New Monte Carlo Based Online Planner

[LOL!]

Awesome client support!

In the output, the "Percent Expenses Funded" value (tracked per year) gives you an idea of how much pain (in the form of belt tightening) the plan is imposing on you in order to get the probability of success that's shown.

How so?  My output always says "97% - 115%"  no matter if I have allowed for 50% or not.  I guess you are writing that 97% is the median of 10,000 runs?

Finally, if you click show "% of expenses funded" just above the graph, you can see the year-by-year effects of the spending policy overlaid on the graph.

There is no Y-axis label for this graph, so I cannot tell what the plot shows.  Same goes for "% of failures" which I guess would use the same y-axis label.

Here's more of my inner thoughts: I am thinking that I could survive on $60K a year, but I usually spend $100K a year.  I could cut out $20K of eating out and $20K of vacations pretty easily I think.  So would I then making the spending adjustment multiplier some huge number?

 

[FRPGuy]

How so? My output always says "97% - 115%" no matter if I have allowed for 50% or not. I guess you are writing that 97% is the median of 10,000 runs?

Yes, the report and graph shows the median value, so you can't see the handful of really bad trials where you cut expenses by 50% and still ran out of money. Also, for more detailed data than just a range, make sure you click the Detailed Output tab and look at the year-by-year data (that's the same data that gets graphed).

There is no Y-axis label for this graph, so I cannot tell what the plot shows. Same goes for "% of failures" which I guess would use the same y-axis label.

Yeah, I've gotten a few complaints about this, but the graph is so cluttered already, I hate to add more text to it. Also, if you click on the graph, the dashed red line will move to where you click and the details to the right of the graph update to show the data for the selected year. There you can see the value for % expenses funded. If you click-drag the red bar across the graph, the right hand data will update as you drag.

Here's more of my inner thoughts: I am thinking that I could survive on $60K a year, but I usually spend $100K a year. I could cut out $20K of eating out and $20K of vacations pretty easily I think. So would I then making the spending adjustment multiplier some huge number?

Yes, you can try that, but in the testing I've done it seems that the spending multiplier can only boost the success probability by a few points at the most. Maybe there are cases where it really helps. Also, if you're going that route, you probably should set the ceiling close to 100 or 120%.

 

[FRPGuy]

OK, maybe a dumb question but...
If I take the results of this panner and follow it as my retirement plan, how will I know when I am in a "worst case scenario" (according to the planner) and should cut my spending back to 50%?

Not dumb at all. In the research, this is one of the things they talk about as a constraint on how fancy you can make the so called "decision rule".

I think the decision rule that's used in the FRP could easily be used by a retiree or their financial planner. In any year where your portfolio balance is less than last year and less than it was when your retirement started (in real dollars), you don't get a cola that year.

The floor is a little tougher to implement in real life since the dollar amounts will all be moving around due to inflation. A financial planner could easily do the calculation, but non-quant types might need to wing it and make something up for the floor - (like if you have to buy cat food to eat you've hit the floor). You could keep a notebook and track your annual "percent of spending funded" and track the inflation adjusted value of your original spending amount. Tracking the adjusted spending amount from year to year just means taking last year's value and multiplying it by (1-inflation rate) in any years where you need to subtract, and (1+inflation rate) in any years where you need to add.

 

[mhoffman]

Hello: I have some questions re the flexible reitirement planner. In one post, it says that all the numbers are inflation adjusted. I ran a scenario and in the planned expenses column it shows the amount I entered for planned retirement spending constant (uninflated) over the entire time period. What's up? Here are some other things I don't understand:

o I entered social security, pension-no cola and miscellaneous income in the additional inputs section. The detailed output shows after tax income going down by about .3-.5%/year. What's going on here?

o I ran a scenario with an 11.7% return and 8.9% standard deviation (dfa 50 portfolio). The balance of my investments increased 12-14%/year in the first few years and after that was in the range of 4-7% increase/year. I can't figure out this pattern. I'm by no means an expert, but if the model were generating random returns based on the std dev, wouldn't the range of returns be much greater? What gives?

o How many cases does the model run through? It gives an xx.x% probability of success, which implies multiple iterations, but it doesn't say how many iterations.

o The "detailed output section", "median portfolio balance" column shows my portfolio increasing by 90% after 5 years (same return, std dev as above). This seems way out of line. When I ran 100 scenarios with a simple model using a random number generator, I get a median increase of about 30% over 5 years. What gives?

