Author Topic: cFiresim SEVERELY overestimates success rates for Mustachians  (Read 15893 times)

tyort1

  • Handlebar Stache
  • *****
  • Posts: 1436
  • Age: 45
  • Location: Denver, Colorado
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #50 on: September 21, 2016, 09:22:44 AM »
On the other hand, if you have a 25k budget, then getting a job that pays even as little as $15k per year will have a huge protection against FIRE failure.

Many people talk about how if they leave their high paying jobs that it'll be impossible to get back to a similar level of salary of things take a downturn.  But the truth is that they don't need a high salary during an extended, severe downturn.  All they need is a job that can cover a portion of their yearly budget, which is only $25k.  Hardly need a CEO or lawyer level salary to do that....
Frugalite in training.

boarder42

  • Walrus Stache
  • *******
  • Posts: 5249
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #51 on: September 21, 2016, 09:27:54 AM »
Predictions are hard to makes, especially about the future.

When the 4% rule fails (and it probably will) it will likely be for reasons no one saw coming and no one prepared for. Doesn't mean it's not a great starting point though and it should hold up in most scenarios.

The biggest threat I see to MMM style early retirement is there is so much inflexibility to cutting further spending. If you are retiring on 25k/yr it is much harder/impossible to cut spending by 10k than if you are spending 70k. All the fat has already been cut out of the budget. There is more danger of sequence of return risk. Also everyone is healthy in their 30's (more or less). As I see insurance become thinner and more restrictive over time I worry about a medical event causing failure. It is hard to be frugal or go back to work when someone in the family is going through multiple rounds of chemotherapy. In this case even though the 4% rule may work, it may fail if you are unable to keep spending at 4%. We all talk about being flexible and spending less in times of market stress, but few talk about the potential need to spend MORE during these times.

At the end of the day it's about being comfortable with risk. The more redundant systems in place the better.

this may be why i feel safe b/c with our mortgage we spend 60k a year so we have lots of fats that could be cut should we have to.  i also side hustle ... there are so many ways to make money or conserve money. 

there are 2 easy safety nets for both low spending and high spending scenarios.

1. extremely low spending - easy to pick up a job part time to cover expenses - i mean 10k a year is a 10 dollar per hour job working 50% of the typical 40 hour work week. 
2. extremely high spending(by mustachian levels) - easy to trim some of that extra fat like vacations and eating out or whatever your hobby or vice is you cant cut. 

PM me about how to save 6% on your annual grocery Bill!

There is a 35k starwood bonus right now as well. PM me for the info.

BTDretire

  • Handlebar Stache
  • *****
  • Posts: 1191
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #52 on: September 21, 2016, 10:29:35 AM »

Everything old is new again. 

Wait around long enough, bell bottoms will be back in style

 Oh to see 17 year old Debra in her hip hugger bell bottoms again.
But then, that was 1972.

BTDretire

  • Handlebar Stache
  • *****
  • Posts: 1191
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #53 on: September 21, 2016, 10:44:19 AM »
Interesting take.  I don't think DH and I are more likely to declare FIRE in an up market for psychology reasons, but it makes sense that an up market might be enough to push people's stache up to the FIRE target, thus increasing the likelihood of FIRE during an up year. 
That is happening to me, we had a pretty good stache 6 years ago, but our average
yearly stock market gain over the last 6 years is $110k, plus another $40k of savings each year. These gains plus the stache we had is giving me the NW to retire at the end of this year, at a market peak.

Edit 7-19-2017. I learned I don't know if this is a market peak and won't know until the peak has passed.
So replace market peak with all time market high. btw, the S&P and Nasdaq both closed at all time highs yesterday.

  I have three advantages, I can get full SS in 5 years, I have excess stache, and my wife wants to work 5 more years.
 One large disadvantage, my daughter starts dental school soon, at $40k+ per year,
that's why my wife wants to continue working. We should be able to live on her income plus, pay the tuition.
« Last Edit: July 19, 2017, 10:55:52 AM by BTDretire »

dragoncar

  • Walrus Stache
  • *******
  • Posts: 7119
  • Registered member
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #54 on: September 21, 2016, 11:10:50 AM »
Interesting take.  I don't think DH and I are more likely to declare FIRE in an up market for psychology reasons, but it makes sense that an up market might be enough to push people's stache up to the FIRE target, thus increasing the likelihood of FIRE during an up year. 
That is happening to me, we had a pretty good stache 6 years ago, but our average
yearly stock market gain over the last 6 years is $110k, plus another $40k of savings each year. These gains plus the stache we had is giving me the NW to retire at the end of this year, at a market peak.
  I have three advantages, I can get full SS in 5 years, I have excess stache, and my wife wants to work 5 more years.
 One large disadvantage, my daughter starts dental school soon, at $40k+ per year,
that's why my wife wants to continue working. We should be able to live on her income plus, pay the tuition.

Same here, although for me it hasn't been 6 years of huge growth it's just that in the last 1-2 years my investments finally caught up to where I would expect them to be (mean reversion after years of underperformance).

But I think you guys are confusing the idea of a "new high" in the market with a "peak."  We aren't at the peak, or at least we can't know that until the next downturn happens.  We are at a new high.

There's nothing wrong with retiring at a new high of it's only halfway up to the peak.  For example, retiring in 2005 or 2013, I think you'd be fine(although I know we don't have the full 30 years yet). 2008, on the other hand, only time will tell.

rantk81

  • Stubble
  • **
  • Posts: 219
  • Age: 35
  • Location: Chicago
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #55 on: September 22, 2016, 04:06:06 AM »
I calculated that Plan A has 98% success rate while Plan B only has 40%!! (Python yahoo-finance package)

Thank you for posting this and bringing it to my attention.  By the way, do you mind posting your Python code?  I'd like to see it and possibly make some tweaks and run some scenarios.


Mr. Green

  • Handlebar Stache
  • *****
  • Posts: 1269
  • Age: 33
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #56 on: September 22, 2016, 07:14:01 AM »
Your percentages are different because cFIREsim isn't really set up to deal with the accumulation phase. There's no setting where you can say, "I want to FIRE as soon as my portfolio reaches $X, then run for 30 years to see if it'll last in retirement." I don't think comparing the 40% chance of success based on accumulation time to the success rates of a portfolio to go the distance in retirement holds water. It's bad math.

