Withdrawal Rates at 50-50 Stocks/Bonds

[samclem]

But, I am wondering if firecalc somehow factors in the cost of switching fixed income products if interest rates rise?

Obviously FIRECalc doesn't micro-model each asset class for various investor behaviors, but you don't need that anyway. It's taken care of by the way the data behaves if we assume periodic rebalancing (you're going to do that, not try to time the market, right?). FIRECalc assumes you'll do that rebalancing annually. You've described correctly how bond prices respond to interest rate changes. Think about how that plays out with annual rebalancing in response to their value/returns and you'll see that the buying/selling/value changes are already baked into the data FIRECalc is using.

 

[rayvt]

Okay, download this spreadsheet https://www.dropbox.com/s/cbzvg74iyeyfwt6/SPX-monthly-1950-2013-and-IUL-test.xls

It was built for a different purpose but will perhaps tell you what you want to know. It's not Firecalc, but it will tell you some things that Firecalc doesn't. It has data for S&P500 from 1950 to present.
For your exploration, ignore the 10mSMA and IUL sections, only look at the B&H section.

You have not mentioned what is probably the most important thing. The *average* return does not matter. What matters is the *worst* return. An average return of 8.5% doesn't help you if you hit a string of -5% years and run out of money before a string of 15% years brings the average back up.

The worst 50 year return for S&P had CAGR of 8.5% and began 4/1/59.

The worst 30 year return for S&P had CAGR of 8.9% and began 11/1/55.

I plugged 4/1/59 into the start & withdraw dates, $100,000 to the initial value, withdraw of $333/mo (4% of $100K), 3% for withdrawal inflation rate, weightings 5 & 5.

Go down to cell R789 and put in 53, for a end date of Mar 2012.
Final value: $1,646,157. Total withdrawn: -$505,388.
The lowest value: $97,995. So you're fine.

Pick a start date just before a bad bear market, like 5/1/69
Final value: $1,163,648.
The lowest value: $82,744. So you're fine. Although it may have been nail-biting time.

Unlike firecalc, it doesn't show you every portfolio start date, it only shows what the portfolio did for one starting date.

However, the "rolling N yr" sheet shows the best/worst/avg/med for every N year period (actually N * 12 month period) you care to examine.
----

Virtually certain? If 4% is the 30 year SWR, the indefinite duration SWR is closer to 3% based on every academic paper I've seen. That's a significant difference...

Yes, and moreover those papers have data going back to the 1800's or earlier. I'm not sure how applicable the 1850 investment market statistics are to a 2013 portfolio. But, whatever, the data I have for S&P since 1950 indicates that "virtually certain" is arguably right.
The rolling N-yr returns the above spreadsheet shows for a 60/40 are:

[FONT=Courier New]Yrs --> 1       5       10      15      20      25      30[/FONT]
[FONT=Courier New]Avg     9.2%    9.1%    9.2%    9.5%    9.6%    9.8%    9.9%[/FONT]
[FONT=Courier New]Med     10.3%   9.1%    8.9%    9.2%    9.3%    9.6%    9.8%[/FONT]
[FONT=Courier New]Min     -27.1%  -2.2%   0.0%    4.6%    6.3%    7.2%    8.1%[/FONT]
[FONT=Courier New]Max     39.3%   21.9%   15.6%   15.3%   14.6%   14.0%   12.3%[/FONT]

The average return varies hardly at all. The one we care about for portfolio survival, Min, is monotonically increasing.

The most dangerous stage for a retirement portfolio is the early years. There's been a couple of recent papers that talked about this. Just doing naive math in your head:
If you grow at 6.3% and withdraw at 4%, that's a new 2.3% growth. You just have to get past the early years, so that you have enough years that the average is above 4%. The killer is when you've just started out and get hit with 3 years of -8% loss while withdrawing 4%. (And, yeah, that happened. May'00 to May'03.)

 

[sdfire]

Obviously FIRECalc doesn't micro-model each asset class for various investor behaviors, but you don't need that anyway. It's taken care of by the way the data behaves if we assume periodic rebalancing (you're going to do that, not try to time the market, right?). FIRECalc assumes you'll do that rebalancing annually. You've described correctly how bond prices respond to interest rate changes. Think about how that plays out with annual rebalancing in response to their value/returns and you'll see that the buying/selling/value changes are already baked into the data FIRECalc is using.

Right. Thanks. I've got to get into the thought pattern of annual rebalancing when redeeming. I'm used to buying new products with new money to rebalance. But, I'm not used to selling to rebalance. With the expectation of routine stock outperformance compared to fixed income, I could be purchasing new fixed income products every year. I see the point.

This makes me think about the tax consequences of rebalancing. But, I'll think about that for a while and start a new thread with any thoughts/questions.

 

[bmcgonig]

FIRECALC has several spending choices, not just constant dollar/inflation adjusted. Sounds like you want to model 3% of remaining portfolio, you can do that with FIRECALC as well (see pic below). FIRECALC can give you more objective answers than anything a member here can. What's right for you and what's right for me can be considerably different even if we have the same portfolio, spending and retirement duration. There are members here who sleep fine retiring with a 75% success rate, and others who need 200% to pull the trigger. Beyond average returns, are you familiar with the impact sequence of returns has on a portfolio? It's relatively negligible during accumulation, but highly significant once contributions end and distribution/withdrawal begins. I didn't see any acknowledgment in your OP. The fundamental question for retirement withdrawal planning is what actual returns and sequence of returns will be. Assuming 8.5% or any other constant return is guaranteed to be wrong, you can be sure it won't happen like that. Looking at market history, Monte Carlo or some variation of returns will do more to help you understand how much is right for your risk tolerance and time frame.

Something I learned over the weekend from reading Dirk Cottons blog: sequence of returns is not a problem if you are modeling n% of remaining portfolio. Only if ur spending a constant percentage of original portfolio.

 

[Midpack]

Something I learned over the weekend from reading Dirk Cottons blog: sequence of returns is not a problem if you are modeling n% of remaining portfolio. Only if ur spending a constant percentage of original portfolio.

Fair point and true in terms of probability of success. Sequence of returns can lead to failure with constant dollar/inflation adjusted withdrawals. While sequence of returns does not lead to failure with % of remaining portfolio withdrawals (failure is mathematically impossible), it can sure produce dramatic fluctuations in income. Using longer withdrawal periods can smooth that greatly too though, there's a very smart member here who uses 5-year rolling periods.

 

[Henry Lili]

An interesting (and sobering) FIRECALC exercise to run is extending your timeline by increments of 5 years observing the lowest balance each time. Start with 5 years (not that 5 yrs is a real goal) and go five years at a time up to your goal of 50yrs. You’ll see how Mr. Market is a long term play. The shorter time periods look worse than the longer ones given your portfolio has less time to recover from a big drop.

 

[Henry Lili]

Fair point and true in terms of probability of success. Sequence of returns can lead to failure with constant dollar/inflation adjusted withdrawals. While sequence of returns does not lead to failure with % of remaining portfolio withdrawals (failure is mathematically impossible), it can sure produce dramatic fluctuations in income. Using longer withdrawal periods can smooth that greatly too though, there's a very smart member here who uses 5-year rolling periods.

Midpack, I didn't see this note regarding the 5 year rolling periods before I posted my similar note. I wasn't trying to take the credit for the 5 yr rolling approch but it is somthing I have done as well.