FIRECalc question

[Rustic23]

As I am retired, I did not vote. I would concur that the type of retirement income, cola'd pension vs 401k would have an effect. However, I would also think the faith one has in FireCalc, the 4% rule and other guidelines also if a factor. The current economy, while bad, is not as bad as previous economies. It may get there, but I for one do not think so.

It also brings me to the point about FireCalc/4% I don't understand. If I had $1M in 2008 and needed $40K to retire, I am good. However, in 2009 I only have $600k. If I were going to retire and need $40K, I can't retire, but the guy that retired in 2008 is OK. We both need the same amount, we both have the same amount? It does not seem like 1 year should make that much difference.

basic firecalc answer
1,000,000 need 40,000 for 30 years 94% chance of success
600,000 need 40,000 for 29 years 39% chance of success!
980,000 need 40,000 for 29 years 94% chance of success!

 

[REWahoo]

It also brings me to the point about FireCalc/4% I don't understand. If I had $1M in 2008 and needed $40K to retire, I am good. However, in 2009 I only have $600k. If I were going to retire and need $40K, I can't retire, but the guy that retired in 2008 is OK. We both need the same amount, we both have the same amount? It does not seem like 1 year should make that much difference.

Remember that FIRECalc takes your starting balance and runs it against historical market performance to answer the question, "will I run out of money in X years, spending Y per year invested in Z asset allocation".

If you retired with $1M in 2008 and FC said you are 100% successful at 30 years that means your portfolio didn't go to zero through the worst downturns on record. If you dropped to $600k in 2009, one year into retirement, you've already begun to experience a downturn and, unless this is worse than anything in FIRECalc's history back to 1871, your portfolio will survive.

You might stain your shorts while doing so, however.

OTOH, if you started with $600k and then experienced one of the big historical downturns, FIRECalc tells you what the odds are your portfolio will recover and last you 30 years. Not 100% unless you reduce your spending or die sooner...

 

[Architect]

It also brings me to the point about FireCalc/4% I don't understand. If I had $1M in 2008 and needed $40K to retire, I am good. However, in 2009 I only have $600k. If I were going to retire and need $40K, I can't retire, but the guy that retired in 2008 is OK. We both need the same amount, we both have the same amount? It does not seem like 1 year should make that much difference.

basic firecalc answer
1,000,000 need 40,000 for 30 years 94% chance of success
600,000 need 40,000 for 29 years 39% chance of success!

That's a very interesting question! Initial capital is obviously critical. I think what you are seeing is that since a monte carlo simulation is statistical, it's 'not expecting' that in the first year you ER the portfolio drops by 40%. It predicts that the first year will be some average of the previous 100.

But reality is a little hard, and in this case the $1M ER DID get a 40% hit (a somewhat unusual event), which turns out is really bad. The chances of outliving your money is much higher, and the calculator can now give an updated, more accurate answer. Think about the flip side, say stocks jumped by 20% the first year of ER. Run Firecalc again - the chances of survival go way up. Bernstein talked about this - how your investments behave during the early part of your retirement are critical.

It also brings up another important point - it's a good idea to run Firecalc every year, with your updated capital, and take that as your new baseline as if you had just newly retired.

PS. This highlights the importance of a conservative portfolio going into retirement. Early losses can be devastating.

 

[LOL!]

...
It also brings me to the point about FireCalc/4% I don't understand. If I had $1M in 2008 and needed $40K to retire, I am good. However, in 2009 I only have $600k. If I were going to retire and need $40K, I can't retire, but the guy that retired in 2008 is OK. We both need the same amount, we both have the same amount? It does not seem like 1 year should make that much difference.

basic firecalc answer
1,000,000 need 40,000 for 30 years 94% chance of success
600,000 need 40,000 for 29 years 39% chance of success!
980,000 need 40,000 for 29 years 94% chance of success!

The guy with $600K fell into the 6% failure mode. Note that 6% is somewhat larger than 0%.

 

[REWahoo]

That's a very interesting question! Initial capital is obviously critical. I think what you are seeing is that since a monte carlo simulation is statistical,..

FIRECalc doesn't use a Monte Calro simulation unless you intentionally select that option. See How FIRECalc works..

... it's 'not expecting' that in the first year you ER the portfolio drops by 40%. It predicts that the first year will be some average of the previous 100.

Not correct. Once again see How FIRECalc works.

It also brings up another important point - it's a good idea to run Firecalc every year, with your updated capital, and take that as your new baseline as if you had just newly retired.

Once again, not correct. See How FIRECalc works..

PS. This highlights the importance of a conservative portfolio going into retirement. Early losses can be devastating.

On this we can completely agree.

 

[FIRE'd@51]

It also brings me to the point about FireCalc/4% I don't understand. If I had $1M in 2008 and needed $40K to retire, I am good. However, in 2009 I only have $600k. If I were going to retire and need $40K, I can't retire, but the guy that retired in 2008 is OK. We both need the same amount, we both have the same amount? It does not seem like 1 year should make that much difference.

basic firecalc answer
1,000,000 need 40,000 for 30 years 94% chance of success
600,000 need 40,000 for 29 years 39% chance of success!
980,000 need 40,000 for 29 years 94% chance of success!

