Why Firecalc isn't good enough (and other comments)

[SnowballCamper]

In another thread a poster wrote a good question, and that thread got me thinking more broadly. I've gathered up some thoughts to share, but here's the initial question:

"DCA allows you to buy more on dips and less on peaks, it makes volatility your friend during accumulation.
Can't the process be run in reverse, selling more on the peaks, selling less on the valleys?"

I think the reverse process would be to sell a fixed number of shares each time period, but this yields more money when the market is up and less money when the market is down. It isn't quite selling more on the peaks. In order to sell more on the peaks you have to know it is a peak e.g. you would have to know the future.

That thread started with some conclusions made from a back-test of market data, and many members pointed out the problem with back testing. The problem is that a subset of the data is selected, and conclusions drawn from that subset don't reflect all the data.

Many of us prefer to use Firecalc, and this is much better than back testing from a subset of the data. However, Firecalc is really just a super backtest using all the readily available data over multiple time periods. When we properly interpret the result as a summary of historical data things are good.

The problem is that the result is presented in a single percentage, and percentage is also used to quantify probability. It's just too slippery to slide from "90% of 30 year periods in the data set resulted in retirement success" to the erroneous, "I have a 90% chance of a successful retirement." Reading a percentage should always be followed by the question "Of what?" and answered accurately.

Since we do have to plan for the uncertain future, a look at history is certainly prudent, but we should do more. A popular approach is to do a simulation where one or more market mechanisms is modeled with a probability distribution (random number generator). This approach has the conceptual advantage of recognizing that past performance is not necessarily indicative of future results. However, a common criticism starts with the question, "Well, has that ever happened before?" And of course the answer is no. But the whole purpose of the simulation is to look beyond what has happened and to consider what could happen in a mathematically rigorous way.

Mathematical rigor is difficult, and so criticisms come in sloppy, misleading statements such as, "Since standard deviation applies only to Gaussian distributions, we get breathless reporting of "100 year" market declines."

The real criticism, popularized by Nassim Nicholas Taleb, is that the probability distributions used in the model are unjustified because time and time again reality demonstrates events that were extraordinarily unlikely from any probability distribution with a finite variance. This is the shotgun blast through the "mathematical rigor," but the difficulty remains. To get a glimpse of the difficulty, find a probability distribution visualizer online such as https://statdist.ksmzn.com Then look at the Normal (Gaussian) distribution and compare it to the Cauchy distribution. You can fidget with the parameters to make them look almost the same. You'll find the the variance for the Cauchy distribution is described as "undefined," a euphemism for infinitely large. The standard deviation is the square root of the variance and certainly applies to more than just the normal distribution, as the variance is listed for every other distribution in the visualizer.

So what should one do? Thankfully, a member of the forum pinned me down on this some years ago. I think the best way to evaluate a retirement plan from a quantitive perspective is with the following equation

(Retirement Assets) / (Annual Spending) = Years of funded retirement

But it must come with two questions:
How are you going to grow retirement assets to keep up with inflation?
And How are you going to grow retirement assets to last until you die (presuming you expect to live longer than you have currently funded)?

This approach invites the users to examine what they have, what they want to spend, and in most cases will have a stark difference between how many years are funded, and how much longer they expect to live. It invites many more questions with "How...", but this is a feature. The user has to think about the answers instead of just wondering if ninety whatever percent is good enough.

 

[NW-Bound]

We don't have enough history of the market return to make a really meaningful model.

And it is not likely that this random process is stationary, meaning its statistical properties do not change with time. History has shown that past empires peaked then declined, and so did their economies.

And there are so many extraneous variables. I have not seen REWahoo mention asteroids recently, but even lesser events such the recent pandemic could throw a curve ball to any model. And there are so many unpredictable political developments, both domestic as well as international.

All I can do is to have a lot of reserve, and deal with life as the future unfolds.

 

[DrRoy]

I would say that the main reason FC is not mathematically predictive is that the future will not look like the past.

 

[marko]

I would say that the main reason FC is not mathematically predictive is that the future will not look like the past.

I've posted a few times here on this. I don't see how going back to the 1950s markets, let alone the early 1900s is helpful. Even back to 1970, who would've guessed that the likes of GE, Exxon, Philip Morris et al would be so different. We now have mutual funds, ETFs, AI, hedge funds and hundreds of non market cultural differences as well.

