Dipping under 100% on Firecalc?

[corn18]

The comparison of the two scenarios is a distinction without a difference. The data are the data. What you do with it is what matters.

 

[daylatedollarshort]

Are you saying that the data entered for both people is EXACTLY the same? I don't think it is. The year each started retirement is different, a key element of the input data.

It is a hypothetical example to illustrate a point, so it is identical because that is part of the example. Bob retired one year ago with $1M with an end date of 2051 and now has $850K. He had a firecalc score of 100% last year and now it is 85%. Mary worked until today and has $850K with a planned end date of 2051 and a firecalc score of 85%. Other than that, everything is identical in their portfolios and assumptions.

Going forward, do they have the same probability of retirement success? They have identical portfolios, identical assumptions and the same planned end dates. Mary's score is 85%. Is Bob's score now still 100% or is it 85%, the same as Mary's, since going forward they have identical time frames, asset allocations and portfolios. I won't post about Bob and Mary any more unless I think of some new twist to add. In this case I put in specific end dates and added portfolio numbers, as those are key.

 

[RunningBum]

I've never felt that Firecalc was the ultimate decider on my retirement finances. It's a tool, based on history from the first of each year for 100+ years. There's no guarantee things won't be worse, so if I did retire at the start of this year I would be concerned, but not overly so. It's never comforting to lose more money in the market than you are spending.

We are not on the track of one of the historical runs Firecalc uses. Every run is different. Probably it's not going to be worse than all previous runs but it could be. Given that Firecalc only runs from the start of each year (as I understand it, and no one corrected me when I said this before), Firecalc doesn't even capture the very worst runs in the last 100+ years. You can argue about what 100% success in Firecalc means, but it's an imperfect tool to use to assess your plan for your retirement future so I don't see the point.

 

[ERD50]

It is a hypothetical example to illustrate a point, so it is identical because that is part of the example. Bob retired one year ago with $1M with an end date of 2051 and now has $850K. He had a firecalc score of 100% last year and now it is 85%. Mary worked until today and has $850K with a planned end date of 2051 and a firecalc score of 85%. Other than that, everything is identical in their portfolios and assumptions.

Going forward, do they have the same probability of retirement success? They have identical portfolios, identical assumptions and the same planned end dates. Mary's score is 85%. Is Bob's score now still 100% or is it 85%, the same as Mary's, since going forward they have identical time frames, asset allocations and portfolios. I won't post about Bob and Mary any more unless I think of some new twist to add. In this case I put in specific end dates and added portfolio numbers, as those are key.

OK, I've got it. I've got a way to nail this down... Look at it this way:

Bob had $1M in year xxxx when he retired, and his spend rate gave him 100% success in FC for 30 years.

Two years later, the market took a hit, and Bob is down to (whatever, say) 650K and a FC run for 28 years shows a lower success percent.

Mary, at that time, also has $650K portfolio and the same spend as Bob. So why is her success % lower, if Bob was at (and presumably still is) 100% success?

The reason is... , we now have 2 years of "knowns". We "know" that Bob has had his portfolio decline, and we know that historically, none of the strarting years failed with Bob's starting numbers, his portfolio will recover.

However, when you run Mary's (or Bob's with the latest values), FC makes the run against every starting year.

Now.... imagine if we could tell FC to filter out every starting year that had a portfolio balance above $650K in year 2. Then you would see 100% success rate. Because the FC runs that were below $650K at that point went on to succeed. But some of the starting years that were above $650K in year 2, would not succeed with $650K, they needed more.

That's why you can't just plug in that lower number, and expect the same success rate.

What does this mean for someone running FC now, with that info.... stay tuned...

-ERD50

 

[Dtail]

No, that's not the case (speaking strictly with the historical data at hand, and the algorithms).

It's common for the calculated success rate to drop over the course of the retirement. The only way that would not be true is your portfolio outpaced your spending the very first year, and stayed ahead the entire time. But normal cycles have ups and downs.

So for any 100% successful run, the retiree did not need to cut spending or make any adjustment. The 100% was under those conditions and succeeded. Period.

Now, if you are talking a real life retiree, not a historical one, sure - the future is uncertain and unknown. But DLDS was questioning the tool results, and that can only be evaluated on the data that is on hand, and our understanding of just what it means.

