New version of FIRECalc for testing

Status
Not open for further replies.

dory36

Early-Retirement.org Founder, Developer of FIRECal
Joined
Jun 23, 2002
Messages
1,841
I've posted a new version of FIRECalc at
http://capn-bill.com/fire2/index.asp
for testing.

The underlying computations have not changed yet -- I'll be adding updated data soon.

What has changed is the layout of the questions to be filled in, and the example data. Now, it assumes a somewhat realistic portfolio (rather than $1000), and puts some figures in for future income changes in the example. A few defaults changed slightly also.

Comments please!

Dory36
 
Great job Dory36.

I made a few "limited" runs (don't have my actual data here with me at this time) and it appears to be more intuitive. I'll try to use it with some real live examples in the next day or so and let you know.

One immediate observation. It would be useful to have at least two or three more "Additional Withdrawal Changes". I personally have a situation where my withdrawal needs will change more than once over my retirement lifetime.

Several calculators I've seen allow you to assume a reduced (or increased) withdrawal at various points in retirement. That appears to be logical to me. I don't think I'll need as much when I'm 80 as I do when I'm 55. The addition of a few more Additional Withdrawal Changes would accomodate that type of planning assumption.

Hope this helps. I'll try to give some more feedback soon.

Moguls
 
The new version is working well and is very useful. Can't wait till you update the data to see if there are significant changes in my situation.
 
Dory36,

I tried the new version. I don't know if I'm doing something wrong, or there's a bug in the new version, or it's in the rounding from the old version. But I get different results.

I get higher survivable values in the new model (hope it's the correct one!!!)

Here's some sample data. Try it and see if you get different results.

Init port of $1,000,000
Init withdrawal of $50,000
Life span of 40 yrs.
In year 0 decrease withdrawal by $20,000 (no inflation adjustment)
In year 12 decrease withdrawal by $6,000 (inflation adjusted)
In year 11 increase withdrawal by $33,000 (inflation adjusted)
In year 2 one time cash inflow of $2000
In year 15 one time cash inflow of $10,000

Use 75% equities, 25% commercial paper with expense ration of .18%.

First year withdrawal immediately.

In the original model I get a survivable rate of 89.2%. In the new model I get 93.1%.


Hope this helps.


Moguls
 
Thanks for the feedback!

Moguls, I first found the same discrepancy you found when I ran your numbers. I started over, and carefully went line by line through your message and entered the info again, and then the results matched between old and new.

I am guessing it is due to one of the defaults I changed between old and new, and I must have missed one of the default settings the first time through.

The old one defaulted to first year withdrawals immediately (different from Intercst's spreadsheet, and source of much confusion); the new one treats first year expenses as outside the model, unless you override it.

Also, the default equity split was 77% on the old, and 75% on the new one.

The whole program is split between a handful of modules. The input is collected with one, and that's the one I updated, but I didn't intentionally change any of the input variables. The computations all take place in a few modules that have not been touched yet. So, if there is a remaining discrepancy, it has to be in the data collection module.

If you still see the discrepancy, please try running the "DETAIL RESULTS" option, and see what the year-by-year data look like.

I'll look at additional withdrawal changes, but I'm a bit worried that the input is too long already. I'll put this one into my list of "how can I do this" updates. Maybe an optional additional window? An interim idea for investigating your situation is to treat your portflio as two portfolios, one with 3 changes, and the other with three different changes. You might be able to

I do have one question though -- why change the withdraway in year zero? Oops, I just saw it was non-inflation adjusted. Must be a fixed pension.

Stevelb, I hope to incorporate new data in a few weeks. All the percentages will shift slightly, as we'll be using 132 years of data instead of 131, but those shifts will be pretty small.

What devastates portfolios is bad early years, not bad late years. When the portfolio gets clobbered two or three years in a row at the beginning, it seems to fail early. If it gets off to a good start, it soars.

So, I'm not expecting dramatic changes once the model includes the 2001 data. The "bad years" of the 1960s (1964-69) are already showing up as bad years, and withdrawals much over 4% are already showing as failing years for retirements in those years. Thus, the 2001 numbers will not change the outcome by much at all.

Thanks again for the feedback! Keep it coming!

Dory36
 
Ok. I tried it again and still got different results. BUT, this time I investigated a little further and here's what I found.

It appears that the order in which the data is entered may vary the results.

