Protracted recession and it's effects

[kcowan]

Originally Posted by Canadian Grunt

The problem is for those retiring in their late 30's early 40's and relying on a 4% withdrawal rate. If the market goes sideways for an extended period of time or we get stagflation there will be problems for them moving forward if they put too much trust in Monte Carlo predictions.

ISTR that Firecalc is best for a 30 year projection. If you are around 40, you probably are foregoing your high earning years during your high spending years. Not a good combination.

When I was forty, I had an acreage with a 5000sq.ft house, all the toys, 2 kids in school, four cars and lots of other high expenses. Taking a holiday was a major production and costly. Food expense was a major item.

[FUEGO]

Originally Posted by Canadian Grunt

The problem is for those retiring in their late 30's early 40's and relying on a 4% withdrawal rate. If the market goes sideways for an extended period of time or we get stagflation there will be problems for them moving forward if they put too much trust in Monte Carlo predictions.

I fall into the category of "late 30's or early 40's" early retiree (or that's the plan anyway). I tend to agree with your assertion that a 4% SWR with annual CPI inflation adjustments could be problematic over the course of what may be five or six decades.

I personally plan to use a more flexible withdrawal strategy that ties some or all of my withdrawal amount to portfolio value.

And I haven't ruled work in some form out. Based on analysis I have conducted (some of which I have posted on this forum), if you make it through the first 10 to 15 years with your portfolio substantially intact, then you are likely to be ok thereafter. For me, it wouldn't be that hard to get back to working within the first 10 to 15 years. Even a part time job or something that paid little but provided health insurance would drop my withdrawal rate significantly.

It seems like most folks that are ER-ing at 35-45 have some sort of back up plan and/or a flexible withdrawal strategy. I would suggest that the fact that one is able to embark on an ER at age 35-45 is evidence that they are pretty good at planning (or they got really lucky!). If I were to retire at 40 and see my portfolio cut in half by age 45, I would probably not continue to take the annual 4% inflation adjusted withdrawals - I would start implementing plan B.

I don't think you will find anyone here arguing that the 4% rule is a hard and fast adamant rule that is guaranteed to work. It is just a general guideline that established a high degree of certainty based on historical data. Of course a 3% withdrawal rate is safer than a 4% rate, but at what point do you feel secure enough to retire? A "black swan" event could wipe any of us out tomorrow...

I'd also like to point out that FIREcalc, the tool that gives the proverbial "4% SWR", is not a Monte Carlo simulation, but instead a tool based on historical time-series data. Monte Carlo sim doesn't include the time series correlation from year to year. As a result, Monte Carlo produces worse estimates of success. In the FIREcalc data set (going back to 1871??), a series of bad years are typically followed by a series of good years. In Monte Carlo, you could have a long series of years that have below average returns.

[MasterBlaster]

I think a 4% withdrawal rate going forward is suspect.

The problem is for those retiring in their late 30's early 40's and relying on a 4% withdrawal rate.

In SWR101 class we learned that the 4% SWR rate is for a 30 year term. If you are 30-something then you may need to plan for a 50-60-70 year retirement. For those retirement durations the 4 % SWR is not valid.

For a retirement that long, at best you could have a SWR in the low 3 percent range. Bernsteins chart (for a stock-bond portfolio) is shown below. Other studies showed similar results.

[CaseInPoint]

Originally Posted by haha

The most vocal ER cheerleaders have COLA pensions, usually taxpayer supplied. That is an cheap viewpoint for them to hold.

My own POV is that a reasonably good job that isn't tearing you up is better than retirement, at least until you are old. I was not a contented worker, I admit it and have no apologies for that. But I also admit that very early ER is an exercise in thin-ice skating, unless one is quite well fixed and has no taste for heavy spending. There are plenty people here who also fit that description.

But people ride motorcycles, climb mountains, sleep with tarts- why not retire early if you want to? Not everyone is extremely risk averse, and the modern job environment can really suck.

ha

I completely agree, and I hope that people who come to this forum to research the possibility of ER will get the chance to read your posts.

On the point of the taxpayer-paid pensions, I didn't realize how common that is until I joined this forum. Maybe I was naive. Still, I can't help but think that these programs are yet another form of welfare and forced wealth redistribution. No wonder our government operates at such a deficit, and no wonder anyone in the private sector has to work so many extra years just to pay in taxes that go to support government pensions.

Originally Posted by MasterBlaster

You just may be right about the coming apocalyspse. If we move into a few decades of no growth or very slow growth then we are in trouble. Ask me in a few decades and I'll let you know.