Those comparisons look back through an Industrial Revolution lens, a time that we, and our markets no longer inhabit.

In short, our world and the markets that react to our world is entirely different than even a few decades ago. While I fully appreciate the FC outputs, I always wonder about the value of looking back 50 or 100 years to such a different time...how relevant can that analysis be? We might as well look to ancient Rome for comparison. IMO.

 

[SunGuy]

Aside from unexpected medical costs, a lot of people actually end up over-estimating how much they need for retirement.

My parents have drastically reduced what they felt were otherwise pretty tame spending habits while working.

The cost of working and daily commutes really does play a huge factor in costs.

I always let people know about the book 'Your Money or Your Life' which has a great way to analyze your work costs.

As the modern shift for white collar jobs lets more people work hybrid, and fully remote jobs, people will be able to get a better understanding of their fixed costs.

My grandparents on both sides were the same way, they both would retire in their 50s before I was born, except my grandfather and grandmother on my mother's side worked some, he as a lawyer doing the odd will/estate planning, and she would sell a few houses a year as a real estate agent.

And my grandparents on my father's side had real estate that they built up while retired, up to 5 SFH's and had a part time business that made maybe a few $k a year.


I had a fun conversation with my parents as they went on another trip, their 8th this year, they apparently had budgeted as if they were still going to need all the included costs, and have done very well in the markets, thanks in part to a very good financial advisor.

They have at least an extra $1k a month coming in from their accounts, and are just going to save it for future medical costs.

They are very happy to be in a place of over-preparing for retirement, and now under-spending their nest egg.

 

[Gumby]

As the famous saying goes "It is hard to make predictions, especially about the future." The only thing I know for certain is that in all places and at all times, it has proven better to to have more money than less money. That's about the extent of my retirement planning.

I know how much secure retirement income I have (but I must assume that it will continue). I know for sure how much I spend now and I can make a relatively safe prediction of how much I will spend in retirement (and it has proven to be fairly accurate over 4 years of retirement). But I can make only the roughest prediction of investment returns and I can only put a reasonable outer bound on my life expectancy.

So yes, there is uncertainty. There always has been uncertainty and there always will be uncertainty. The data is uncertain and the models are all wrong (but some are useful). If you cannot act in the face of uncertain, incomplete, vague and/or ambiguous information, then your safest bet is to either keep working or commit to offing yourself the day you run out of money.

 

[marko]

As the famous saying goes "It is hard to make predictions, especially about the future." The only thing I know for certain is that in all places and at all times, it has proven better to to have more money than less money. That's about the extent of my retirement planning.

I know how much secure retirement income I have (but I must assume that it will continue). I know for sure how much I spend now and I can make a relatively safe prediction of how much I will spend in retirement (and it has proven to be fairly accurate over 4 years of retirement). But I can make only the roughest prediction of investment returns and I can only put a reasonable outer bound on my life expectancy.

So yes, there is uncertainty. There always has been uncertainty and there always will be uncertainty. The data is uncertain and the models are all wrong (but some are useful). If you cannot act in the face of uncertain, incomplete, vague and/or ambiguous information, then your safest bet is to either keep working or commit to offing yourself the day you run out of money.

+1 After all the bills are paid and my year-end balance is (even slightly) larger than my year-start balance. I'm in good shape and happy.

Some years are better than others, but I tend to look at a 3 and 5 year average. We live on our dividends (and SS) so the market fluctuations are less of a concern. Having said that, our balance has doubled over the past 18 years since ER.

 

[Midpack]

So yes, there is uncertainty. There always has been uncertainty and there always will be uncertainty. The data is uncertain and the models are all wrong (but some are useful). If you cannot act in the face of uncertain, incomplete, vague and/or ambiguous information, then your safest bet is to either keep working or commit to offing yourself the day you run out of money.

+1, that’s what it comes down to no matter what, there is no approach that is “good enough.” The OP was interesting, but I think FIRECALC is outstanding for the price.

After spending years looking for the best retirement spending scalpel, I came to the conclusion some of the simplest axes are as good as any. I stumbled on Pfau’s Funded Ratio a few weeks ago, it’s much like the OP’s simple equation. You project lifetime assets / lifetime spending and the % is an indication of your probability of success. If you’re at 100%, you’re going to have to be prepared to make substantial adjustments during retirement - up or down. If you’re at 200%, you have the opportunity to spend more, but how much and when remains uncertain.