-ERD50

Actually I was referring to the retiree in a real life situation. I didn't make that clear, not the data results in that situation.
If one runs Firecalc on an ongoing basis and the score drops to 70% at any given time for example, one would not likely be satisfied if the historical sequencing data shows that he should make it in the end.

 

[ERD50]

Actually I was referring to the retiree in a real life situation. I didn't make that clear, not the data results in that situation.
If one runs Firecalc on an ongoing basis and the score drops to 70% at any given time for example, one would not likely be satisfied if the historical sequencing data shows that he should make it in the end.

OK, I'm getting a little confused with some posts I think talking about the mechanics of what FC shows, versus the human reaction to the results.

But if one is surprised that they are seeing a drop to 70% success after a little hiccup in the market, that means they didn't understand it in the first place. I started a thread long ago about the "scary dips" - follow the lines, and even with 100% success, there are drops to ~50% in just a few years. But they recover (if you set spending for 100%).

So sure, those later dropped values as starting numbers will lead to failure. But as I pointed out, the ones that fail from that point didn't have that big drop and a subsequent run up to recovery. Without that run up, they fail from that lower point.

Sure, people may still not be satisfied, we'd all like to see rainbows and unicorns, but the numbers are the numbers, it's all you can do.


-ERD50

 

[daylatedollarshort]

OK, I've got it. I've got a way to nail this down... Look at it this way:

Bob had $1M in year xxxx when he retired, and his spend rate gave him 100% success in FC for 30 years.

Two years later, the market took a hit, and Bob is down to (whatever, say) 650K and a FC run for 28 years shows a lower success percent.

Mary, at that time, also has $650K portfolio and the same spend as Bob. So why is her success % lower, if Bob was at (and presumably still is) 100% success?

The reason is... , we now have 2 years of "knowns". We "know" that Bob has had his portfolio decline, and we know that historically, none of the strarting years failed with Bob's starting numbers, his portfolio will recover.

However, when you run Mary's (or Bob's with the latest values), FC makes the run against every starting year.

Now.... imagine if we could tell FC to filter out every starting year that had a portfolio balance above $650K in year 2. Then you would see 100% success rate. Because the FC runs that were below $650K at that point went on to succeed. But some of the starting years that were above $650K in year 2, would not succeed with $650K, they needed more.

That's why you can't just plug in that lower number, and expect the same success rate.

What does this mean for someone running FC now, with that info.... stay tuned...

-ERD50

In real life, regardless of firecalc scores, two people with identical starting portfolios on 5/29/2022, parameters and the same retirement end dates have the same probability of portfolio failure or success going forward. It is just math. Different firecalc scores aren't going to change the actual probability outcome.

 

[corn18]

OK, I'm getting a little confused with some posts I think talking about the mechanics of what FC shows, versus the human reaction to the results.

But if one is surprised that they are seeing a drop to 70% success after a little hiccup in the market, that means they didn't understand it in the first place. I started a thread long ago about the "scary dips" - follow the lines, and even with 100% success, there are drops to ~50% in just a few years. But they recover (if you set spending for 100%).

So sure, those later dropped values as starting numbers will lead to failure. But as I pointed out, the ones that fail from that point didn't have that big drop and a subsequent run up to recovery. Without that run up, they fail from that lower point.

Sure, people may still not be satisfied, we'd all like to see rainbows and unicorns, but the numbers are the numbers, it's all you can do.


-ERD50

Oh, I understood it before I retired and understand it very well now. I understand it enough that I built it into my spreadsheet. I know exactly what it does and how it does it.

You correctly cite that it is a human reaction to the results.

The question is what should you do about it? Absolutely nothing.

 

[mikes425]

Still don't know get how you can accurately reflect your portfolio in FireCalc if there's no provision in the Portfolio options to enter international equities and bonds, and FI funds that don't fall under any of the provided categories - And as far as bonds.. My PF has a range, of short, intermediate and long duration holdings. It doesn't provide any option to really define that.

 

[youbet]

In real life, regardless of firecalc scores, two people with identical starting portfolios on 5/29/2022, parameters and the same retirement end dates have the same probability of portfolio failure or success going forward. It is just math. Different firecalc scores aren't going to change the actual probability outcome.

Aaaah..... finally something we can agree on. Yes, two identical scenarios passing through the same times will end with the same results. We have no idea of what the results will be, but they will be identical. FireCalc testing against historical data is independent of results going forward. Any link between the past and the future is up to the individual user to to accept or deny. Personally, I think testing against historical data is the best we can do. If we can identify some assignable cause for the future to be different from the past (the past as used by FireCalc) then we'll have to find a way to work that into the scheme.