Thanks for your help.

M Hoffman

 

[FRPGuy]

M Hoffman,

Thanks for the detailed questions. These are a great help in making the planner better...

it says that all the numbers are inflation adjusted. I ran a scenario and in the planned expenses column it shows the amount I entered for planned retirement spending constant (uninflated) over the entire time period. What's up? Here are some other things I don't understand:

o I entered social security, pension-no cola and miscellaneous income in the additional inputs section. The detailed output shows after tax income going down by about .3-.5%/year. What's going on here?

Maybe what I should have said was that all values are entered and displayed in 2006 dollars. That means that in the output tab, I'm discounting all the amounts back to 2006 before displaying them.

In the case of the non-cola adjusted income cashflow, the "real" value of that cashflow (in 2006 dollars) should go down by the inflation rate each year. I know this is confusing, since the actual payout amount will be fixed, but this is the nature of dealing with the time value of money. The output column is showing the value of each payment in terms of what it's worth in 2006 dollars.

Does that make sense or do I need to try again?

o I ran a scenario with an 11.7% return and 8.9% standard deviation (dfa 50 portfolio). The balance of my investments increased 12-14%/year in the first few years and after that was in the range of 4-7% increase/year. I can't figure out this pattern. I'm by no means an expert, but if the model were generating random returns based on the std dev, wouldn't the range of returns be much greater? What gives?

I'm not sure what other cash flows are configured in your plan, but perhaps there's an expense cash flow that's kicking in later in the plan that's keeping the portfolio growth down. One thing to note is that by default, the "Retirement Expenses" value on the main input section starts getting subtracted from the portfolio balance in the first year of retirement and then every year for the rest of the plan. You can zero this out if you like, and manually enter your retirement expenses (and timing) on the "additional inputs" tab if you need more granularity.

o How many cases does the model run through? It gives an xx.x% probability of success, which implies multiple iterations, but it doesn't say how many iterations.

yeah, I probably should say that somewhere in the app. The number is 10,000. In my defense, the number is shown a bunch of times in the documentation on the web site - on the other hand, I should know that no one reads that!

o The "detailed output section", "median portfolio balance" column shows my portfolio increasing by 90% after 5 years (same return, std dev as above). This seems way out of line. When I ran 100 scenarios with a simple model using a random number generator, I get a median increase of about 30% over 5 years. What gives?

Could this be another case of the default cashflows confusing things? By default, there's several thousand added to the portfolio every year until retirement.

I set up the defaults so you could click "run simulation" without typing anything in and see some results. maybe that was a bad idea, if what's burning you are all these preset cash flows. Or maybe I'm barking up the wrong tree and something else is wrong.

Jim

 

[FRPGuy]

On a related note, I posted a bunch new stuff to the web site including a much expanded FAQ and documentation section, a page with some "safe withdrawal rate" research links (you folks have probably read them all), and snippets of source code from the planner (gulp!).

The idea on the source code posting was that there's a lot of criticism of planners generating differing outputs because of different assumptions. I wanted to answer that and maybe try to prod others to let their code be subjected to peer review.

While the code I posted doesn't show everything, it does include the heart of the MCS loop and the support routines. I also documented the random number methodology (and code).

Anyhow, I wonder what people think about the idea of encouraging sw vendors to make their planner code (or at least pseudo-code) public.

Jim

 

[FRPGuy]

OK, last post...

I wanted to remind people that there's an email address on the planner web site if they want to contact me with questions or comments about the planner.

I love answering questions here, but I'm sensitive to the fact that I'm using dory36's board to sort of hawk a competing product. I have been trying to follow a "don't speak unless spoken to" model, so people wouldn't think I was trying to keep this topic up high on the list.

In any case, my fear is that I'll violate the ediquette and get slapped down, and frankly that would hurt. This is a great board and I respect this crowd and your views.

