Firecalc knowledge of the past?

[mrWinter]

Hopefully I'm able to articulate my thought here. I've still many years before I can retire, probably a couple decades. As such I run numbers in firecalc and other simulators with 20 years of savings then 40 or so years of retirement.

I've also out of curiosity pretended I'm retiring this year and see what balance I need to retire, what's my 'magic number'. What's striking me as odd is that in the scenario where I save for 20 years first, I'll get many simulations that have a balance at my retirement year that is lower than my 'magic number' yet still end up succeeding in the long term.

For example take this input, but go to the investigate tab and check the box to export the data. See the data for 1930, you'll see the inflation adjusted balance right before retirement is $1,875,288. Yet if I use the investigate tab to see what the minimum balance is required to guarantee 100k withdrawals for 40 years, I get $2,992,703.

So say I start with that same plan saving for 20 years then retiring for 40. The simulation says in 100% of scenarios I'm successful. Yet if I followed that plan and followed along with one of the scenarios, starting this all in 1930, then ran the calculator again at my retirement year 1950, having a balance of $1,875,288, the same calculator would say I had only a 45.8% chance of success.

One of these two forecasts has to be wrong. I believe it's the forecast made in 1950, because when running the calculator then, I'm actually giving it less information. The only scenarios that would yield a balance that low after 20 years of saving 20k a year on top of a 750k starting balance is one where the market had performed very poorly leading up to retirement, which then eventually rebounded, carrying the portfolio along with it. I should be able to retire in confidence despite a 45.8% prediction because I followed the savings plan that produced a 100% success over the longer term.

I feel like people may be getting overly pessimistic forecasts. If I was on track to retire in 2008 based on all simulations (which include ones where there is a big crash right at the beginning of retirement), but as we know everything went to crap in 2007, I was probably still okay to make the jump in 2008, despite the fact that my balance was now likely far below the number I needed to achieve 100% success in simulations going forward.

So how do we avoid this? How can we, at the time of retirement, calculate the magic number, but include data on all our past savings history? Or is it past market performance we need to consider? And how do we do this using data based simulations to get a better picture, rather than just off-the-cuff gut feelings of market conditions?

Or is my thinking here wrong and the simulation run at retirement age is actually more accurate? A retirees Monty Hall problem perhaps? I need some statisticians or experts in probabilities to chime in.

[pb4uski]

It is unnecessary to consider past savings history or market performance.... when you do a run as of your retirement date all of that is reflected in your portfolio balance.

[ERD50]

If I'm following you (sorry, I didn't study your post in great detail), I think what is happening is you are essentially running two periods, one after the other? If you do that, you get a really bad 20 years as the worst run, followed by that same really bad 20 years in the second run. But the worst 40 years is not as bad as the worst 20 year period played back-to-back.

Does that fit what you described?

Also, 60 year runs take out some data - only complete runs from ~ 1957 on can be used, or they won;t have a full 60 years.

-ERD50

[mrWinter]

Quote:

Originally Posted by ERD50 If I'm following you (sorry, I didn't study your post in great detail), I think what is happening is you are essentially running two periods, one after the other? If you do that, you get a really bad 20 years as the worst run, followed by that same really bad 20 years in the second run. But the worst 40 years is not as bad as the worst 20 year period played back-to-back. Does that fit what you described? Also, 60 year runs take out some data - only complete runs from ~ 1957 on can be used, or they won;t have a full 60 years. -ERD50

No that's not what I'm doing. I'm comparing a 60 year simulation which includes 20 years pre-retirement that yields 100% success, with a 40 year simulation of just post-retirement years that yields 100% success. The 40 year simulation requires 3 mil at retirement to give that 100%, while the 60 year scenario has scenarios that have a smaller balance, <2mil at retirement yet still end up succeeding. Meaning I could make a plan for the next 60 years that has 100% success, then follow the first 20 years of that plan, then rerun the simulation with up-to-date data at my target retirement year, to find that I now have <100% success, or even <50% success, even though I've followed what was a 100% success plan. That doesn't make sense. The forecast should not get more grave as time goes on unless reality is dishing out performance worse than the worst case simulation.