 

[youbet]

Still don't know get how you can accurately reflect your portfolio in FireCalc if there's no provision in the Portfolio options to enter international equities and bonds, and FI funds that don't fall under any of the provided categories - And as far as bonds.. My PF has a range, of short, intermediate and long duration holdings. It doesn't provide any option to really define that.

Of course not. FireCalc tests against historical data. There is limited data for the categories you mention. You can't test against historical data that doesn't exist.

You enter your portfolio data as close as you can approximate it against FireCalc's choices and go from there.

This is just one of the reasons the range of outputs for any given set of inputs is so broad. You may end up broke or fabulously wealthy. You're just testing to see if, historically, you might avoid running out of money.

Don't try to measure with a micrometer and cut with an ax.

 

[daylatedollarshort]

Still don't know get how you can accurately reflect your portfolio in FireCalc if there's no provision in the Portfolio options to enter international equities and bonds, and FI funds that don't fall under any of the provided categories - And as far as bonds.. My PF has a range, of short, intermediate and long duration holdings. It doesn't provide any option to really define that.

We have a lot of TIPS and none of the planners I used allowed for that. I made a retirement spreadsheet that had real interest rates and inflation as variables. It worked out well because we were able to model and stress test both high inflation and low real returns, like we ended up having today. We also had international stocks and bonds up until recently but not enough for me to bother adding them as a separate variable. But with an Excel type spreadsheet that would be easy enough to add.

The results usually came out pretty close to the Fidelity retirement planner under conservative assumptions, which I used as my reasonableness check.

 

[youbet]

In real life, regardless of firecalc scores, two people with identical starting portfolios on 5/29/2022, parameters and the same retirement end dates have the same probability of portfolio failure or success going forward. It is just math. Different firecalc scores aren't going to change the actual probability outcome.

I'd just drop the word "probability" from your description. FireCalc results will not predict or determine the actual future outcome. Probabilities are something we have to assign. The two people will have the same results, but there really isn't anything we can say about the probability of any particular result apriori.

 

[corn18]

Of course not. FireCalc tests against historical data. There is limited data for the categories you mention. You can't test against historical data that doesn't exist.

You enter your portfolio data as close as you can approximate it against FireCalc's choices and go from there.

This is just one of the reasons the range of outputs for any given set of inputs is so broad. You may end up broke or fabulously wealthy. You're just testing to see if, historically, you might avoid running out of money.

Don't try to measure with a micrometer and cut with an ax.

If you want to roll your own model, you can use the data from Simba back testing spreadsheet. That's what I use for my historical model. I can tailor what data I use to more closely match my portfolio.

https://www.bogleheads.org/forum/viewtopic.php?p=6506086#p6506086

 

[Koolau]

I'd love to be 17 years younger, but the good news is I no longer am running FIRECalc for the first time. In fact, I don't run FC any longer. So, when encountering today's equity slide (and even the high inflation) I don't get too excited. I'll admit to a bit of angst for the world my kids and GKs are inheriting. Heh, heh, they don't know about FIRECalc yet, so I guess they're okay. Seriously, I do plan on alerting them to FC. YMMV

 

[youbet]

Our interesting discussion has brought me to the point of being able to comment on your early questions.

It is a hypothetical example to illustrate a point, so it is identical because that is part of the example. Bob retired one year ago with $1M with an end date of 2051 and now has $850K. He had a firecalc score of 100% last year and now it is 85%. Mary worked until today and has $850K with a planned end date of 2051 and a firecalc score of 85%. Other than that, everything is identical in their portfolios and assumptions.

Going forward, do they have the same probability of retirement success?

Yes, but we don't know what that probability is and the outcome will be independent of FireCalc outputs.

They have identical portfolios, identical assumptions and the same planned end dates. Mary's score is 85%. Is Bob's score now still 100% or is it 85%,

When Bob's original data was entered into FireCalc, it back-tested to be 100% successful. When the new data with the lower portfolio value was entered, it back-tested as being 85% successful. The two outcomes are independent events. Some of the difference can be explained by the fact that Bob's 2nd run consisted of a greater number of 29 year runs beginning with a different year. But I'd guess most of the difference is from reducing Bob's portfolio by his initial years tough losses and then testing that smaller portfolio against the same tough sequences that were included in his first run.