Naturally you're more likely to retire on an up year if you're plan is to hit a number and then quit working. There's also a whole bunch of data that shows the sequence of returns in the early years of retirement are the most important. It goes without saying that if your returns are 0% over the first 10 years of retirement that you are in a much riskier position than the person who averaged a 6% return during those same years. Someone who is a little gun shy may work beyond their number, but remember that when cFIREsim tells you you have a 95% chance of success it has already factored in all the times in history where a retirement is immediately proceeded by bad returns. And in a few of those cases the portfolio fails, as evidenced by the percentage not being 100%.

What cFIREsim doesn't take into effect is our ability to use our brains. On threads like these I see so many people who make comments that allude to the idea that they want some type of method for their spending in retirement that will automatically account for that risk of a downturn after retiring. I know there are a lot of engineers here so I understand (I am one). But no method is required, just thinking. Since we know there is a very strong correlation between poor initial returns and portfolio failure, it's as simple as being proactive if you find yourself in this position.

For example, I FIRE'd in June. Many people would say they're expecting a market downturn soon. Let's say 5 years from now the average return for those 5 years was 0%. Based on sequence of returns risk, I already know that my portfolio now has a substantially higher risk of failure over 30 years, assuming I spent all the money I had planned to during those 5 years. If I didn't curtail my spending at all, I would consider picking up a part-time job to give my portfolio an assist. I'm only spending $40,000 a year from my initial $1 million stash, so a few years of 10-20k income will give me a boost. The next few years' returns would likely dictate how long I kept the part-time job. I could also lower my spending a bit if I didn't want to go back to work. Our bare bones expenses are $25,000 so if I'm really conscious of the sequence of returns risk maybe I just cut my spending hard in the year where the returns are particularly bad. No set method required, just a little brain power and flexibility.
FIRE, Take Two.

Mr. Green

  • Handlebar Stache
  • *****
  • Posts: 1269
  • Age: 33
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #57 on: September 22, 2016, 07:27:16 AM »
There's nothing wrong with retiring at a new high of it's only halfway up to the peak.  For example, retiring in 2005 or 2013, I think you'd be fine(although I know we don't have the full 30 years yet). 2008, on the other hand, only time will tell.
Some data to put this in perspective. The average return for the first 10 years of someone retiring in 2005 was 5.41%, adjusted for inflation. At no time in history has a portfolio with that return during the first 10 years gone on to fail over 30 years. At worst, the portfolio ended up being less than the initial value after 30 years (the orange cells in the link). If you live in the US, the importance of anything longer than 30 years is greatly reduced because SS kicks in and buffers the portfolio, unless you were on the ball from day one and retired in your 20's. So, unless we have some unprecedented bad returns over the next 20 years, a 2005 retiree has already won the game (being their portfolio doing the distance over 30 years). The jury is still out on 2013 because the first 10 years are still happening.

Someone could take my chart one step further and extrapolate what returns would be needed over the next 2 years to make 2008 have have an initial 10 year run that keeps you on a good track. What the hell, I have the excel spreadsheet so I went ahead and did it. If the S&P stays flat and the market ends up 6% for 2016 and then earns 0% in 2017, the first 10 year period for a 2008 retiree will have averaged a 4.53% return. Under no historical circumstances has a portfolio averaged more than a 3% return over the first 10 years and failed over 30 years. Under the scenario I just described, a 2008 retiree has still won the game, baring unprecedented bad returns over the last 20 years of the 30 year period. How 'bout them apples!

NOTE: The caveat is that my excel spreadsheet assumes someone is 100% invested in stocks. I'm sure a 901/10 or 80/20 portfolio would be very similar.
« Last Edit: September 22, 2016, 07:41:18 AM by Mr. Green »
FIRE, Take Two.

SachaFiscal

  • Stubble
  • **
  • Posts: 158
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #58 on: September 22, 2016, 07:47:09 AM »
I wonder how did the OP come up with a period of 5 years after you hit your number to reach 90% success? Could you also look at it in terms of a percentage of your stash buffer? Like if you retired at a peak year but had reached your number plus 20% would that be sufficient to avoid a sequence of returns portfolio failure? If your number was 1m that would just mean you should wait until you have 1.2m and you would be fine.

ender

  • Magnum Stache
  • ******
  • Posts: 3783
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #59 on: September 22, 2016, 08:03:27 AM »
Something to keep in mind is that the market is within a small percentage of it's peak for the majority of the time period cFiresim reflects.

The "what if you retire near a market peak?" is already more or less built into the numbers it runs.

boarder42

  • Walrus Stache
  • *******
  • Posts: 5249
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #60 on: September 22, 2016, 08:38:44 AM »
Something to keep in mind is that the market is within a small percentage of it's peak for the majority of the time period cFiresim reflects.

The "what if you retire near a market peak?" is already more or less built into the numbers it runs.

yes the market is near a peak now ... if its flat/ up a couple percent over the next 2 years it will still be near a peak but the shiller PE ratio will have adjusted
PM me about how to save 6% on your annual grocery Bill!

There is a 35k starwood bonus right now as well. PM me for the info.

Shane

  • Pencil Stache
  • ****
  • Posts: 690
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #61 on: September 22, 2016, 01:19:16 PM »
What cFIREsim doesn't take into effect is our ability to use our brains. On threads like these I see so many people who make comments that allude to the idea that they want some type of method for their spending in retirement that will automatically account for that risk of a downturn after retiring. I know there are a lot of engineers here so I understand (I am one). But no method is required, just thinking. Since we know there is a very strong correlation between poor initial returns and portfolio failure, it's as simple as being proactive if you find yourself in this position.

For example, I FIRE'd in June. Many people would say they're expecting a market downturn soon. Let's say 5 years from now the average return for those 5 years was 0%. Based on sequence of returns risk, I already know that my portfolio now has a substantially higher risk of failure over 30 years, assuming I spent all the money I had planned to during those 5 years. If I didn't curtail my spending at all, I would consider picking up a part-time job to give my portfolio an assist. I'm only spending $40,000 a year from my initial $1 million stash, so a few years of 10-20k income will give me a boost. The next few years' returns would likely dictate how long I kept the part-time job. I could also lower my spending a bit if I didn't want to go back to work. Our bare bones expenses are $25,000 so if I'm really conscious of the sequence of returns risk maybe I just cut my spending hard in the year where the returns are particularly bad. No set method required, just a little brain power and flexibility.