[ERD50]

Maybe my second part explains it?

A 60 year run only has data from start dates of ~ 1957 and before.

A 40 year run includes start dates of ~ 1977 and before.

~ 1966 was one of the 'killer' start dates, included in the 40 year run, but not the 60 year run.

-ERD50

[cathy63]

Quote:

Originally Posted by mrWinter For example take this input, but go to the investigate tab and check the box to export the data. See the data for 1930, you'll see the inflation adjusted balance right before retirement is $1,875,288. Yet if I use the investigate tab to see what the minimum balance is required to guarantee 100k withdrawals for 40 years, I get $2,992,703.

You are comparing two different things here. While $1.875288M was a fine number to retire with in 1950, and the single-year spreadsheet for 1930 shows success, there is no reason to think that nest egg would be sufficient for every other 40 year period. In particular, if you had retired in 1966, you would have needed $2.992703M. Fortunately, if you had started your savings program in 1946, you would have amassed $4.385914M by retirement so you would have succeeded in that year also.

Quote:

Originally Posted by mrWinter So say I start with that same plan saving for 20 years then retiring for 40. The simulation says in 100% of scenarios I'm successful. Yet if I followed that plan and followed along with one of the scenarios, starting this all in 1910, then ran the calculator again at my retirement year 1930, having a balance of $1,875,288, the same calculator would say I had only a 45.8% chance of success.

But as in the previous example, $1875288 is just the amount you would have had in 1950 after saving for 20 years starting in 1930. You can't assume that would have been enough to retire in 1930 or 1966 or any other year, just because it was enough in 1950.

[cathy63]

Quote:

Originally Posted by mrWinter No that's not what I'm doing. I'm comparing a 60 year simulation which includes 20 years pre-retirement that yields 100% success, with a 40 year simulation of just post-retirement years that yields 100% success. The 40 year simulation requires 3 mil at retirement to give that 100%, while the 60 year scenario has scenarios that have a smaller balance, <2mil at retirement yet still end up succeeding. Meaning I could make a plan for the next 60 years that has 100% success, then follow the first 20 years of that plan, then rerun the simulation with up-to-date data at my target retirement year, to find that I now have <100% success, or even <50% success, even though I've followed what was a 100% success plan. That doesn't make sense. The forecast should not get more grave as time goes on unless reality is dishing out performance worse than the worst case simulation.

Yep, this absolutely could happen. The future could be extremely dismal and things could happen that have never happened before. Firecalc can only tell you whether your strategy would have survived everything that's happened in the past periods for which it has valid data. It is not a forecasting tool; it's a backtesting tool.

1966 was a very bad year to retire, but it was preceded by 20 extremely good years for saving, so that is all included in your 60 year runs. You have to determine for yourself whether you think the next bad retirement years are more likely to be preceded by good savings years or bad ones. We definitely could have an unprecedented 60-year interval where your savings time is more like 1930 to 1950 and then your retirement time is more like 1966 to 2006. Firecalc doesn't have a way to say "ignore everything that happened between 1950 and 1966", so things like Monte Carlo simulators and other randomization algorithms might work better for you right now.

Also, 60 years is a very long horizon. You're essentially trying to predict events that'll cover 60-75% to of your entire lifespan. There's no tool that can do that accurately.

[mrWinter]

Quote:

Originally Posted by cathy63 You are comparing two different things here. While $1.875288M was a fine number to retire with in 1950, and the single-year spreadsheet for 1930 shows success, there is no reason to think that nest egg would be sufficient for every other 40 year period. In particular, if you had retired in 1966, you would have needed $2.992703M. Fortunately, if you had started your savings program in 1946, you would have amassed $4.385914M by retirement so you would have succeeded in that year also. But as in the previous example, $1875288 is just the amount you would have had in 1950 after saving for 20 years starting in 1930. You can't assume that would have been enough to retire in 1930 or 1966 or any other year, just because it was enough in 1950.

Right, exactly, I agree with all you've said. But had I run the calculator right before retiring, with $1875288, the calculator would have said I'd have <50% chance of success, but I didn't, I had a better chance than that.