So if some of the more senior members want to provide some guidance, or gently tell me to buzz off, I'll respect that.

Jim

 

[REWahoo]

So if some of the more senior members want to provide some guidance... I'll respect that.

I'm sure you've made a healthy financial contribution to Dory in support of his board, right?

If not, here's the link: http://www.fireseeker.com/results.php

 

[donheff]

Anyhow, I wonder what people think about the idea of encouraging sw vendors to make their planner code (or at least pseudo-code) public.

Jim

Transparancy is alway good -- "a thousand eyes," or whatever the open source quote is. Maybe you can be the first guy on the block to sponsor an open source retirement planner application development project - get a few financial software gurus on the team to tweak. Put the project up on Source Forge, good to go.

 

[Delawaredave]

When I try to run the calculator, I get a big blank gray square in a blue border.

I clicked on the java download - downloaded it, and the check says OK.

Any idea what I'm doing wrong ?

Thanks.

 

[Delawaredave]

When I try to run the calculator, I get a big blank gray square in a blue border.
I clicked on the java download - downloaded it, and the check says OK.
Any idea what I'm doing wrong ?
Thanks.

I have it working now.  Windows XP.  The program works under Internet Explorer, but does not work under Maxthon browser.   Maxthon is a tabbed browser shell over top of Internet Explorer (great browser http://www.maxthon.com).

It's very rare that I find a page that works under IE but not under Maxthon.   Oh well.

Now to run the tool.

 

[obryanjf]

FRP guy

Only speak for myself but I think you are doing it right.  I am writing code in C++, to create an executable.  It is to be for my pleasure and benefit and if others care to use it then great.  I was also intending to include the code so others that gave a damx could review it.  I've run yours once and don't have much time now to devote to yours or finishing mine.

good luck.

job

 

[nellieb]

I get the same blank screen with an x.

 

[Tadpole]

Tadpole,

Thanks for the Java dump.  (I Think -  )  I knew I'd get one of those eventually. 

Unfortunately (or maybe fortunately), the error is before the planner got launched by Java.  It's complaining of a versioning error trying to launch the main class, but I don't exactly understand what's going on.

One thing to try would be to close all your browser windows, and relaunch the planner from the home page.

If you could tell me what type of system (OS) and browser  type/version you're on that might help too.  I can see from the dump that your plug-in version is 1.4.2 which should be fine (the latest is 1.5).

Also, is this your first visit or a return visit?

Finally, if you'd rather you can email me at info@flexibleRetirementPlanner.com

I do appreciate the information and am eager to get to the bottom of it.

Thanks,

Jim

Jim, I did a google search for the problem. I came to the conclusion that this is either caused by something you did or did not do to tell the program to allow execution in older versions or it was caused by the version of 1.4 that I have. Since I found out that 1.4 has some problems that 1.3 doesn't even have, I updated to 1.5. That fixed the problem but you might want to research some more on this error as you might be able to easily fix it for all versions. The x that people are seeing is probably the same problem as that is what the screen shows when the error occurs. Have them open their Java console and look.

I ran the program and decided I needed to read the manual to confirm what I think your inputs are. Why the conversion of 0.5% as a std value to 50%? It would help if each field of input had either a pop up or mouse passover explanation.

 

[FRPGuy]

Tadpole,

Thanks for the help on this. You're 2 for 2.

1) It turned out that I inadvertently set my development tool (JBuilder) to require version 1.5 of the Java VM. The bad news is that all of my testing is with 1.5, so before this update it wouldn't work with an old version, now it might work badly. I don't think I'm using any of the 1.5 Java goodies, so it should be ok. I guess I should have read the exception dump more carefully.

2) As far as the percent parsing stuff, I'm calling that a bug. I was trying to model the way Excel does percentage input in a cell that's formatted as a percent. I guess I never tested a std dev of .5%. I must be more pessimistic than you as far as volatility goes

Anyhow, it should be better now.

I posted an updated version and as usual, beware that the browser may cache the old version. The source changes were tiny, so hopefully they didn't mess anything up.

Thanks again,

Jim