 

[daylatedollarshort]

Our interesting discussion has brought me to the point of being able to comment on your early questions.

Yes, but we don't know what that probability is and the outcome will be independent of FireCalc outputs. When Bob's original data was entered into FireCalc, it back-tested to be 100% successful. When the new data with the lower portfolio value was entered, it back-tested as being 85% successful. The two outcomes are independent events. Some of the difference can be explained by the fact that Bob's 2nd run consisted of a greater number of 29 year runs beginning with a different year.

You can run it in the same year, lower the portfolio value by an average bear market drop and go from a 100% to 70% score.

There was a thread on Boglehead's about if your firecalc scare was 100% , why haven't you retired? I don't recall any of the replies being along the lines of because that score could easily be 70% within a year's time, and they wouldn't feel comfortable at retiring with a 70% score.

Per ERD50s post "Bob's odds did *not* go to 86% - he started at 100% historical success, and he stays at 100%." Which circles us back Bob and Mary.

 

[youbet]

If you want to roll your own model, you can use the data from Simba back testing spreadsheet. That's what I use for my historical model. I can tailor what data I use to more closely match my portfolio.

https://www.bogleheads.org/forum/viewtopic.php?p=6506086#p6506086

Thanks, but like Koolau, I'm too darn old to worry about the details of my FIRE portfolio lasting. It's been 16+ years since I had any earned income and we've done OK. So these days I just focus on portfolio performance, keeping expenses reasonable but still doing most of what we want to do and trying to maintain health (getting challenging).

I do play around with FireCalc and other tools from time to time but more for guessing what my terminal portfolio value might be for estate planning purposes.

But, maybe I'll go over there and take a look just the same. Thanks.

 

[ERD50]

Originally Posted by daylatedollarshort View Post
In real life, regardless of firecalc scores, two people with identical starting portfolios on 5/29/2022, parameters and the same retirement end dates have the same probability of portfolio failure or success going forward. It is just math. Different firecalc scores aren't going to change the actual probability outcome.

I'd just drop the word "probability" from your description. FireCalc results will not predict or determine the actual future outcome. But probabilities are something we have to assign. The two people will have the same results, but there really isn't anything we can say about the probability of any particular result apriori.

Yes, as youbet says, if we are talking about two people, same profile, same number of years, starting now and going forward, the results *will* be the same, not 'probably'.

So we agree on that. Now, what was the question?

There is no paradox here. You just described two people under the same scenario, with the same future ups/downs. But FC runs you against a bunch of scenarios. But you compare Bob and Mary under two different scenarios - his first run that included the early big dip (but he's at 100% for that first run), and then a second scenario where he (and Mary) gets hit by that big early dip, but with a smaller starting portfolio. So of course they *both* have the same lower success rate.

But you can't compare that to Bob's first run with his full portfolio. That's a different time/place/situation. He already hit that early dip - you are sticking him with it twice. That's not in the data set for any starting year.

If you were to pick a single starting year for Bob on the first run, and then select a starting year two years from that 1st run, with Bob's lower balance at year 2, you would get 100% for both for Bob and Mary. That's analogous to your "same future" scenario.

Per ERD50s post "Bob's odds did *not* go to 86% - he started at 100% historical success, and he stays at 100%." Which circles us back Bob and Mary.

The "first run Bob" is still at 100%. History didn't change. But *you* changed history when you decided that you should be able to pull Bob's first run numbers after a downturn, and start over, hitting that dip a second time (which didn't happen in history) with the already damaged portfolio, and he should still be at 100%... because.

I take it you have not absorbed my runner analogy. That's really all there is to it, nothing more, nothing less.

-ERD50

 

[youbet]

You can run it in the same year, lower the portfolio value by an average bear market drop and go from a 100% to 70% score.

There was a thread on Boglehead's about if your firecalc scare was 100% , why haven't you retired? I don't recall any of the replies being along the lines of because that score could easily be 70% within a year's time, and they wouldn't feel comfortable at retiring with a 70% score.

Per ERD50s post "Bob's odds did *not* go to 86% - he started at 100% historical success, and he stays at 100%." Which circles us back Bob and Mary.

I think you're overstating the likelihood of a success rate dropping from 100% to 70% other than rarely.

BTW, Bob's "odds" do not change. "Odds" implies probabilities of future outcomes, something we're not talking about. His historical portfolio success rate changes.