[cathy63]

Quote:

Originally Posted by mrWinter Right, exactly, I agree with all you've said. But had I run the calculator in 1950 with $1875288, the calculator would have said I'd have <50% chance of success, but I didn't, I had a better chance than that.

Yes, if you'd run Firecalc in 1950 with a starting value of $1875288 and withdrawals of $100K*, you'd have been looking at 40-year intervals starting from 1870 to 1910 and 23 out of 40 would have failed. Firecalc would not have been able to tell you that the next 40 years were going to be fine and you actually had a 100% success rate. It cannot predict the future. Just as the future might be more dire than anything that's happened in the past, it might also be much better than anything that's happened in the past.

* Aside: we're mixing inflation adjusted and non-adjusted numbers here, so this analysis is not really correct. I haven't thought about it deeply enough to figure out whether we ought to take your 1950 portfolio and adjust it to a 2018 size before running this scenario or maybe we need to reduce the $100K spend to a corresponding 1950 level.

[pb4uski]

Quote:

Originally Posted by mrWinter Hopefully I'm able to articulate my thought here. I've still many years before I can retire, probably a couple decades. As such I run numbers in firecalc and other simulators with 20 years of savings then 40 or so years of retirement. I've also out of curiosity pretended I'm retiring this year and see what balance I need to retire, what's my 'magic number'. What's striking me as odd is that in the scenario where I save for 20 years first, I'll get many simulations that have a balance at my retirement year that is lower than my 'magic number' yet still end up succeeding in the long term. For example take this input, but go to the investigate tab and check the box to export the data. See the data for 1930, you'll see the inflation adjusted balance right before retirement is $1,875,288. ....

For that 1930 scenario the portfolio balance just prior to retirement is $2,292,280.... cell J22.

[mrWinter]

Quote:

Originally Posted by pb4uski For that 1930 scenario the portfolio balance just prior to retirement is $2,292,280.... cell J22.

My interpretation of the spreadsheet is that cell J22 is the non-inflation-adjusted starting portfolio of 1949. There is no column for inflation-adjusted starting portfolio, so I used the inflation-adjusted ending portfolio of the previous year (1949), cell U22, to get the inflation adjusted starting portfolio of 1950, the year of retirement.

[mrWinter]

Seems like my thinking here isn't gaining much traction with others. Maybe I'm not making sense or maybe my thinking is just wrong.

Let me try to once more, with fewer variables, if that doesn't demonstrate what I'm thinking then I'll just layoff and assume I'm just not firing on all cylinders, but I think I can explain this more clearly:


Let's say it's September of 2007 and I'm planning to retire in 2009. My portfolio looks comfortably better than the minimum needed, forecasts show 100% success in all simulations. Cool. But then the recession happens.

Fast forward to January 2009, my proposed retirement date. My portfolio has dropped 40%. But can I still retire? Well I ran simulations back in 2007 that included the worst market drops in history (1929, 2001) happening right before or right after I retire. Looks like one of those scenarios is what's playing out for me. But the calculator showed success despite those drops. In theory, this was accounted for, I should be good to go still!

To confirm I now run the calculators again, I plug in my portfolio which is 40% of what it was when I last ran it in 2007, and what do I get? A less than 100% simulation success rate, probably starkly so. Why is that? Because these simulations include scenarios starting in 1929 or 2001, where I immediately get hit by another huge market drop. Nothing wrong about that given the information I have provided the calculator. But there has never been a period in all of history where we had a 2007 style recession, immediately followed by a 1929 drop or a 2001 drop again.

Could we in theory have something worse than the worst that ever happened yet in history? Sure, but that's not what firecalc tries to consider. If you want to consider things worse than the worst case yet encountered then go run a montey-carlo simulation. Unless you don't believe in any tenancy of reversion to the mean, then the fact that I am running a simulation in 2009 right after a huge recession should change the forecasts moving forward. Simulating two of the worst market drops in history, one right after the other, is being overly pessimistic, that would be market performance significantly worse than has ever been seen in the past.