^+1

This is my point as well. If you're flexible and use your brain, things will turn out fine.

If you require 100% proof that things will be okay before you jump, you'll never do it.

Cutting expenses and/or taking a fun part time job or earning some income from a side hustle ≠ failure.

gerardc

  • Bristles
  • ***
  • Posts: 294
  • Age: 33
  • Location: SF bay area
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #62 on: July 15, 2017, 03:14:24 AM »
Alright, there were errors in my previous numbers. The effect is smaller than I initially thought.

What I proposed in the OP is to improve cFIREsim-like simulations, which only consider drawdown periods starting at uniformly distributed years, to more realistic simulations that also consider accumulation periods where investment contributions are made until a FIRE target is reached, at which point the drawdown period starts. This more realistic scenario (plan B) leads to different FIRE starting year distributions, with higher probabilities in peak market years, which in turn lower success rates.

Below are success rates (SR) for 50-year drawdown periods (90% stocks / 10% bonds) for different withdrawal rates (WR), calculated either as by cFIREsim, or by considering MMM-typical (10 years) or worst-case accumulation phases:

WR (%)cFIREsim SR (%)typical SR (%)worst-case SR (%)
3100.0100.0100.0
494.793.081.9
569.162.847.2
642.637.225.0
820.214.01.4

The figure below gives a more complete picture. The main thing to undertand is that very fast accumulation phases reduce down to uniformly distributed FIRE starting years as in cFIREsim (far right of the x axis). As such, the effect of this correction is more pronounced for long, slow accumulation phases and higher WRs, but usually small in typical FIRE scenarios.

I can post the code if someone is interested.



boarder42

  • Walrus Stache
  • *******
  • Posts: 5249
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #63 on: July 15, 2017, 05:07:26 AM »
For the retiring at peaks issue you can just look at kitces work on the shiller pe to swr relationship.  Right now it would indicate a 3.3 swr. But it's deflated by the horrible earnings of 08-09.  So it's still a little high but probably closer to 3.7-3.8. Which is right in the realm of error for the study to make a 4% swr work.

But at the end of the day these things are all just historically based tools to help you feel comfortable with your FIRE plans. Only each individual can decide what makes sense for them. Personally I'll probably work 60% of one extra year when we hit our FIRE number. Plus we'll make money when we are retired and inherit money etc.
PM me about how to save 6% on your annual grocery Bill!

There is a 35k starwood bonus right now as well. PM me for the info.

maizeman

  • Handlebar Stache
  • *****
  • Posts: 1095
  • Location: The World of Tomorrow
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #64 on: July 15, 2017, 08:29:43 AM »
Thanks gerardc. The new numbers seem like a more reasonable effect size but the bias you identified is clearly still at play.

I can post the code if someone is interested.

Thanks, I'd be fascinated to take a look at it.
"Itís a selective retirement," Richard explained, "a retirement from boring s**t."

My source code & my journal

fuzzy math

  • Stubble
  • **
  • Posts: 164
  • Location: PNW ---> Midwest (for now)
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #65 on: July 15, 2017, 08:44:41 AM »
Thanks everyone for such a great thread! I, for one, am thrilled when someone has a new take (or in the eyes of others, recycles) an old idea. It allows new comers such as myself the chance to learn.

This might be a silly and unrelated question, but are all of you number crunchers engineers? I know Go Curry Cracker and MMM were. I am always fascinated by the backgrounds of the people who do the hard work of sorting this stuff out for people like me to come capitalize off their ideas for free :D


radram

  • Bristles
  • ***
  • Posts: 495
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #66 on: July 15, 2017, 09:05:22 AM »

- Life Insurance.  My wife and I each have a 500,000 universal life policy on each other. 

Not to completely take over this thread, but why in the world do you have these policies, and what are they costing you?

With so many backup plans, and the fact that since you are on this site you are likely to FIRE at an early age, you are probably one with the least need for a whole life plan(with my overall opinion being that NOBODY needs whole life).

A FIRE couple needs NO life insurance because you are a liability in FIRE, not an asset. Your remaining stashe need only fund one person instead of the planned 2 people.

I would cash out a whole life plan and buy a term life plan until FIRE to get the surviving spouse to FIRE upon the others death.

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 25299
  • Age: -999
  • Location: Traveling the World
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #67 on: July 15, 2017, 09:16:12 AM »
Alright, there were errors in my previous numbers. The effect is smaller than I initially thought.

What I proposed in the OP is to improve cFIREsim-like simulations, which only consider drawdown periods starting at uniformly distributed years, to more realistic simulations that also consider accumulation periods where investment contributions are made until a FIRE target is reached, at which point the drawdown period starts. This more realistic scenario (plan B) leads to different FIRE starting year distributions, with higher probabilities in peak market years, which in turn lower success rates.

Below are success rates (SR) for 50-year drawdown periods (90% stocks / 10% bonds) for different withdrawal rates (WR), calculated either as by cFIREsim, or by considering MMM-typical (10 years) or worst-case accumulation phases:

WR (%)cFIREsim SR (%)typical SR (%)worst-case SR (%)
3100.0100.0100.0
494.793.081.9
569.162.847.2
642.637.225.0
820.214.01.4

The figure below gives a more complete picture. The main thing to undertand is that very fast accumulation phases reduce down to uniformly distributed FIRE starting years as in cFIREsim (far right of the x axis). As such, the effect of this correction is more pronounced for long, slow accumulation phases and higher WRs, but usually small in typical FIRE scenarios.

I can post the code if someone is interested.

Thanks for following up with this, and with chart and graph!  :)

Very interesting.  Makes sense that faster FIRE time = less affected.  Once again, high savings rate wins out!
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with a kid.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

PizzaSteve

  • Bristles
  • ***
  • Posts: 493
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #68 on: July 15, 2017, 09:20:59 AM »
Yes, this is mostly another way of illustrating sequence of returns risk.

My hunch is this: the shorter your investing timeline, the more likely you are to be prone to sequence of returns risk, and vice-versa. But even then, you aren't necessarily more likely to be at the top of the market, as there's no way to predict that. I agree that you're less likely to retire in a bear market, but that doesn't mean the opposite is true.

It's easy to cherry pick peaks and assume that one retires on the peaks and then see how likely those 30 yr periods are likely to fail due to sequence risk. It's very possible that your overall savings increases faster than market returns, and you pull the plug well before the peak so that you're less exposed to sequence risk.