Using the data described in my previous post (in the new model) I entered the withdrawals first in the order of Yr 0 under "your social security", yr 12 under "spouce's social security" and yr 11 under "additional withdrawal change". This yielded a survival rate of 89.2.

Then I switched yr 0 and yr 12. Entering yr 12 under "your social security" and yr 0 under "spouces social security". This yielded a survival rate of 93.1.
Yes, I did remember enter the change for inflation adjustment indicator.

Interesting.

I tried it three times (doesn't mean I didn't make the same mistake each time).

PS. The same problem exists in the old version. That's why I was getting different results.

Hope this helps.

Moguls
(still trying)
 
Moguls,

THANKS! That identified a bug in the original, which I now located and fixed in the new version. I'll go back and fix it in the original as well.

Due to a programming typo, withdrawal 2 was using the inflation/no-inflation selection from withdrawal 1.

I fixed that in the new version now.

Also, I took your suggestion, and added 2 more changes in withdrawals over the years, and 1 more lump sum.

Please keep the feedback coming!

Dory36
 
The old version fix has been applied as well.

Try again, Moguls. Thanks for your feedback and patience.

Dory36
 
Wow! I like it. It appears to meet all of my needs.

One comment - with sample data in all the withdrawal and one time cash fields, the user is forced to override with zero if they don't have entries for every option. Would it be more intuitive if the default were zero for all the entries? If the user missed overriding one of the areas it could give some significantly misleading results. Not a problem for me (I use all the entries anyway). Just a thought.

I tried some trial data and like the new look and feel. I'll try my real data over the weekend and give you some feedback.

You do good work Dory36! I'll be sending a donation.

Thanks.

Moguls
(who appreciates all the help he can get)
 
Tried it, and it is more intuitive. I have a few suggestions and one request. First, optional withdrawal changes should have 0 defaults - no reason to touch the field if not needed. second, the old calculator (didn't check this one) claimed that PPI was more conservative than CPI, but the PPI option gave better survival % numbers than the CPI...seems backward.

Finally, the real request - please add some options for additional income streams. SS is nice to have in the model, but I also have a fixed pension and others may have annuities (not necessarily by unconstrained choice). It makes a significant difference in my spreadsheets how I model it. Modelling it as a reduction in needed dollars is not accurate. For example, if I have a pension of 40k/year, and withdrawal of 50k/year, then after 100% inflation I need a withdrawal of 100k/year. If I set my need at 10k/year by reducing the need, then after 100% inflation my need would be 20k/year - not accurate at all. If the pension were inflation adjusted, then just subtracting it works, but most pensions are not inflation adjusted. In the above example my withdrawal from assets starts at 10k/year and rises to 60k/year after 100% inflation.

Thanks for all the work you've done on this!

Wayne
 
Tried the PPI vs. CPI on the new test calculator, and it seems to be acting as I expect - higher survival % for the CPI. So ignore the comment in the previous post.

Wayne
 
Dory36,

In reference to Wzd's request could you clarify how the model calculates withdrawal impact by inflation?

It seems to me that there are two basic approaches. Both are valid depending on the individual circumstances.

The first approach would be to calculate the increase in inflation effect on the starting withdrawal (as stated in initial withdrawal field) then adjust the amount actually withdrawn from the portfolio based upon the withdrawal adjustments as specified in the social security and change withdrawal fields (after these fields were adjusted for inflation if appropriate).

This would cover the situation where one is assuming a fixed income stream is needed over the entire retirement period. In this approach the changes in withdrawal are treated like changes in cash flows rather than actual reduced expense needs.

It was my assumption (you know what assume makes!) that this was how the model currently works.

The second approach would be to allow for changes in the base withdrawal. In this senario, the changes specified in the withdrawal change fields (except for Social Security) would be subtracted from the initial withdrawal first and then the resulting amount increased for inflation.

This second alternative would allow for planned reduction in actual expense needs such as mortgage being paid off, kids finally off payroll, anticipated lower expenses needed at 80 vs. 45, etc.

As Wzd points out, these are distinctly different approaches (both valid depending on the circumstances). The idea model would allow for both, but the instructions need to be very clear how the underlying model is handling the withdrawal change impact.

Don't you just love it when everyone tells you how to design your system!!!

Thanks again, and keep up the great work. We appreciate it.


Moguls
 
Whoa!

Dory36, did you make a major change to the (old and new) model? I tried running some of my real data through and got dramatically different results in both of them, than I did just a day or so ago.

Moguls
 
Good points -- I'm resetting the defaults to zero where folks might overlook them.