What ERD50 meant was that Bob's first run back-tested at 100% successful. His run a year later back-tested at 85% successful. Neither run has an effect on the other as they are independent. Even after the 2nd run back-tested at 85%, the results of the first run were still 100% and nothing has happened yet to Bob that falls outside of the first run results.

I can't remember what your personal situation is. Are you contemplating FIRE with an exact 100% FireCalc success rate and concerned if you test again next year it will be 70% Or are you retired and thinking about returning to the work force due to our market downturn and increase in prices?

The question you started off with is a good one in these volatile times. If our portfolio values are going to be swinging up and down dramatically over relatively short periods of time taking us from good financial shape to not-so-good financial shape as tested historically, what can we do about it? How do we know if living off passive income is going to be viable? I guess we don't. Just ask those economist folks who try to forecast the future.........

 

[daylatedollarshort]

I think you're overstating the likelihood of a success rate dropping from 100% to 70%.

We've only had a 20% or so market drop so far this year and corn18's score has dipped to 86. That isn't a big drop by historical standards. Anyway I've made my points along with some others here and you've made yours. The other posters can decide for themselves what to do with the information presented. I don't use firecalc for the various reasons stated in this thread by myself and others.

 

[mikes425]

Of course not. FireCalc tests against historical data. There is limited data for the categories you mention. You can't test against historical data that doesn't exist.

You enter your portfolio data as close as you can approximate it against FireCalc's choices and go from there.
...
Don't try to measure with a micrometer and cut with an ax.

No historical data for 2 vs 5 vs 10 year bonds? No historical data for non-US global market performance exists? That's hard to believe.

I get that FC doesn't include it, but obviously a calculator that DOES -in addition to the existing US stock/bond categories would be much more realistic and useful -allowing for a more accurate estimation of results reflecting ones actual PF composition. The analogy of the micrometer and ax doesn't seem applicable to my point.

Just saying, if one has exposure in categories that aren't represented in the choices FC offers, it's hard to just approximate outcomes using the choices offered if there are unrepresented components of the PF. They certainly figure importantly in the equation. I also can't believe others haven't thought of this variable not being addressed when entering their PF data in FC. It just seems one is getting an incomplete picture that is 'skewed' in not being able to reflect everything one has in one's PF.

Thanks!

 

[youbet]

No historical data for 2 vs 5 vs 10 year bonds? No historical data for non-US global market performance exists? That's hard to believe.

I get that FC doesn't include it, but obviously a calculator that DOES -in addition to the existing US stock/bond categories would be much more realistic and useful -allowing for a more accurate estimation of results reflecting ones actual PF composition. The analogy of the micrometer and ax doesn't seem applicable to my point.

Just saying, if one has exposure in categories that aren't represented in the choices FC offers, it's hard to just approximate outcomes using the choices offered if there are unrepresented components of the PF. They certainly figure importantly in the equation. I also can't believe others haven't thought of this variable not being addressed when entering their PF data in FC. It just seems one is getting an incomplete picture that is 'skewed' in not being able to reflect everything one has in one's PF.

Thanks!

I didn't say "NO" historical data.

Of course having more options to slice and dice holdings would be preferable. We've discussed it a lot here and limited data is the reason I remember. If you know the necessary data is available, why don't you bring it up to the administrator? Perhaps you'll get a better answer.

 

[mikes425]

I didn't say "NO" historical data.

Of course having more options to slice and dice holdings would be preferable. We've discussed it a lot here and limited data is the reason I remember. If you know the necessary data is available, why don't you bring it up to the administrator? Perhaps you'll get a better answer.

Ok my error- limited data. Sorry. But i suspect that some historical benchmark tracking data in non-US markets would be well established resources by now and yes, agreed - it would be a good improvement to suggest to the FC admin. for reasons mentioned. Curious others havent raised this as most well diversified portfolio models have some variation in the duration of bonds - and international exposure.��

 

[youbet]

Ok my error- limited data. Sorry. But i suspect that some historical benchmark tracking data in non-US markets would be well established resources by now and yes, agreed - it would be a good improvement to suggest to the FC admin. for reasons mentioned. Curious others havent raised this as most well diversified portfolio models have some variation in the duration of bonds - and international exposure.��

As I mentioned in my last post, others have raised this issue. It was from those discussions that I recalled "limited data" being mentioned and I passed that on to you. Perhaps someone else will chime in who can give you specifics.