When you run Firecalc with 600K for 29 years, you are running it through all the 29 year periods, including those that started when the market was higher. To really get a handle on this, I think you would have to single out the paths in Firecalc that began after a large market decline. Unfortunately, there are only be a few of them, so there would be no statistical significance.

 

[ziggy29]

To really get a handle on this, I think you would have to single out the paths in Firecalc that began after a large market decline. Unfortunately, there are only be a few of them, so there would be no statistical significance.

I tend to agree. The only way to get enough data points is probably a Monte Carlo simulation, but even that has its own issues.

 

[Architect]

Once again, not correct. See How FIRECalc works..

Regarding re-running the simulation with updated data, take an extreme example for illustration. Somebody starts out with $1M and the next year it crashes to $40k (like I said - extreme example). So now that retiree can continue to spend his SWR 4%/$40k per year secure in the knowledge that he started with a 97% likelihood of success? Is it much different for the guy who suffered a 60% crash, or 40%? In any simulation updated information improves your predicitability.

Reminds me of the smart WS folks who priced the CDO's with the knowledge that house prices haven't declined in the US on a nationwide basis (in the post war period). Once it was clear that house prices were declining, does that mean they don't have to re run the models incorporating that new fact?

 

[FIRE'd@51]

I tend to agree. The only way to get enough data points is probably a Monte Carlo simulation, but even that has its own issues.

Yep, a MC simulation is truly independent of what happened before, so it has no way of knowing there was a previous large market decline. The only way that I can think of to offset this is to raise the expected return (i.e. the distribution average) to a number more representative of the types of returns following large declines.

 

[REWahoo]

Regarding re-running the simulation with updated data, take an extreme example for illustration. Somebody starts out with $1M and the next year it crashes to $40k (like I said - extreme example). So now that retiree can continue to spend his SWR 4%/$40k per year secure in the knowledge that he started with a 97% likelihood of success?

If the portfolio dropped from $1M to $40k, then I'd say the retiree just experienced the 3% failure rate FIRECalc gave him.

In any simulation updated information improves your predicitability.

Only if you understand how the simulation works and what the simulation is telling you.

Reminds me of the smart WS folks who priced the CDO's with the knowledge that house prices haven't declined in the US on a nationwide basis (in the post war period). Once it was clear that house prices were declining, does that mean they don't have to re run the models incorporating that new fact?

No rational retiree would put their head in the sand when faced with a significant hit to their portfolio. But re-running FIRECalc with new numbers a year or more after retiring isn't the right way to determine what adjustments you need for financial survival.

You might want to check out a few threads on the FIRECalc Support forum and do a search for FIRECalc discussions elsewhere on the board. This subject has been discussed on several previous threads.

 

[Rustic23]

OK,
Here is a thought. It may be the difference between possibility and odds.

Example, A guy that sits down at a blackjack table with $1M has a better probability of leaving with say $500k than a guy that sits down with $500k. However, if both are setting there with $500k they both have the same odds on the next hand and on future hands. That is assuming they are of equal skill. I am sure there might be other things I might have to be the same, but it was good enough for me.

Having said that, I don't have a clue how it would relate to FireCalc or retirement!

 

[Kabekew]

But re-running FIRECalc with new numbers a year or more after retiring isn't the right way to determine what adjustments you need for financial survival.

Doesn't that violate fundamental statistics (Bayes theorem?) My original withdrawal rate and likelihood of success (if I'm assuming past performance will be repeated) is a prior probability compared to my updated withdrawal rate the following year -- and by Bayes theorem my standard deviation would continue to increase over time if I stuck to that initial predicition, compared to continually refining it given my updated portfolio balance (until eventually reaching 0 in my last year)?

That's pretty fundamental to probability theory, but I think there's a disconnect here because the number given by FIRECalc only represents how a portfolio would have performed in the past. And true, for analysis of portfolio performance against a finite set of data points, you only have to "perform" it once.

But statistically speaking, if you assumed past performance indicates some amount of predictive capability of future performance (correlation > 0, whatever it may be), then using the output of FIRECalc would give you a probability P with standard deviation O for a given withdrawal rate, let's say, and running it again the following year for a period one year less than the initial, given the new information (new portfolio balance after one year), according to Bayes theorem the only way you could get the same P and O for the same withdrawal rate would be if the liklihood function was 1, i.e. you could predict the future with absolute certainty. You can't, therefore for better accuracy you would necessarily need to run FIRECalc again continuously.

The FIRECalc page though says outright it doesn't attempt to predict future results, so maybe it's moot. But I think some people use it with some assumption it may predict future performance -- if so, mathematically speaking it's more accurate to refine it periodically.

 

[REWahoo]

The FIRECalc page though says outright it doesn't attempt to predict future results, so maybe it's moot.

Yep.

However, rerunning your numbers and wringing your hands does provide an occasional diversion while you're worrying about having to go back to work...

 

[REWahoo]

BTW, for those who do like to run new FIRECalc numbers again and again after retiring, Dory36 was kind enough to provide a streamlined version just for you: FIRECalc Simple