While you're much more likely to hit "your number" in a bull market, that doesn't necessarily mean that it's a peak.

The other aspect that worries me about achieving "FIRE" is that the guidance here seems to be to just shoot for 25x of your expenses. What's more accurate is that you need 25x of your predicted expenses in retirement. Rosily subtracting out taxes and other job-related expenses really seems to ignore the fact that future health care costs (as one example) will substantially increase as you age. Maybe some people are okay with taking that chance and relying on the government to bail them out, but I really don't see how these people retiring with a 500K net worth will have what's considered successful retirements if we could look back in 50 years.

Most people's numbers do include taxes as well as a buffer for potential increases in spending. If you're retiring on $500k, it's unlikely you'll be paying any taxes at all. And in all likelihood, you will spend less and have an even larger buffer than you thought.
This (above).

1. Excellent post for those who don't yet understand sequence risk. Educational and clear.

2. I note that people should remember that investment plan failure is not retirement plan failure.  It simply means your income from the stash is less than planned income, thats it.  Usually that means backup strategies kick in...no problem. Typical backup plans are listed above: spend less, save more of a cushion, rely on inheritance or SS, go back to work.  None of these are inherently better than others, and some have no back-up plan, which can be ok too, just accepting more risk, should the sequence work against you.

3. Finally, people should hopefully understand that backtesting has limits too, even when well designed.  It is guaranteed that the future sequence of returns will not be equal to any of the historical sequence of returns.  FUTURE RETURN ARE AN UNKNOWN.  This is true by definition and is a fundamental limitation to all investment strategies built upon backward looking assumptions.  It doesnt invalidate the strategy, but it is a risk to understand.  Insurance company actuaries are experts in this, but also occasionally fail. Even an insurance company can go bankrupt, so skate on the ice that is the thickness you can accept.

4. I am sooooo tempted to reply to young, naive investors citing Shiller PE and saying that the market is at a peak (having no clue what the right side of the chart will look like, but will pass.  One sequence of returns is unprecidented prosperity due to technology innovation, robotics, AI, biotech, materials advances and the efficiencies these will bring.  Shiller, Sharpe, and all economists are real people i have met, looked at their papers, etc.  WIKI why it is called the Dismal Science.  LOL. Again, none of us know.  We can only predict.  My plans trust in the global corporate/govt money machine to fund my retirement.  I hope we all can fund the life we want in the future via our frugality and delayed gratification ethos.

PS likely to be among my last posts, so no replies needed.  PMs will be read.  Cheers.
« Last Edit: July 15, 2017, 09:41:08 AM by PizzaSteve »
All posts are opinions of the author subject to independent verification by the reader.  No representations of fact are asserted regarding commercial products or services.

gerardc

  • Bristles
  • ***
  • Posts: 294
  • Age: 33
  • Location: SF bay area
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #69 on: July 15, 2017, 02:00:53 PM »
Below is a clearer graph of success rate in function of savings rate instead of contribution rate. At typical 4% SWR and 50% savings rate, cFIREsim predicts 94.6% success, which goes down to 91.6% when considering the effect of market fluctuations on FIRE start year.



My hunch is this: the shorter your investing timeline, the more likely you are to be prone to sequence of returns risk, and vice-versa. But even then, you aren't necessarily more likely to be at the top of the market, as there's no way to predict that. I agree that you're less likely to retire in a bear market, but that doesn't mean the opposite is true.

Below is a plot of the FIRE year distribution for accumulation phases starting at random years for different savings rates. At 90% savings rate, FIRE density is roughly uniform; at 10% savings rate, FIRE density is almost all concentrated around market peaks right before crashes; in between (50%), FIRE density is surprisingly also at the mercy of market fluctuations, which explains the correction above.




Thanks, I'd be fascinated to take a look at it.

Here you go. I wrote this quickly so there might be bugs.

Code: [Select]
import json
import matplotlib.pyplot as plt
import numpy as np

START_YEAR = 1871
END_YEAR = 2015
PERIOD = 50
STOCK_FRACTION = 0.9
BONDS_FRACTION = 0.1

# From https://github.com/boknows/cFIREsim-open/blob/master/js/marketData.js
market_data = json.load(file('market.json'))
market_data = {int(year): properties for year, properties in market_data.iteritems()}

def Simulate(start_value, start_year, end_year, accumulation, target):
  value = start_value
  die = None
  reached = None
  for year in xrange(start_year, end_year):
    properties = market_data[year]
    next_properties = market_data[year + 1]

    value += accumulation
    if value > 0:
      stock_gains = (1. + properties['dividends']) * (1. + properties['growth'])
      bonds_gains = 1. + properties['fixed_income']
      value *= STOCK_FRACTION * stock_gains + BONDS_FRACTION * bonds_gains
    value *= float(properties['cpi']) / next_properties['cpi']

    if value <= 0 and die is None:
      die = year
    if target is not None and value >= target and reached is None:
      reached = year + 1
    #print year, value
  return {'end_value': value, 'die': die, 'reached': reached}

def SuccessRate(period, accumulation, withdrawal):
  successes = []
  durations = []
  fire_years = []
  for start_year in xrange(START_YEAR, END_YEAR - period):
    accumulation_results = Simulate(0.0, start_year, END_YEAR, accumulation, 1.0)
    if accumulation_results['reached'] is None:
      continue
    reached_year = accumulation_results['reached']
    fire_years.append(reached_year)
    if reached_year + period > END_YEAR:
      continue

    drawdown_results = Simulate(1.0, reached_year, reached_year + period, -withdrawal, target=None)
    die_year = drawdown_results['die']
    success = die_year is None
    duration = period if die_year is None else die_year - reached_year
    successes.append(success)
    durations.append(duration)
  return {'success_rate': np.mean(successes), 'mean_duration': np.mean(durations), 'fire_years': fire_years}

def Accumulation(savings_rate, withdrawal):
  # Assuming expenses (as fraction of target) = withdrawal rate
  # savings_rate = accumulation / (accumulation + withdrawal)
  return withdrawal / (1. / savings_rate - 1.)

