof refers to making sure that anyone who goes into the special section for a sensitivity study knows that he is there and not in the normal FIRECalc section.

Without special precautions, someone could be sent back to the original section to continue making his inputs. If he forgot that he had changed things in the special section, he could come up with all sorts of nonsense.

Who might forget? Forget what? (humor)

I imagine that you yourself might enjoy such a section. Captain Bill aka Dory36 does a marvelous job making user interfaces. I think that he could make it much easier for you to use than for me to explain.

You might want to know what could happen if went through something like the sixties and seventies over the next two decades. Looking at the year by year results might help you figure out what to look out for and when to feel comfortable with your portfolio.

Have fun.

John R.

 

[Ted]

I feel a little guilty recommending things that would make the software for FIRECalc more complicated, since I'm not the one who would have to do the work. But here's another idea that I don't think would be too difficult to implement.

FIRECalc assumes rebalancing a portfolio at the end of every year to a specified percentage of stocks (with the remainder of the portfolio going into the specified fixed income asset). One recommendation (by JWR1945) was to permit the frequency of rebalancing to be changed; however, I think that that would be difficult mathematically and would not really provide any useful guidance as to what frequency of rebalancing is likely to produce the best results in the future. As far as anyone can tell, doing it once a year should be adequate.

The conceptual problem is that, as has been recognized in various posts, the "optimal" safe percentage allocation to stocks declines as a person ages and their "planning period" declines. But FIRECalc doesn't provide a means of accounting for this (except by some manipulations involving sequential runs that I won't try to explain.) For example, if a person inputs a 40 year planning period, with a 50% allocation to stocks, FIRECalc rebalances the portfolio to 50% stocks at the end of each year. In practice, however, as the person aged, they would want to periodically reduce their allocation to stocks, such that when they were, say, 90 years old and their planning period was down to 10 years, they would optimally have only 15% or so in stocks (and the rest would best be in TIPs).

I think that FIRECalc could be modified fairly easily to permit the allocation to stocks to be changed about three times, to a specified revised percentage after a specified number of years.

Many thanks to Dory36 for his efforts. It's a valuable service.

 

[dory36]

Ted -- thanks for the suggestion, and don't feel guilty -- I am perfectly capable of ignoring suggestions that I don't feel like implementing!

I'll look at the feasibility of letting people change allocation a few times when I get around to a rewrite.

(For those who haven't picked up on this, I'm cruising full time on a boat. I generally travel for a day, stop a few days, travel a day, etc. These stops don't lend themselves to programming. But every 6-8 weeks, we'll stop for a month or so, and that's when I pay attention to updating FIRECalc or other stuff that requires brain cells.)

Dory36. currently in Islamorada FL, at 85 degrees

 

[JWR1945]

My preference so far is Ted's suggestion in reply #20:

Here's a suggestion that I suspect would be easy to implement: allow a different "expense ratio" to be entered for each category of asset.

Therefore, while Dory36 makes it clear with good humor related to a later suggestion that:

Ted -- thanks for the suggestion, and don't feel guilty -- I am perfectly capable of ignoring suggestions that I don't feel like implementing!

I want him to give this particular suggestion of Ted's a priority. (BTW, he did not reject Ted's later suggestion...at least for now.)

While we are at it, is there any reason that FIRECalc is not available directly at this location?

Have fun.

John R.

 

[dory36]

I'll be looking at the changing expense ratios and the changing allocations sometime down the road, err, waterway.

Meanwhile, there is a link to FIRECalc at the top level of the forum. Look at "News" and you'll eventually see: "Do you have enough to retire? Visit FireCalc" --

the word FireCalc is a link.

Not sure where else I can easily put it.

Dory36

 

[JWR1945]

I recommend that you implement switching into at least one version of the FIRECalc.

By switching, I mean that you can vary portfolio allocations according to P/E10 (as defined by Professor Shiller). I have recently investigated switching using the Retire Early Safe Withdrawal [Rate] Calculator, version 1.61, November 7, 2002 (as-is, without any modifications of my own).
http://rehphome.tripod.com/re60.html

Earlier reports are in error. Switching produces a big payoff.

Although the calculator has provisions for two thresholds (and three allocations), it implements only one. Data inputs at F19, B20 and I20 work. Inputs for I19 and F20 do not. That is, the Low PE to Mid PE threshold works and the stock allocations for the Low PE Years and the High PE Years work.