You might be confusing me with someone else. I don't believe in any coming apocalypse. On the contrary, I think that the market will rebound within the next 2 years, maybe even sooner barring another 9/11 scenario, and depending on how Europe's economic problems affect us.

As for the ability to safely withdraw 4%, that might be affected by when, during one's retirement, the market downturn happens.

If the downturn happens at the beginning of ER, it might have more of an impact, because the portfolio's income moving forward will be based on a reduced nest-egg. On the other hand, if the market downturn happens towards the middle or end of the ER period, fewer years are affected.

[Canadian Grunt]

There is no change to my assertion that statistical data from a Monte Carlo simulation or Firecalc spreadsheet is accurate over a 40-50 year timeframe. The data is still insufficient for all the reasons I described in my last post. It's just a guess based on past performance. Neither Firecalc, nor any other statistical model can project a 40-50 year time horizon due to data skew and a constant change in the financial regulations affecting the raw data. The longer the horizon the greater the risk. Predicting a 95% success ratio off Firecalc then retiring at 40 with a possible time horizon of 50 years is a great way to go broke between 70-90 years old.

Unless you have a guaranteed pension, bailing in your late 30’s or early 40’s may inhibit your ability to ER in your 50’s. Most people in their 30’s and 40’s underestimate the difficulty of finding a good job as they age once the connection to the work force has been severed. Age bias is still alive. The good job you may abandon at 40 probably won’t be there if you are forced back to work for financial reasons at 50.

I would also like to state again that survivor bias is present in the raw dats whether you use Firecalc or a Monte Carlo simulation. Your returns will be lower than those projected by the simulation.

[Canadian Grunt]

Originally Posted by MasterBlaster

In SWR101 class we learned that the 4% SWR rate is for a 30 year term. If you are 30-something then you may need to plan for a 50-60-70 year retirement. For those retirement durations the 4 % SWR is not valid.

For a retirement that long, at best you could have a SWR in the low 3 percent range. Bernsteins chart (for a stock-bond portfolio) is shown below. Other studies showed similar results.

By the way, Bernstein is on the record as stating that early retirement (30's and 40's) is a good way to be destitute in later years for a variety of reasons that are being covered. Even a 3% withdrawal rate may deplete all the capital over a 40-50 year horizon.

[FUEGO]

Originally Posted by Canadian Grunt

I would also like to state again that survivor bias is present in the raw dats whether you use Firecalc or a Monte Carlo simulation. Your returns will be lower than those projected by the simulation.

Can you clarify what exactly you mean by survivor bias? Do you mean the S&P 500 or DJIA or average US treasury returns are overly optimistic because they continue to exist whereas other indexes or treasuries have ceased to exist (pre-ww2 germany maybe?)?

If so, you may have a valid point, by the way.

[Abreutime]

Originally Posted by Canadian Grunt

Still here, just letting some discussion flow.

Let me qualify my statement. I think a 4% withdrawal rate going forward is suspect. There are some smart people on this board so I hope they will pick up where my inadequate delivery leaves off.

The problem is for those retiring in their late 30's early 40's and relying on a 4% withdrawal rate. If the market goes sideways for an extended period of time or we get stagflation there will be problems for them moving forward if they put too much trust in Monte Carlo predictions.

Monte Carlo simulations are based on statistical returns over the last 100 years. Going beyond 100 years is a fool’s folly as there were not enough stocks in the market to get an accurate projection. There is not enough variance in the date after 100 years to even bother making a statistical model.

Monte Carlo predictions are based on market statistics broken into monthly/weekly/and daily (depending on the model) slots to provide enough variables. Statistically, there is not enough “usable” data. As an investor you need 1 yr, 5yr, and 10 yr distributions to make an accurate model to project distributions into the future. What I am trying to say is, that a Monte Carlo prediction would not merit statistically. It is a financial tool to make estimates.

1 yr distributions only give 100, 1 yr data points (nothing statistically unless you are flipping coins), 5 and 10 year distributions are slightly less. It is statistically wrong to shift the data points by one day to get a second set of variables as returns are historically measured on a yearly basis not a daily basis. Statistically, there is not enough variance to predict future performance. Add on the fact that the stock market has fundamentally changed from 1929, and 1987, due to regulation and market controls the case for Monte Carlo predictions breaks down.

Realistically, Monte Carlo simulation is just a prediction used to aid people who are going into normal retirement and have shorter time horizons. The shorter life spans aid to reduce variance in the models by smoothing the distributions.

As another point, Monte Carlo predictions do not take into account survivor bias. You can guarantee that returns were lower than predicted due to this variance alone.