def PlotSuccessRate():
  savings_rates = np.arange(0.1, 1., .05)
  withdrawals = [0.03, 0.04, 0.05, 0.06, 0.07]

  legend = []
  for withdrawal in withdrawals:
    success_rates = [100. * SuccessRate(PERIOD,
                                        Accumulation(savings_rate, withdrawal),
                                        withdrawal)['success_rate']
                     for savings_rate in savings_rates]
    plt.plot(100. * savings_rates, success_rates)
    legend.append('%i%% WR' % (100. * withdrawal))
    print '%i%% WR:' % (100. * withdrawal),
    print ' '.join('(%i,%.1f)' % t for t in zip(100. * savings_rates, success_rates))

  plt.xlabel('accumulation phase savings rate (%)')
  plt.ylabel('success rate (%)')
  plt.legend(legend)
  plt.title('Historical success rates over %i years after accumulation phase\n(%i%% stock / %i%% bonds portfolio)' %
            (PERIOD, STOCK_FRACTION * 100, BONDS_FRACTION * 100))
  plt.grid(True)
  plt.show()

def PlotFireYearDensity():
  savings_rates = [0.1, 0.5, 0.9]
  withdrawal = 0.04

  years = np.arange(START_YEAR, END_YEAR, 0.1)
  smoothing_years = 2.
  def kernel(fire_year):
    return np.exp(-0.5 * ((years - fire_year) / smoothing_years) ** 2)

  legend = []
  for savings_rate in savings_rates:
    fire_years = SuccessRate(0, Accumulation(savings_rate, withdrawal), withdrawal)['fire_years']
    distribution = np.sum([kernel(fire_year) for fire_year in fire_years], axis=0)
    plt.plot(years, distribution, linewidth=1.0)
    legend.append('%i%% SR' % (100. * savings_rate))

  plt.xlabel('year')
  plt.ylabel('FIRE density')
  plt.legend(legend)
  plt.title('FIRE year distribution under targeted accumulation phase\n(%i%% stock / %i%% bonds portfolio)' %
       (STOCK_FRACTION * 100, BONDS_FRACTION * 100))
  plt.grid(True)
  plt.show()
« Last Edit: July 15, 2017, 02:04:00 PM by gerardc »

SwordGuy

  • Magnum Stache
  • ******
  • Posts: 3015
  • Location: Fayetteville, NC
    • Flipping Fayetteville
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #70 on: July 15, 2017, 02:19:58 PM »
The peaks seem to be smooth slopes on the climb with cliffs on the backside as the markets panic. Based on that, I would expect a stronger aversion to RE in a year like 2008 than a bias towards 2007.

To be accurate, you are saying "I'm more likely to regret retiring in the year before a big crash after I figure out in the next year that's what happened."

msilenus

  • Bristles
  • ***
  • Posts: 485
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #71 on: July 15, 2017, 02:20:56 PM »
garardc, this is a clever way of quantifying forces that I felt underequipped to grapple with.  Thank you for doing this.

SwordGuy

  • Magnum Stache
  • ******
  • Posts: 3015
  • Location: Fayetteville, NC
    • Flipping Fayetteville
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #72 on: July 15, 2017, 02:51:56 PM »
CFiresim assumes you never work again.

That is simply not true.

CFiresim makes no such assumption.

CFiresim allows you to enter in expected income from any source in the years after you "retire".

I, for example, have entered in rental income, income from making art, and occasional consulting or teaching work. 

I could as easily enter in working part or full time at Hooters in the kitchen at minimum wage ('cause I sure am not going to get hired as a Hooter's girl) for several years. 

In fact, it would be instructive to do so, to see if that low-paying a job would take you from a 95% to a 100% success path.

Padonak

  • 5 O'Clock Shadow
  • *
  • Posts: 62
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #73 on: July 16, 2017, 12:10:20 PM »
Below is a clearer graph of success rate in function of savings rate instead of contribution rate. At typical 4% SWR and 50% savings rate, cFIREsim predicts 94.6% success, which goes down to 91.6% when considering the effect of market fluctuations on FIRE start year.

Excellent analysis, thanks a lot for posting this!

I have a question about using PE10 to determine SWR: according to Mad Fientist who I believe used Kitces' analysis, the following formula can be used for SWR:

Safe Withdrawal Rate ~ 1/Shiller PE Ratio

The current PE10 is about 30, so the SWR is 3.333%.

Question: has this relationship been backtested? If so, how extensively? Would you rely on this formula to calculate your safe withdrawal rate?



Classical_Liberal

  • Pencil Stache
  • ****
  • Posts: 697
  • Age: 41
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #74 on: July 16, 2017, 04:19:21 PM »
@gerardc 

I third your very excellent post, graphics and analysis.

boarder42

  • Walrus Stache
  • *******
  • Posts: 5249
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #75 on: July 16, 2017, 04:33:59 PM »
Below is a clearer graph of success rate in function of savings rate instead of contribution rate. At typical 4% SWR and 50% savings rate, cFIREsim predicts 94.6% success, which goes down to 91.6% when considering the effect of market fluctuations on FIRE start year.

Excellent analysis, thanks a lot for posting this!

I have a question about using PE10 to determine SWR: according to Mad Fientist who I believe used Kitces' analysis, the following formula can be used for SWR:

Safe Withdrawal Rate ~ 1/Shiller PE Ratio

The current PE10 is about 30, so the SWR is 3.333%.

Question: has this relationship been backtested? If so, how extensively? Would you rely on this formula to calculate your safe withdrawal rate?

It was derived from historical data therefore back tested. It needs more analysis than just looking at it as it sits now the shiller is very inflated due to 09 earnings plummet in a couple years that will move off and it will normalize some. We're really not as high as shiller makes it look right now.

I as I said above rely on this formula from a simple data point that I'll use when I fire to help determine if I personally think my plan will work.

For example I'd feel very comfortable retiring today with 25x my expenses even with the shiller that high due to the reasons I think it's inflated above.

The individual has to take all data they have and apply it to their situation and determine if they are personally comfortable with the safety of their own fire plan.
PM me about how to save 6% on your annual grocery Bill!

There is a 35k starwood bonus right now as well. PM me for the info.

Padonak

  • 5 O'Clock Shadow
  • *
  • Posts: 62
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #76 on: July 16, 2017, 07:49:54 PM »
Below is a clearer graph of success rate in function of savings rate instead of contribution rate. At typical 4% SWR and 50% savings rate, cFIREsim predicts 94.6% success, which goes down to 91.6% when considering the effect of market fluctuations on FIRE start year.

Excellent analysis, thanks a lot for posting this!