There are now 5 different income streams possible, and 3 lump sum changes.

If you tried the "preserve values" yesterday, and notice that option is missing now, it's because some browsers had a problem with the feature.

Moguls, the inflation works like this:

First I figure out what the new withdrawal should be, based on the previous year's withdrawal and multiplied by the CPI or PPI change. (I save this withdrawal value for next year's calculation.)

Next I do the same for any optional withdrawal changes, if they are supposed to be inflation-adjusted.

Then I subtract both from the portfolio.

I see your point about how the withdrawal calculation is done; if the optional withdrawal is not inflation-adjusted, then the calculation takes too much out of the portfolio, and is therefore slightly conservative. If a mortgage is $10,000 a year, the additional inflation would be something like $200 to $2000 over the years. Not a lot in a year, but cumulative effects might add up. I'll look at changing that next time through the computation section.

Thanks again for the feedback! Keep those cards and letters coming!

Dory36
 
Moguls,

:( Yep -- as I mentioned a couple of days ago,
THANKS! That identified a bug in the original, which I now located and fixed in the new version. I'll go back and fix it in the original as well.

The bug affected the inflation calculation for one of the optional withdrawals. As I was tracking it down due to your report of differences depending on which order the withdrawals were entered, I found a typo in withdrawal 2. The same typo was in 2 different places (so much for copy and paste for function calls!). I fixed it in one spot 2 days ago; in the second spot last night. The fix was applied in the old as well as the new.

Also, the new version now includes 2001 data, but that shouldn't have much of a change in your outcomes.

I hope the change in your results wasn't TOO dramatic!


Dory36
 
It was very dramatic - from 90+% to <60%. Scared the crap out of me!

This didn't agree with my other calculations I had done so I knew there had to be more issues.

Fortunately I had also made an incorrect assumption that over compensated in the other direction. Bottom line was that the % survivable dropped by <5% after the fixes. Not good, but not disasterous.

Since we're redesigning FIRECalc for you :D , here's some more comments.

For the non-equity portion of the portfolio, you may want to consider providing some detail or a reference to information that would provide the historical averages for the options you allow. That would allow the user to pick the one that most closely matches their non-equity portfolio. Many of us have a blend and are unfamiliar with the historical averages of these options and therefore don't know which is the best option.

Hey, I like this designing on the fly and just passing on requests. I'm an old IT guy and now I know why my clients behave like this. I like it :D.

Moguls
(determined to RE)
 
This is reminisent of work - designing on the fly...and redesigning...and redesigning... lol

I think Mogul's note on the distinction between reduced living expense needs and alternate income streams was a good way of characterizing the two ways of looking at it. If I may be so bold as to suggest how to present it, I would use that exactly. A strong note that the first number is the living expenses needed, and then ways to adjust the needed living expense at key points in the future. A clearly seperate section for alternate income should also be there. The alternate income can be a pension (non inflation adjusted, lifetime), a side business (rental property prducing inflation adjusted cash flows for a period of years), social security, annuities, etc.

Retired early (age 49), with kids (5,14,16) and tring to figure out how to stay "retired". Avaya had a downsizing program that was too good not to take advantage of and I took it.

Wayne
 
The benefits of various choices of fixed income are completely beyond my area of expertise. Intercst's index (http://rehphome.tripod.com/chronidx.html) has a link to a planned 6/03 article on "Optimizing the Fixed Income Portion of a Retirement Portfolio", so that may help when it is published.

I'll look at the presentation of income questions -- thanks for that suggestion.

Regarding designing on the fly... I once worked for a computer company (NeXT) and part of my job was to go around and do what we called "public programming. This involved getting on a stage in an auditorium (usually on a college campus), getting a simple application together, and then redesigning it while an audience looked on and perhaps gave suggestions.  

At least this way, via the forum, I can pretend not to have read the message while I try and figure out how in the heck I am going to make the change! :p

Dory36
 
Ok. One last suggestion (at least until the next one) :D.

In addition to the portfolio totals in the detail, it would be nice to see the actual withdrawal amount for each year. For circumstances where many adjustments were made to the withdrawal rate (not to mention the inflation impact), it would be handy to see these details.

And I like Wzd's suggestion on how to present the two alternatives of changes to withdrawal requirements and additional income streams.

Thanks again for the great calculator.

Moguls
 
He's baaack!!!

Ok, in addition to the above suggestion on the output. I have another suggestion and a question.