I think you're mistaking the purpose of using simulations to project outcomes.
Monte Carlo refers to the technique of using a range of possible inputs, randomly selecting data from within this range, and displaying output. This is done thousands of times to simulate a range of outcomes. N

1) The output is only as good as it's input
2) The output is generally given as a range of scenarios of unequal likelihood.
3) Any survivor bias is the result of faulty data. So if you are looking at historical stock returns, make sure the data includes the bankrupt firms. If you improperly use Morningstar databases, then you will have survivorship bias. But most broad historical databases of returns do not.


To your specific points:

Statistically, there is not enough variance to predict future performance.

Give me 1 data point, and I can make a future prediction. With each extra data point, I think I can make a better prediction. Do you disagree?

The shorter life spans aid to reduce variance in the models by smoothing the distributions.

There is no artificial smoothing of distributions. Short time horizons have a smaller range of likely outcomes.

There is not enough variance in the date after 100 years to even bother making a statistical model.

How much is enough for you?

What I am trying to say is, that a Monte Carlo prediction would not merit statistically.

This statement is so vague to be devoid of meaning. It's not predicting anything; it's projecting a series of outcomes based on the inputs you choose. If you think 100 years of returns are worthless, you can use 200. If you think the US since 1987 has fundamentally shifted, use this data. Heck, use monthly data if you want. The projections will reflect whatever data you use. If you don't like the outputs, it's not the fault of a statistical tool/method. Get better data.

It seems like you think the US is entering a brand new paradigm. Is this what you believe? Guess what, off the top of my head, the same thing was said in the 1920s (sustained and permanent prosperity), the 1930s (long Depression with no end in sight), the 1960s (economic expansions can be sustained as long as policy does not interfere), and 1970s (the death of equities, stagflation), 1980s (deficits dont matter), 1990s (The New Economy), too.

[Canadian Grunt]

Originally Posted by FUEGO

Can you clarify what exactly you mean by survivor bias? Do you mean the S&P 500 or DJIA or average US treasury returns are overly optimistic because they continue to exist whereas other indexes or treasuries have ceased to exist (pre-ww2 germany maybe?)?

If so, you may have a valid point, by the way.

Sure, survivor bias falls into two categories.

First, stocks: Can you list how many stocks are still around from 1901?

Most stocks enter a growth phase and then move through maturity to decline. Somewhere through the spectrum they are either acquired, spun off, or go bankrupt, causing a loss in value and a false elevation of actual market returns.

Secondly, Mutual funds, hedge funds, and ETF's that do not perform are closed down, thus removing their poor performance from the records.

All financial models fail to account for this discrepancy between actual market returns and reported market returns. Reported market returns are always higher than the actual market for this reason.


There is also additional evidence that reported market returns by funds and stocks over extended periods are almost never match the investors returns as most funds and stocks are rotated by investors (even if you only adjust the portfolio to match your asset mix). If you miss the best 10-30 days you miss the fund return.

[Canadian Grunt]

Originally Posted by Abreutime

I think you're mistaking the purpose of using simulations to project outcomes.
Monte Carlo refers to the technique of using a range of possible inputs, randomly selecting data from within this range, and displaying output. This is done thousands of times to simulate a range of outcomes. N

1) The output is only as good as it's input
2) The output is generally given as a range of scenarios of unequal likelihood.
3) Any survivor bias is the result of faulty data. So if you are looking at historical stock returns, make sure the data includes the bankrupt firms. If you improperly use Morningstar databases, then you will have survivorship bias. But most broad historical databases of returns do not.



To your specific points:
Give me 1 data point, and I can make a future prediction. With each extra data point, I think I can make a better prediction. Do you disagree?

There is no artificial smoothing of distributions. Short time horizons have a smaller range of likely outcomes.

How much is enough for you?
This statement is so vague to be devoid of meaning. It's not predicting anything; it's projecting a series of outcomes based on the inputs you choose. If you think 100 years of returns are worthless, you can use 200. If you think the US since 1987 has fundamentally shifted, use this data. Heck, use monthly data if you want. The projections will reflect whatever data you use. If you don't like the outputs, it's not the fault of a statistical tool/method. Get better data.

It seems like you think the US is entering a brand new paradigm. Is this what you believe? Guess what, off the top of my head, the same thing was said in the 1920s (sustained and permanent prosperity), the 1930s (long Depression with no end in sight), the 1960s (economic expansions can be sustained as long as policy does not interfere), and 1970s (the death of equities, stagflation), 1980s (deficits dont matter), 1990s (The New Economy), too.