I have a question about using PE10 to determine SWR: according to Mad Fientist who I believe used Kitces' analysis, the following formula can be used for SWR:

Safe Withdrawal Rate ~ 1/Shiller PE Ratio

The current PE10 is about 30, so the SWR is 3.333%.

Question: has this relationship been backtested? If so, how extensively? Would you rely on this formula to calculate your safe withdrawal rate?

It was derived from historical data therefore back tested. It needs more analysis than just looking at it as it sits now the shiller is very inflated due to 09 earnings plummet in a couple years that will move off and it will normalize some. We're really not as high as shiller makes it look right now.

I as I said above rely on this formula from a simple data point that I'll use when I fire to help determine if I personally think my plan will work.

For example I'd feel very comfortable retiring today with 25x my expenses even with the shiller that high due to the reasons I think it's inflated above.

The individual has to take all data they have and apply it to their situation and determine if they are personally comfortable with the safety of their own fire plan.

But isn't the whole point of using PE10 to include recession years in profit calculations, not only good years? Given that normally, every 10 years or so there is a recession.

DavidAnnArbor

  • Pencil Stache
  • ****
  • Posts: 966
  • Age: 51
  • Location: Ann Arbor, Michigan
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #77 on: July 16, 2017, 08:51:35 PM »
There have been some accounting changes in tech companies so I'm not sure how well the PE 10 can be entirely accurate. Nonetheless, 30 seems historically high, so it's definitely good to be aware of that in thinking about how to handle a downturn in the stock market if one is on the verge of retiring.

Shiller PE 10 Ratio by Month
https://www.quandl.com/data/MULTPL/SHILLER_PE_RATIO_MONTH-Shiller-PE-Ratio-by-Month
« Last Edit: July 16, 2017, 09:15:53 PM by DavidAnnArbor »

obstinate

  • Pencil Stache
  • ****
  • Posts: 637
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #78 on: July 16, 2017, 09:22:26 PM »
The probabilistic problem solved by cFiresim/FIRECalc has the following form: You start with $1m at a random year, you spend $40k yearly from then on, what's the probability of success over 50 years?

The actual plan people follow is more like: You start at 0, you save $50k yearly, when you reach $1m you stop working, then you start spending $40k yearly, what's your probability of success over 50 years?

The 2 formulations look very similar, but the probabilities are very different. Can you see why?

I calculated that Plan A has 98% success rate while Plan B only has 40%!! (Python yahoo-finance package)

The reason is that under Plan B, the year you stop working is not actually uniformly distributed -- you're much more likely to retire on a peak year, and unlikely to retire just after a crash. This makes a big difference, because cFiresim assumes you could have retired on $1m in 2009, but that wouldn't happen on Plan B because you'd have had $2m in 2007 and would have retired then. On Plan B, you're more likely to have only $500k in 2009.

A better approach to increase your probability of success is Plan C: same as Plan B, but you're only allow to retire when your target has been reached for 5 years in a row. Success rate is back at 90% but you just lost 5 years... So, please be careful.

Update with analysis and figure
Wow, so, this is absolutely correct. The dependence of the early retirement variable with recent financial performance is super important. Now, assuming you can adjust your spending, you're still probably in OK shape, but this makes engineering tolerances that much more important. I think this adds considerably to the early retirement discussion, and I thank you for posting it.

Classical_Liberal

  • Pencil Stache
  • ****
  • Posts: 697
  • Age: 41
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #79 on: July 17, 2017, 09:39:22 AM »
But isn't the whole point of using PE10 to include recession years in profit calculations, not only good years? Given that normally, every 10 years or so there is a recession.

You are correct, even if you take out the unusual 2009 situation, and take a PE8, for example, CAPE is still very high (about a half point lower).  There are a myriad of other reasons for this phenomenon in the current marketplace.  By far the largest (most obvious) are long-term low yields on Treasuries & long-term low inflationary trends. 

I think accounting standards are another big one, along with availability of accurate data (long term CAPE data includes eras of less regulatory disclosure requirements).  These are factors which are more difficult to directly assess.

There have also been suggestions that indexing, 401k's, and other low cost investment options for the masses have permanently altered publicly traded companies valuations; pushing more risky and volatile investments into the realm of wealthier private investors (which is what the public markets were in the past). Again, this is difficult to look at directly.

runewell

  • Bristles
  • ***
  • Posts: 370
  • Age: 45
  • Location: IA
  • Bridge-playing actuary
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #80 on: July 17, 2017, 09:52:45 AM »
Do your simulations account for the fact that the market has been on a bit of a tear and the long-term upside is expected to be lower?
High CAPE ratios are a good predictor of low returns going forwards. 

http://www.starcapital.de/files/publikationen/Research_2016-01_Predicting_Stock_Market_Returns_Shiller_CAPE_Keimling.pdf

check out pages 10-11

If your simulations start now with a CAPE ratio of 30 and assume you can sample freely from historical returns, you are probably overestimating the return and survival rate.

Looking at the historical CAPE ratios, we are at the level of the great depression just before the stock market crash.
« Last Edit: July 17, 2017, 09:55:02 AM by runewell »

gerardc

  • Bristles
  • ***
  • Posts: 294
  • Age: 33
  • Location: SF bay area
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #81 on: July 17, 2017, 10:29:51 AM »
Do your simulations account for the fact that the market has been on a bit of a tear and the long-term upside is expected to be lower?
High CAPE ratios are a good predictor of low returns going forwards. 

This is a Trinity study-like simulation that only compiles historical success rates of the aforementioned accumulation/withdrawal strategies; it is not Monte Carlo and does not attempt (at least not explicitly) to predict the success rate as of 2017.

runewell

  • Bristles
  • ***
  • Posts: 370
  • Age: 45
  • Location: IA
  • Bridge-playing actuary
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #82 on: July 17, 2017, 10:46:38 AM »
Do your simulations account for the fact that the market has been on a bit of a tear and the long-term upside is expected to be lower?
High CAPE ratios are a good predictor of low returns going forwards. 

This is a Trinity study-like simulation that only compiles historical success rates of the aforementioned accumulation/withdrawal strategies; it is not Monte Carlo and does not attempt (at least not explicitly) to predict the success rate as of 2017.

Hmm that's not good.  People are blindly assuming that returns will follow the historical distribution when research shows that below-average returns should be expected. 