First the suggestion. I occassionally print the summary output page (and I assume others do as well) so it would be good to show on the summary page the selected interest rate for tips, if one was entered. Otherwise you have to remember and write it in.

Now the question. Does the model do anything special with Tips or is this just a way to specify a fixed interest rate?

Moguls
 
hmm... i looked over the detail data for a run and saw for 1929: yr30 : 420, yr31 -37089, yr32 -11417, yr 33 14580, ... yr 40 8704, and the line is not colored red. However the failure rate was supposed to be 6.9% out of 130 or 9 failures. Only 6 were colored red, so I suspect a display bug. However, the numbers are interesting - what a way to recover :)

I used defaults + wd=34000, nestegg=650000, 40yrs, ss 14400 in 15 yrs.

Wayne
 
Only 6 were colored red, so I suspect a display bug. I used defaults + wd=34000, nestegg=650000, 40yrs, ss 14400 in 15 yrs.

Yep -- there is a display bug. The addition of SS allows the portfolio to go back to positive after dipping below zero -- something that would never happen in a plain old withdrawal scenario.

The percentage should be right, but I need to set the "red line" logic to go red whenever ANY year is negative, and not just the ending year.

Thanks for spotting that.

Regarding the other changes -- with the Wall Street Journal mention of FIRECalc yesterday, we're getting a LOT of new users, so I think I will not add features for a few weeks. But keep the suggestions coming -- I'll add them for a next release.

Dory36
 
From an email exchange:
It wasn't clear from the BB notes - are the adjustments from the income for SS and the 3 others done by adjusting the need or subtracting from the needed cash? It should make a difference as inflation adjustments should be applied to the need $$; for inflation adjusted income streams I guess there is no difference, but for non-inflation adjusted streams there is a difference.

The adjustments for income are as follows:

The current year amount for each withdrawal/income stream (and, if scheduled, lump sum) is individually calculated, and all are summed. The sum is subtracted from the portfolio.

If I follow your question, the difference and potential error would be as follows:

Both approaches would start the same:
1,000,000 Year x portfolio
-40,000 Year x w/d
+10,000 Year x income
970,000 Year x balance

The way FIRECalc works now:
1,060,000 Year x+1 portfolio, with growth
-41,000 Year x+1 w/d, with inflation at 2.5%
+10,000 Year x+1 income (not adjusted for inflation)
1,029,000 Year x+1 balance

With a spending reduction before application of inflation:
1,060,000 Year x+1 portfolio, with growth
-40,000 Year x+1 w/d, without inflation
+10,000 Year x+1 income (not adjusted for inflation)
-30,000 Year x+1 pre-inflation spending
-750 Year x+1 inflation adjustment to spending
1,029,250 Year x+1 balance

As you noted, this $250 difference (in this example) would only apply when the $10,000 additional income was fixed, not adjusted for inflation.

Of course, if the fixed income is substantial in comparison to the "normal" withdrawal and portfolio, the seemingly small difference can matter over decades, so my next time through the computation code, I will likely change the code to make the adjustments to the withdrawals before the effects of inflation are applied.


And in response to Moguls:
First the suggestion. I occassionally print the summary output page (and I assume others do as well) so it would be good to show on the summary page the selected interest rate for tips, if one was entered. Otherwise you have to remember and write it in.

Now the question. Does the model do anything special with Tips or is this just a way to specify a fixed interest rate?
Good idea on the summary page -- I'll do that next change.

On the TIPS, the Treasury pays a coupon (that you enter on FIRECalc, depending on what they are offering), AND they adjust for inflation. So a 3.5% TIPS bond might return 6.5% one year, 5.9% the next year, and so forth. But if inflation goes to 20%, the TIPS would return 23.5%. That's their attraction. There's a lot more to them than I can relate here, but that's the view from 10,000 feet.

By the way, I'm inclined to "close" this topic and move the continuing discussion outside of the "testing" subject and into a new topic. New readers would probably like to follow along without reading so much history. So if it is closed tomorrow, you'll know why!

Dory36
 
will do. I am writing up a summary of what I would request in a new topic right now.

Wayne
 
This comment belongs here more than in the new topic - I would recommend changing the current wording to explain the way you calculate income streams today. I had assumed the opposite of what you said above.... Although the way you explain it is exactly what I need.

And regarding the comment "if I understand the question right", you do understand the question exactly...See the example in the new post for a scenerio based on my position.

Wayne
 
Status
Not open for further replies.
Back
Top Bottom