I discovered this when I modified a version to include switching between stocks and TIPS. The relevant code appears in the formula bar (at the top) when you highlight any cell in row 182 Stock Switch by P/E. Before then, I had assumed that all of the thresholds and allocations were active. So had others.

Using a threshold of 12.0 and stock allocations of 80% (below threshold) and 20% (above threshold), I was able to bring the reported 30-year 100% Safe withdrawal rate up to 4.8%, compared 4.12% without switching. The fixed income portion of the portfolio consisted of commercial paper. (Performance is even better with TIPS, but it requires modification of the calculator.)

I routinely report my research along these lines at the SWR Research Group board at www.nofeeboards.com.

Have fun.

John R.

 

[renferme]

For what it's worth, I used Firecalc in this way: put in a dollar amt. as a starting point, 100% in stocks, for a 10 year period. I did this just to see what I would earn just from stocks over every 10 year period from 1871 to 1991. Years 1992 thru 2000 show as incomplete 10 year periods. While the S&P500 has 'earned' about 11 percent from 1926 thru 2000, Firecalc shows from 1871. I can see that the market goes in cycles that sometimes are bad, real bad. These periods are from 1873-1876, 1880-1890, 1901-1914, 1922-1937, and 1964-1973. Of course, the period from 2000 thru 2002 was also terrible, but firecalc doesn't show it. The financial gurus like to talk about the 1920's thru today, omitting pre 1920 market results.

So,,,, is the period 1871 thru 1920 irrelevant ? Should Firecalc not take into consideration the 19th Century

 

[renferme]

For what it's worth, I used Firecalc in this way: put in a dollar amt. as a starting point, 100% in stocks, for a 10 year period. I did this just to see what I would earn just from stocks over every 10 year period from 1871 to 1991. Years 1992 thru 2000 show as incomplete 10 year periods. While the S&P500 has 'earned' about 11 percent from 1926 thru 2000, Firecalc shows from 1871. I can see that the market goes in cycles that sometimes are bad, real bad. These periods are from 1873-1876, 1880-1890, 1901-1914, 1922-1937, and 1964-1973. Of course, the period from 2000 thru 2002 was also terrible, but firecalc doesn't show it. The financial gurus like to talk about the 1920's thru today, omitting pre 1920 market results.

So,,,, is the period 1871 thru 1920 irrelevant ? Should Firecalc not take into consideration the 19th Century

 

[Ed the Gypsy]

I seem to recall that gummi evaluated switching (market timing) based on PE ratio in his web pages and showed that it didn't work very well.

gummi, are you there? Can you help here?

The only problem I have with gummi's otherwise wonderful site is that I have trouble finding things again. There is so much good stuff there.

Ed

 

[JWR1945]

Understand that I am reporting new findings.  The Retire Early Home Page article that I referenced claimed that neither P/E nor P/E10 could improve performance.  I have now shown that P/E10 produces a substantial improvement using the identical calculator.

FIRECalc was originally based on an early version of the Retire Early Safe Withdrawal [Rate] Calculator.  Dory36 added a fabulous user interface, but the underlying calculations were the same.  Since then the two calculators have diverged somewhat.  Undoubtedly, the kind of results that I have just reported would show up on FIRECalc if it had a stock allocation switching capability.

I brought up the details of my discovery within this thread: 3% SWR for 56 Years dated Monday, Oct 13, 2003 at 7:19 pm CDT.
http://nofeeboards.com/boards/viewtopic.php?t=1541

Start here, read this post and the two that follow:
http://nofeeboards.com/boards/viewtopic.php?p=12651#12651

I was able to reconstruct what was reported at the Retire Early Home Page.  I showed that switching stock allocations according to (Yale Professor Shiller s) P/E10 can increase the Safe Withdrawal Rate to 4.85%.  [I used the specialized term Historical Database Rate instead of Safe Withdrawal Rate in those posts.  This term draws attention to the fact that the calculation itself is not a prediction.  Only when you assume that the future is no worse than the past (and when you explain exactly what those words mean!) does it become a Safe Withdrawal Rate.]

The safety of retirement portfolios is strongly correlated to P/E10, which uses the average of ten years trailing earnings instead of just one year.  The correlation with P/E is weak.  I have recently shown that improving (declining) P/E as an early buy signal when P/E10 is high is a very bad idea.