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 25299
  • Age: -999
  • Location: Traveling the World
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #83 on: July 17, 2017, 02:15:37 PM »
Do your simulations account for the fact that the market has been on a bit of a tear and the long-term upside is expected to be lower?
High CAPE ratios are a good predictor of low returns going forwards. 

This is a Trinity study-like simulation that only compiles historical success rates of the aforementioned accumulation/withdrawal strategies; it is not Monte Carlo and does not attempt (at least not explicitly) to predict the success rate as of 2017.

Hmm that's not good.  People are blindly assuming that returns will follow the historical distribution when research shows that below-average returns should be expected.
Historical modeling doesn't assume it'll be like average, it assumes it won't be worst than the worst ever.

Below average returns have happened before. Believe it or not, around half the time. ;)

Years of low or negative returns have happened in the past. That's taken into account.
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with a kid.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

sol

  • Walrus Stache
  • *******
  • Posts: 5223
  • Age: 40
  • Location: Pacific Northwest
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #84 on: July 17, 2017, 04:26:03 PM »
If you assume that future returns will look like average past returns, the correct SWR for a 30 year retirement is north of 6%.

The 4% SWR has been successful in 95% of starting years, not 50% of years.  Once you're below 6%, you've probably already oversaved and are working too long.  That's okay, most of us would prefer to risk leaving a big inheritance instead of risking having to reduce expenses in retirement.  Just don't fool yourself into thinking you HAVE to work longer in order to leave a big inheritance.

Lan Mandragoran

  • 5 O'Clock Shadow
  • *
  • Posts: 68
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #85 on: July 17, 2017, 08:31:48 PM »
Oh really? didnít know that. How does 6% do historically ? 5%?
Gradatim Ferociter
Step by step, ferociously

Vi veri veniversum vivus vici
"By the power of truth, I, while living, have conquered the universe"

gerardc

  • Bristles
  • ***
  • Posts: 294
  • Age: 33
  • Location: SF bay area
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #86 on: July 17, 2017, 08:53:11 PM »
Oh really? didn’t know that. How does 6% do historically ? 5%?

See the graph above in the limit of 100% savings rate (far right). Historically, 5% WR has ~70% success rate, 6% WR has 45%.

However at high WRs, it becomes more interesting to include flexible withdrawal strategies in the picture. I'm working on extending these simulations to quantify how high of a WR we can go with some amount of flexibility.

sol

  • Walrus Stache
  • *******
  • Posts: 5223
  • Age: 40
  • Location: Pacific Northwest
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #87 on: July 17, 2017, 10:15:11 PM »
Oh really? didnít know that. How does 6% do historically ? 5%?

From page 3 of the 4% thread:


runewell

  • Bristles
  • ***
  • Posts: 370
  • Age: 45
  • Location: IA
  • Bridge-playing actuary
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #88 on: July 18, 2017, 08:10:49 AM »
Years of low or negative returns have happened in the past. That's taken into account.

Not sufficiently, in my opinion. 

The authors basically said, oh you have 20 years to eat away at your retirement?  We'll look at as many 20-yr consecutive time periods as we can in our study and see how things fared.  However, they don't account for the possibility that the market starts out very expensive and that subsequent years could perform poorly.  Also, the study is also nearly 20 years old.   

The CAPE research says, what is the long-term returns 10-15 years from now given we are starting at a specific CAPE ratio?  A CAPE ratio of <10 should has an expected return of 11.7% while a CAPE ratio of 30+ has an expected return of 0.5%.  This research is 18 months old, and it explicitly recognizes some CAPE shortcomings and attempts to address them or adjust for them.

http://www.starcapital.de/files/publikationen/Research_2016-01_Predicting_Stock_Market_Returns_Shiller_CAPE_Keimling.pdf

If you read the newer study, I just think we should be skeptical of the 4% rule given that we are in an expensive market and proceed cautiously.

 

Lan Mandragoran

  • 5 O'Clock Shadow
  • *
  • Posts: 68
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #89 on: July 18, 2017, 08:36:09 AM »
Oh really? didnít know that. How does 6% do historically ? 5%?

See the graph above in the limit of 100% savings rate (far right). Historically, 5% WR has ~70% success rate, 6% WR has 45%.

However at high WRs, it becomes more interesting to include flexible withdrawal strategies in the picture. I'm working on extending these simulations to quantify how high of a WR we can go with some amount of flexibility.

Yeah... thanks. That would be very interesting data to see.

For example my wife works as a physical therapist just a few days of the week, but i'm guessing even post FIRE she would work occasionally (maybe 1 day a week idk) so even just that hundred bucks of income/week would likely change things quite a bit. Especially since either of us could always pick up some work during severe dips quite easily.
Gradatim Ferociter
Step by step, ferociously

Vi veri veniversum vivus vici
"By the power of truth, I, while living, have conquered the universe"

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 25299
  • Age: -999
  • Location: Traveling the World
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #90 on: July 18, 2017, 04:44:03 PM »
Also, the study is also nearly 20 years old.

And you think no one has updated it since?

Quote
If you read the newer study, I just think we should be skeptical of the 4% rule given that we are in an expensive market and proceed cautiously.

Stop worrying about the 4% rule.
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with a kid.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

GenXbiker

  • 5 O'Clock Shadow
  • *
  • Posts: 33
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #91 on: July 19, 2017, 07:45:36 AM »
A CAPE ratio of <10 should has an expected return of 11.7% while a CAPE ratio of 30+ has an expected return of 0.5%.
.5% real gains?  At 50 and an expected FIRE at 52 to 54, I would have to delay my retirement an additional 4 years to have the same "extra" amount of cash to spend each month as I would using a SWR of 4%, and that's even factoring the boost I would get in SS at age 62 from working those extra years.   However, either way, I have a decent enough cushion that I wouldn't extend my working career, even if I could be certain of .5% real gains.  I suspect most here expect their retirement to be longer than the 10 to 15 year time period referred to in the document.

runewell

  • Bristles
  • ***
  • Posts: 370
  • Age: 45
  • Location: IA
  • Bridge-playing actuary
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #92 on: July 19, 2017, 09:27:54 AM »
And you think no one has updated it since?

I don't know for sure.  But, I don't trust MMM or the general public to either (1) read the actual study or (2) sufficiently critically consider its merits and drawbacks.

I think there is a lot of good information in the study, such as the benefit from maintaining a reasonably high allocation of equities.