[pb4uski]

Quote:

Originally Posted by mrWinter My interpretation of the spreadsheet is that cell J22 is the non-inflation-adjusted starting portfolio of 1949. ....

Why would you inflation adjust the $2,292,280?

I believe what it is saying is that if you retired at 12/31/49 with a $2,292,280 portfolio and your first year spending was $148,538 (the spending power of $100,000 at 12/31/1929 on 1/1/1950) that 12/31/1989 you would have $8,537,869.

[pb4uski]

Quote:

Originally Posted by mrWinter Seems like my thinking here isn't gaining much traction with others. Maybe I'm not making sense or maybe my thinking is just wrong. Let me try to once more, with fewer variables, if that doesn't demonstrate what I'm thinking then I'll just layoff and assume I'm just not firing on all cylinders, but I think I can explain this more clearly: Let's say it's September of 2007 and I'm planning to retire in 2009. My portfolio looks comfortably better than the minimum needed, forecasts show 100% success in all simulations. Cool. But then the recession happens. Fast forward to January 2009, my proposed retirement date. My portfolio has dropped 40%. But can I still retire? Well I ran simulations back in 2007 that included the worst market drops in history (1929, 2001) happening right before or right after I retire. Looks like one of those scenarios is what's playing out for me. But the calculator showed success despite those drops. In theory, this was accounted for, I should be good to go still! To confirm I now run the calculators again, I plug in my portfolio which is 40% of what it was when I last ran it in 2007, and what do I get? A less than 100% simulation success rate, probably starkly so. Why is that? Because these simulations include scenarios starting in 1929 or 2001, where I immediately get hit by another huge market drop. Nothing wrong about that given the information I have provided the calculator. But there has never been a period in all of history where we had a 2007 style recession, immediately followed by a 1929 drop or a 2001 drop again. Could we in theory have something worse than the worst that ever happened yet in history? Sure, but that's not what firecalc tries to consider. If you want to consider things worse than the worst case yet encountered then go run a montey-carlo simulation. Unless you don't believe in any tenancy of reversion to the mean, then the fact that I am running a simulation in 2009 right after a huge recession should change the forecasts moving forward. Simulating two of the worst market drops in history, one right after the other, is being overly pessimistic, that would be market performance significantly worse than has ever been seen in the past.

Your rationale makes sense. What some people do is where they are looking at retirement readiness at the end of a strong bull market is that they haircut their beginning portfolio a tad to try to normalize it... IOW, they make a judgemental reversion to the mean adjustment in the input.

I think one could make the same argument if one is starting at the end of a strong bear market... especially since the withdrawal rates are implicitly based on bad case scenarios... which is why some many of the trials result in substantial investment balances. The other way judgementally adjust it would be to accept a lower success rate.

[ERD50]

I agree with pb4uski, your thinking is sound, what you are describing is a form of that 'paradox' of the two retirees retiring a year apart. Earlier I kinda got lost in your post, and I don't look at those spreadsheet outputs much, so I didn't quite follow you.

I'll give a slight twist/addition from what pb4uski said, and highlight that FIRECalc is using the most pessimistic runs to determine failures. So going strictly by the numbers, and assuming the worst of the future is not worse than the worst of the past (don't be distracted by averages) you should not have to make any downward adjustments if you had portfolio that passed in 2007. Yes, the lower amount you have in 2009 may fail if run now, but that is like simulating two really bad runs, one right after the other.

Another way to think of that is - look at some of the successful squiggly lines - some of them make a big dip in the first few years, some are ~ 50% down. If you were to calculate their WR at that time, if they started with ~ 4%, they are now ~ 8%, and they succeeded. So when you enter your lower portfolio, you kind of look like those people a few years in, who are taking 8%, and doing OK in the long run. So if you are really in that position, you can take 8% now. But FIRECalc doesn't 'know' where you are in the cycle (and either do we, until after the fact!), so it has to be neutral, and say ~ 4% is what you can do for all cases.

So you should not need to make any downward adjustment in WR% for where we are now, that's baked into the cake. You may want to make a downward adjustment to account for a future that may be worse than the worst, or just for cushion, etc.

I think you get it.