There’s no removing uncertainty. Some of us will be end up with less than expected, some with more, some near guesstimates - and most of us will have to course correct until the end (unless our goal is a large $ legacy). And we all have to decide what cushion we’re comfortable with.

 

[marko]

After spending years looking for the best scalpel, I came to the conclusion some of the simplest axes are as good as any.

+1 Man, what a great quote! Concise and spot on.

Kudos!!

 

[REWahoo]

After spending years looking for the best retirement spending scalpel, I came to the conclusion some of the simplest axes are as good as any.

+1

"Measure with a micrometer, mark with a grease pencil, cut with an axe" is the reality of retirement planning vs. the retirement experience.

 

[Rianne]

I have questions about FireCalc. The bond section doesn't give enough options and isn't specific about what bond allocation you have. Interest rates are so different than even a year ago. Our bond ladder should put us at a higher success rate than if I simply entered 50% 5-year treasuries. What interest is FireCalc using?

 

[RS53576GG]

FireCalc provides an exact answer for what happened in the past. It has no idea what will happen in the future. So what good is it? I used it as a way to get an idea how much I'd need to retire with confidence that I won't run out of money before I run out of health

Here was my process:

1) Estimate how much I'd want to have as retirement income. Generously overestimate everything.

2) Estimate how long I could live. Generously overestimate this too.

3) See how much money I'd need to have a 100% success ratio.

4) Multiply the needed money by 1.5 at the least, preferably 2.

5) Work hard to get that much money put away.

6) Start goofing off.

It wasn't easy to accomplish Step 5 but I'd rather work hard today and avoid being broke and homeless if the worst case scenario turns out to be be worse than anyone ever imagined.

Conclusion:

Using FireCalc for detailed analyses may not be 100% dependable so it's probably a good idea to use it for general planning and leave it at that.

 

[OldShooter]

FireCalc provides an exact answer for what happened in the past. It has no idea what will happen in the future. So what good is it? I used it as a way to get an idea how much I'd need to retire with confidence that I won't run out of money before I run out of health ...

I don't use it, but I think that class of gadgets, properly understood is useful.

Unfortunately I interpret evidence here to indicate that most people don't understand these gadgets or their limitations. Too much agonizing over the difference between 100% and 90% for example. IMO those two numbers are effectively the same. But obviously it makes money for the site operators, so I guess "All opinions guaranteed worth price paid." applies.

 

[NW-Bound]

... The only thing I know for certain is that in all places and at all times, it has proven better to to have more money than less money. That's about the extent of my retirement planning...

Hear, hear... What I also like to have more of is health, and handsomeness. These are way harder to get. But money is easier to accumulate. So, I settle for that.

"Measure with a micrometer, mark with a grease pencil, cut with an axe" is the reality of retirement planning vs. the retirement experience.

In my engineering work, I did a lot of detailed engineering simulation. Some were quite numerical sensitive and required extended-precision floating point numbers (in contrast with single-precision and double-precision floating point).

However, I used the results of the simulation in a qualitative way and not quantitative. I will make up an example to explain this.

For example, suppose I have built a simulation of an ICE car and it shows the car MPG is 35. Now, I can tweak some parameters and see the effect on the MPG. Is it a tad higher or lower? If higher than 35, then I am going in the right direction. And I will go out and see if the improvement shows up in real-life testing. If the real car goes from 31 mpg to 31.2 mpg, then I pat myself on the back and go back to tweak something else.

If the physical sciences are hard, social sciences are way harder. You can never repeat the same experiment but tweaking something to observe the effects. Suppose Volcker increased the interest rate more or less than he did in 1980. What would happen? We never know for sure. Everything is just theoretical or a surmise. The book I recently re-read, "Paper Money", made fun of econometrics. It's a good read.

 

[ls99]

When I retired at 59.5, Firecalc predicted 74% success. Damn the torpedoes, full speed ahead. I was never fearful of failure, ever. Still not.

 

[skyking1]

I don't use it, but I think that class of gadgets, properly understood is useful.