Gummy hosts a board of his own at www.nofeeboards.com .

Have fun.

John R.

 

[wzd]

Hi Dory -

I would suggest two items - one has been mentioned above by others, and by me in the past - and that is that the modelling of alternative income streams needs to be more intuitive. I suggest seperate boxes for alternative income streams (called out-year revenue in some posts) and changes in withdrawals due to lifestyle changes. I know that they are computationally equivilent, but knowing that should not be required of users. The fixed pension modelling especially should be more obvious.

The second is really just because of my need to look at the details to make myself comfortable - but I think there are quite a few others like me. That is to allow me to select a specific year, and see the detailed calculations for that year.

Let me know if you would like any help on FireCalc. I have been retired over a year now, and have recovered from being totally burnt out on computer work and have been playing around with programming again.

Wayne

 

[dory36]

wzd - thanks for the offer!

I'm debating how I will do the next rewrite.

One option is rewriting FireCalc using PHP instead of ASP/VB so I can move it to a server that has more capacity and less cost. How's your php?

I'm guessing I will get heads-down into the rewrite in about a month or so, after recovering from the move to land. :-/

Dory36

 

[Mr._johngalt]

Hey Dory, and how is the move to land coming. We are
interested.

 

[dory36]

This week we got somewhat settled -- parked the RV on the outskirts of nowhere, TX, got a real phone and real internet (via microwave) for the first time since I retired, and are starting to get stuff put away.

Leaving Dory[/] (the boat) was not too difficult, but watching hundreds of boats headed south, including dozens of boats we knew, was traumatic.

On the bright side, we got a music CD from a cruising singer, and found a song that can be the ER theme song. I've asked for the OK to put the whole song on this site, with a link to the order page. I won't say more until I get a response.

Dory36

 

[Feawen]

I hope that you can add a factor for people who have pensions, in addition to social security. I didn't know how to add that factor in.

Feawen :

 

[Cut-Throat]

Feawen,

Just add your Pension to SS and use the same box or use Spouse SS box. Or use a withdrawal box and input a negative number.

Quite a few ways to do this. Be creative!

 

[dory36]

Cut-Throat is right. The purpose of the three withdrawal change boxes is to allow for pensions or the like. (They are no different computationally from the SS boxes, except that you have to option to indicate whether the pension is inflation adjusted or not.)

You enter your required living expenses - say $40,000.

Then using one of the withdrawal change boxes, you reduce your needed withdrawal by 25,000 in year 8 because that is when your pension for that amount will kick in. You indicate in the box provided if the pension is inflation-adjusted.

What will happen is your portfolio will be hit for $40,000 the first 7 years, then only 15,000 the 8th year, since your pension will make up the differnce.

Dory36

 

[wzd]

And to continue to beat a dead horse, please change one of the withdrawal boxes to add instructions on inputting a pension, and also retitle the box to indicate 'pension or other withdrawal adjustment'.

Wayne

 

[dory36]

You know, I get this question about once a week -- how do I enter a pension. It's all right there -- withdrawal changes 1-3.

But having seen that the floggings will continue until morale improves, I've modified the text to discuss pensions.

Not sure what that expression has to do with the topic, but I haven't used it since I retired 3 years ago, and it was getting lonely from disuse!

Dory36

 

[sgeeeee]

You know, I get this question about once a week -- how do I enter a pension. It's all right there -- withdrawal changes 1-3.

But having seen that the floggings will continue until morale improves, I've modified the text to discuss pensions.

Great. Right after you get that finished, can you tell us how to include pensions?

Just kidding . . .

 

[wzd]

Thanks Dory. I see the question come up quite often, so answering it in advance is the right thing to do. It seems obvious to us, but seemingly not to everyone.

I'll stop beating you with a wet noodle now. (and where did that expression come from?).

Wayne

 

[BunsGettingFirm]

I have already made the same suggestion over at the FAQ section, but the FIRE calc needs two nest egg entries, one for taxable and one for tax-deferred. For most, ER means retiring before 59, so it would be nice to know how long the taxable account will last and how much the tax-deferred account will be when I'm finally able to withdrawl from it.

 

[wzd]

I have used ORP, which is at www.i-orp.com, for that type of planning, and you may want to have a look if you haven't. It is more for planning withdrawals from taxable vs. non taxable accounts.

Wayne