As investment returns benefit from diversification, I also think it's safer to find wisdom in a lot of different viewpoints rather than to cling to a single piece of research. 

I also worry that the very name of that thread "Stop Worrying about the 4% Rule" encourages people to accept it blindly. 
« Last Edit: July 19, 2017, 09:37:03 AM by runewell »

sol

  • Walrus Stache
  • *******
  • Posts: 5223
  • Age: 40
  • Location: Pacific Northwest
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #93 on: July 19, 2017, 10:46:10 AM »
I also worry that the very name of that thread "Stop Worrying about the 4% Rule" encourages people to accept it blindly.

I don't even accept gravity blindly, but I do accept it and encourage others to do the same.

It astounds me that anyone could read through the hundreds of pages of careful mathematical analysis this forum has done about safe withdrawal rates, and then accuse us of blind faith because of a thread title summarizing our findings.

Playing with Fire UK

  • Handlebar Stache
  • *****
  • Posts: 1501
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #94 on: July 19, 2017, 10:57:07 AM »
Thanks for the interesting post.

I'm having to consider this as my 'stache' has spiked in value (invested internationally, measured in GBP) following the currency being devalued. There was a moment where my stache (in £) covered 25x my barebones expenses for the previous year (when £ had higher purchasing power, and I took a cheaper holiday than I normally budget for).

Of course it was pretty obvious that something was up, so I although I hadn't explicitly put it into my plan, I needed to know what my expenses were going forward rather than only looking backwards. But I'd also have needed to plan for this if I'd had a particularly cheap or expensive year, and that wasn't in my plan either.

So this is a real thing, but I'm confident in our community's ability to identify when something is up. Like that glitch in PC which had people's mortgages showing as positive, or that pricing anomaly on AIM stocks a while ago. I didn't hear of anyone quitting work years early based on a glitch without thinking it through.

I think most people get 'close' and then set a date for when they are stopping work; this has a built-in safety buffer form severe market movements (assuming that you don't announce your intentions by shitting on your bosses desk or other career returning move). 

For example my wife works as a physical therapist just a few days of the week, but i'm guessing even post FIRE she would work occasionally (maybe 1 day a week idk) so even just that hundred bucks of income/week would likely change things quite a bit. Especially since either of us could always pick up some work during severe dips quite easily.

Careful that you remember that market dips are not independent of difficulty finding work. If you are looking for work as a bailiff or in a payday loan shop you are likely to be in luck (gut feeling, no data to back this).

Playing with Fire UK

  • Handlebar Stache
  • *****
  • Posts: 1501
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #95 on: July 19, 2017, 11:03:38 AM »
I also worry that the very name of that thread "Stop Worrying about the 4% Rule" encourages people to accept it blindly.

I don't even accept gravity blindly, but I do accept it and encourage others to do the same.

It astounds me that anyone could read through the hundreds of pages of careful mathematical analysis this forum has done about safe withdrawal rates, and then accuse us of blind faith because of a thread title summarizing our findings.

And you think no one has updated it since?
I don't know for sure.  But, I don't trust MMM or the general public to either (1) read the actual study or (2) sufficiently critically consider its merits and drawbacks.
I think there is a lot of good information in the study, such as the benefit from maintaining a reasonably high allocation of equities.
As investment returns benefit from diversification, I also think it's safer to find wisdom in a lot of different viewpoints rather than to cling to a single piece of research. 
I also worry that the very name of that thread "Stop Worrying about the 4% Rule" encourages people to accept it blindly. 

I'm still nursing my hurt feelings from being called a member of the general public. We have a higher than average spread of data nerds here on MMM and are way above average in our tendency to share our learning and challenge each other to learn more. It would seem unlikely that a community of people who are almost entirely resistant to advertising would be swayed by the name of a single internet thread for something as importance as our financial security.

runewell

  • Bristles
  • ***
  • Posts: 370
  • Age: 45
  • Location: IA
  • Bridge-playing actuary
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #96 on: July 19, 2017, 11:05:20 AM »
It astounds me that anyone could read through the hundreds of pages of careful mathematical analysis this forum has done about safe withdrawal rates, and then accuse us of blind faith because of a thread title summarizing our findings.

If the there is no reason to worry about the 4% withdrawal rule (as per the thread title), why is hundreds of pages of careful mathematical analysis going on then?

sol

  • Walrus Stache
  • *******
  • Posts: 5223
  • Age: 40
  • Location: Pacific Northwest
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #97 on: July 19, 2017, 11:38:57 AM »
It astounds me that anyone could read through the hundreds of pages of careful mathematical analysis this forum has done about safe withdrawal rates, and then accuse us of blind faith because of a thread title summarizing our findings.

If the there is no reason to worry about the 4% withdrawal rule (as per the thread title), why is hundreds of pages of careful mathematical analysis going on then?

Ahh, that got a chuckle.  First the forum is wrong because they didn't think about it, them they're wrong because they thought about it too much.  Well, which is it?

Whatever man, I've done more than my fair share of free internet helping.  At this point, if you don't like my advice then you should ignore it and do your own thing. Smoothly functioning markets require a diversity of opinions, and every trade has a winner and a loser.  If you don't like a 4% SWR, then by all means choose something else and please post about it so we can all keep tabs on you.

For example, I'm shooting for 5% and I still expect to die with more money than I started with.

runewell

  • Bristles
  • ***
  • Posts: 370
  • Age: 45
  • Location: IA
  • Bridge-playing actuary
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #98 on: July 19, 2017, 12:19:02 PM »

Ahh, that got a chuckle.  First the forum is wrong because they didn't think about it, them they're wrong because they thought about it too much.  Well, which is it?


I don't think you should characterize the forum as being wrong, I don't think they like it.

My concern was that a title like "Stop worrying about the 4% rule" could encourage the masses to blindly accept something without giving it careful consideration.   Perhaps a title like that will instead lure people into a discussion in which case they educate themselves.  I just think it's ironic that the title says stop worrying but the amount of research in the thread says the exact opposite.  It's false advertising.

DavidAnnArbor

  • Pencil Stache
  • ****
  • Posts: 966
  • Age: 51
  • Location: Ann Arbor, Michigan
Re: cFiresim SEVERELY overestimates success rates for Mustachians
« Reply #99 on: July 19, 2017, 12:27:04 PM »
My middle name is Worry.