Unfortunately I interpret evidence here to indicate that most people don't understand these gadgets or their limitations. Too much agonizing over the difference between 100% and 90% for example. IMO those two numbers are effectively the same. But obviously it makes money for the site operators, so I guess "All opinions guaranteed worth price paid." applies.

It was very useful for me, a relative noob at looking at all the pieces of our retirement.
This site taken in its entirety is a very useful tool, because the members are quite knowledgeable and active in their retirement management.
Various tools and methods are mentioned, debugged, advocated for.
I have tossed our numbers in several different "useful" models and as a result of what I saw, I am comfortable with the situation.
As a young lad I swung a splitting maul and like the axe analogy above, and that is basically what we are doing.

 

[NW-Bound]

When I retired at 59.5, Firecalc predicted 74% success. Damn the torpedoes, full speed ahead. I was never fearful of failure, ever. Still not.

Holly moly! You are more financially courageous than I am. Makes my active investing look like nothing.

 

[Montecfo]

Analysis is great. I have spent a career doing a lot of it. But it has limits: we do not know the future of the world, of our country or of ourselves.

But ultimately you have to become comfortable with the risks and, most importantly, have some levers you can adjust in case things go differently than expected.

And save more than the calculators say that you need.

 

[ls99]

Holly moly! You are more financially courageous than I am. Makes my active investing look like nothing.

Set and forget at Vanguard, and moving TIAA funds there worked wonders. I now far more than I ever could have imagined. My nieces will be filthy rich when I kick the bucket. They just don't know it

 

[11522914]

But ultimately you have to become comfortable with the risks and, most importantly, have some levers you can adjust in case things go differently than expected.

And save more than the calculators say that you need.

+100

One of my levers right now is having the option of starting to collect my SS. Everything is going very well in my 100% income portfolio. I'm getting ready to spend 13k on home maintenance and my net worth will still be positive YTD after that expense. Life is good. YMMV

 

[skyking1]

One of my levers is part time work.
I can make about $2000 a month net without kicking myself out of the pension, in my trade.
If I go outside my trade I can do whatever, bu this is the easy button and I am in demand.

 

[audreyh1]

I really appreciate the FIREcalc tool. I’ve always understood that it’s historical, no predictions. But the data goes all the way back to the late 1800s and covers a very wide range of economic periods including some very volatile ones.

I’ve used it to model a different type of withdrawal method than the typical one and it gave me lots of good information and really valuable insight.

Ultimately as an overall reality check I track my total net worth against inflation since retiring. Things are still growing (knock on wood) even though I’m allowed to spend down a bit. Flexibility is key as is early detection of shrinking compared to inflation.

 

[Sojourner]

I don't see how going back to the 1950s markets, let alone the early 1900s is helpful.
[...]
In short, our world and the markets that react to our world is entirely different than even a few decades ago. While I fully appreciate the FC outputs, I always wonder about the value of looking back 50 or 100 years to such a different time...how relevant can that analysis be? We might as well look to ancient Rome for comparison.

I've had these exact same thoughts in regards to FIRECalc, and that's why I always change the starting date for backtesting from its default of 1871 (when the U.S. had only 37 states!). Of course, the question then becomes which year to start from. I usually go with 1933, since this is when the U.S. implemented major reforms to the securities markets that are still in effect to this day. But who knows?

Given the limitations of backtesting tools like FIRECalc, my approach is to use the most conservative inputs I reasonably can (e.g. greatly reduced SS, no inheritance, 100% historical success) and to shave about 15-20% off the SWR that comes out. This, to me, seems like the best I can do considering that I am not going to "unretire", and I have to live in the real world with all its uncertainties and risks.

 

[marko]

It was very useful for me, a relative noob at looking at all the pieces of our retirement.
.

+1 It was for me as well when first starting out in (unplanned) ER. However flawed it may/may not be, it gave me a level of comfort knowing that I wasn't entirely out in the weeds, confirmed by other retirement tools.

 

[LateToFIRE]

...I don't see how going back to the 1950s markets, let alone the early 1900s is helpful...In short, our world and the markets that react to our world is entirely different than even a few decades ago. While I fully appreciate the FC outputs, I always wonder about the value of looking back 50 or 100 years to such a different time...how relevant can that analysis be? We might as well look to ancient Rome for comparison. IMO.

+1, Totally agree. It's intellectually interesting info, but certainly not conclusive.