What I am trying to say is the statistical data imputed by the model accounts for a series of data points from A-Z and AB-YZ and on and on, where realistic data points should start from a specific reporting point giving only a series of 100 data points. Do you report your yearly returns daily or at a specific point in the calendar such as the end of the year. Statistical data that reports a one year return that is calculated using daily data is skewed. It creates artificial data points.

As for condensation of time variance, the shorter the span the more likely the historical average variance will match the time horizon. As the time variance increases so do the standard deviations. Most statistical variations plot as a bell curve. So the shorter the curve the more likely the statistical data will match given a reasonable expectation of an outcome

[Abreutime]

Originally Posted by Canadian Grunt

What I am trying to say is the statistical data imputed by the model accounts for a series of data points from A-Z and AB-YZ and on and on, where realistic data points should start from a specific reporting point giving only a series of 100 data points. Do you report your yearly returns daily or at a specific point in the calendar such as the end of the year. Statistical data that reports a one year return that is calculated using daily data is skewed. It creates artificial data points.

Do you only make investment decisions on 1 day of the year, or do you make them daily? I don't understand how this is an artificial data point. What is artificial about looking at possible investment returns from March 1, 2006? That seems pretty real to me, and relevant, too.

As for condensation of time variance, the shorter the span the more likely the historical average variance will match the time horizon. As the time variance increases so do the standard deviations. Most statistical variations plot as a bell curve. So the shorter the curve the more likely the statistical data will match given a reasonable expectation of an outcome

I'm having trouble understanding what you're saying here. What do you mean by "time variance," "shorter curve," and data will "match?" I think it's in reference to your earlier statement:

The shorter life spans aid to reduce variance in the models by smoothing the distributions.

[OAG]

Originally Posted by CaseInPoint

I didn't realize how common that is until I joined this forum. Maybe I was naive. Still, I can't help but think that these programs are yet another form of welfare and forced wealth redistribution."..."just to pay in taxes that go to support government pensions.

I get one of these. BTW would you like to "trade" about 7 years of your life to spend that amount of time in the Far East, Middle East, and SE Asia (some of which you could hear and feel stuff blowing up around you)? We all make our choices and some of us "earned" those "pensions". And I agree with your admission of possibly suffering from being a bit naive but you go on and keep paying those taxes for my "welfare and forced wealth redistribution". Have a nice day -- now where did they hide the "ignore" button.

[Canadian Grunt]

Originally Posted by OAG

I get one of these. BTW would you like to "trade" about 7 years of your life to spend that amount of time in the Far East, Middle East, and SE Asia (some of which you could hear and feel stuff blowing up around you)? We all make our choices and some of us "earned" those "pensions". And I agree with your admission of possibly suffering from being a bit naive but you go on and keep paying those taxes for my "welfare and forced wealth redistribution". Have a nice day -- now where did they hide the "ignore" button.

I get one of those as well. My war was the Balkans. And I have a few of those exploding things still in my body

[Abreutime]

Originally Posted by OAG

Have a nice day -- now where did they hide the "ignore" button.

Sick burn.

[Nords]

Originally Posted by Canadian Grunt

Neither Firecalc, nor any other statistical model can project a 40-50 year time horizon due to data skew and a constant change in the financial regulations affecting the raw data. The longer the horizon the greater the risk. Predicting a 95% success ratio off Firecalc then retiring at 40 with a possible time horizon of 50 years is a great way to go broke between 70-90 years old.

(FAQ archive): Bernstein's "Retirement Calculator from Hell" articles

Originally Posted by Canadian Grunt

Unless you have a guaranteed pension, ...

Originally Posted by Canadian Grunt

I get one of those as well. My war was the Balkans. And I have a few of those exploding things still in my body

I don't even have to click on "View Posts" to see that once again the COLA pension recipients get snide comments implying that they've somehow simultaneously cheated the early-retirement system and the taxpayers.

I'm not aware that there are many govt-funded COLA pensions on this board. There's roughly 100 military veterans who are active posters, and the U.S. military's retirement rate is roughly 15%, so it's quite possible that there are fewer than two dozen people posting here with that benefit.

But for those of you whining about COLA pensions, you can go buy your own from Vanguard. Or you could try the alternate approach used by Canadian Grunt and others to earn theirs... although that method has a much different type of "survivor bias".

Or you could continue your whining-- I'm glad that you're feeling frisky with your first-amendment rights. Their free, uninhibited exercise by guys like you make guys like me feel all warm & fuzzy that our service was worth the effort.