-ERD50

[Dtail]

Quote:

Originally Posted by ERD50 I agree with pb4uski, your thinking is sound, what you are describing is a form of that 'paradox' of the two retirees retiring a year apart. Earlier I kinda got lost in your post, and I don't look at those spreadsheet outputs much, so I didn't quite follow you. I'll give a slight twist/addition from what pb4uski said, and highlight that FIRECalc is using the most pessimistic runs to determine failures. So going strictly by the numbers, and assuming the worst of the future is not worse than the worst of the past (don't be distracted by averages) you should not have to make any downward adjustments if you had portfolio that passed in 2007. Yes, the lower amount you have in 2009 may fail if run now, but that is like simulating two really bad runs, one right after the other. Another way to think of that is - look at some of the successful squiggly lines - some of them make a big dip in the first few years, some are ~ 50% down. If you were to calculate their WR at that time, if they started with ~ 4%, they are now ~ 8%, and they succeeded. So when you enter your lower portfolio, you kind of look like those people a few years in, who are taking 8%, and doing OK in the long run. So if you are really in that position, you can take 8% now. But FIRECalc doesn't 'know' where you are in the cycle (and either do we, until after the fact!), so it has to be neutral, and say ~ 4% is what you can do for all cases. So you should not need to make any downward adjustment in WR% for where we are now, that's baked into the cake. You may want to make a downward adjustment to account for a future that may be worse than the worst, or just for cushion, etc. I think you get it. -ERD50

This concept counters the OP concerns which has been stated in varying degrees in other forums/articles.
Otherwise, you are supposedly "in trouble" if you have a bear market just before retirement or just into retirement, or will have issues retiring deep into a bull market.
So when can you retire safely?

[REWahoo]

Quote:

Originally Posted by Dtail So when can you retire safely?

Never, so keep working and doing your part to help solve the looming SS crisis. Those of us who have "retired unsafely" will be greatly appreciative.

[Independent]

Quote:

Originally Posted by mrWinter Could we in theory have something worse than the worst that ever happened yet in history? Sure, but that's not what firecalc tries to consider. If you want to consider things worse than the worst case yet encountered then go run a montey-carlo simulation. Unless you don't believe in any tenancy of reversion to the mean, then the fact that I am running a simulation in 2009 right after a huge recession should change the forecasts moving forward. Simulating two of the worst market drops in history, one right after the other, is being overly pessimistic, that would be market performance significantly worse than has ever been seen in the past.

Right. FireCalc does not "believe in reversion to the mean".

Suppose I'm ready to retire. FireCalc has input fields for my current assets, income, spending targets, etc.

It does not have an input for "recent market performance" or even "current market P/E ratio".

FireCalc is missing something in its calculations - recent market performance. It does not "believe in" reversion to the mean (at least not wrt averaging performance prior to the run start date with performance after).

[USGrant1962]

Kitces published a study of the SWR paradox and suggests that market valuation may be considered in picking a SWR. So in the case where there is a bear just before retirement, presumably valuations are lower and you can select a slightly higher SWR. This partially closes the gap in the paradox.

From https://www.kitces.com/wp-content/up...t-May-2008.pdf

Quote:

Unfortunately, market valuation doesn’t completely eliminate the impact of the starting level of the market on a safe withdrawal rate. It’s a refinement, but does not yield guidelines that precisely equalize the spending amount in each year due to portfolio fluctuations and changes in market valuation (if it did, the Notquiteyets would have had a safe spending amount when they retired of the exact same $46,350 that the Retirenows had). Nonetheless, accounting for market valuation does significantly mitigate the impact of fluctuations in the starting account balance on a prospective retiree’s safe spending amounts. As mentioned earlier, the Notquiteyets were able to spend 11% more than they could have under the original framework, and this in turn reduced the gap between the recommended spending of the two couples from 21% to less than 10%.

[Dtail]

Quote:

Originally Posted by REWahoo Never, so keep working and doing your part to help solve the looming SS crisis. Those of us who have "retired unsafely" will be greatly appreciative.

Hey there - my comment was rhetorical. I retired last year and couldn't be happier.