Sequence of returns and retirement

 Posts: 318
 Joined: Sun Aug 23, 2009 1:10 pm
Sequence of returns and retirement
It's often said that 1966 was the worst year to retire (or maybe 1929 or 2000, but let's say 1966 for this question). I've set up a spreadsheet that calculates how much a retiree would have had each year with a 4% initial withdrawal rate, starting with Year 1 of retirement being 1966 with each year's investment returns and inflation rates. But since we are trying to account for "bad case scenario going forward," do you have to start Year 2, Year 3, Year 4, etc. of retirement as 1966 again? If Year 1 of retirement actually ended up being a good year for stocks (unlike 1966), should you start Year 2 as being "1966" on the spreadsheet? Or is year 2 supposed to be "1967" (which was a great year for stocks)? I hope I'm making sense.
Last edited by OkieIndexer on Wed Oct 04, 2017 5:23 pm, edited 1 time in total.

 Posts: 2467
 Joined: Sat Aug 11, 2012 8:44 am
Re: Sequence of returns and retirement
I'd consider flexible withdrawal methods such as our Wiki's VPW. "Variable percentage withdrawal (VPW) is a withdrawal method that adapts to the retiree's retirement horizon, asset allocation, and portfolio returns during retirement. It combines the best ideas of the constantdollar, constantpercentage, and 1/N withdrawal methods to allow the retiree to spend most of his portfolio using returnadjusted withdrawals. By adapting withdrawals to market returns, VPW will never prematurely deplete the portfolio."
Bogleheads investment philosophy 
Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflationindexed)bonds 
VCN/VXC/VAB/ZRR

 Posts: 424
 Joined: Sat Feb 21, 2015 2:29 pm
Re: Sequence of returns and retirement
Whether you are making sense depends on what you want to calculate.
If you want to calculate what would have happened with 1966 as the starting year, then year 1 is 1966, year 2 is 1967, year 3 is 1968, and so on. 1966 as the starting year is a bad case scenario not only because of the return and inflation rate of that year, but also for the return and inflation rate of each of the later years using the actual sequence of years.
For a different starting year, say 1970, year 1 is 1970, year 2 is 1971, year 3 is 1972, and so on. The return and inflation rate for 1966 does not enter into the calculation that uses 1970 as the starting year.
If you want to calculate what would have happened for some hypothetical sequence of returns, rather than an actual sequence of returns, then that is a different calculation.
If you want to calculate what would have happened with 1966 as the starting year, then year 1 is 1966, year 2 is 1967, year 3 is 1968, and so on. 1966 as the starting year is a bad case scenario not only because of the return and inflation rate of that year, but also for the return and inflation rate of each of the later years using the actual sequence of years.
For a different starting year, say 1970, year 1 is 1970, year 2 is 1971, year 3 is 1972, and so on. The return and inflation rate for 1966 does not enter into the calculation that uses 1970 as the starting year.
If you want to calculate what would have happened for some hypothetical sequence of returns, rather than an actual sequence of returns, then that is a different calculation.

 Posts: 1958
 Joined: Fri Feb 23, 2007 8:21 pm
Re: Sequence of returns and retirement
This question provides a great opportunity to make a plug for Monte Carlo Simulation. This OP is trying to model a single sequence of returns. A MCS can generate 10,000 alternative histories and provide a sense of both Mean outcome and dispersion of potential outcomes. Also MCS can show the influence of changes in asset allocation and spending on the odds of success.
Dave
Dave
Re: Sequence of returns and retirement
I'm not a spreadsheet person, but a spreadsheet and information to test everything is available here:
https://earlyretirementnow.com/2016/12/ ... t1intro/
The series of articles even discusses Monte Carlo Simulation and using Historical data.
I would not even try to invent my own spreadsheet.
https://earlyretirementnow.com/2016/12/ ... t1intro/
The series of articles even discusses Monte Carlo Simulation and using Historical data.
I would not even try to invent my own spreadsheet.
Re: Sequence of returns and retirement
It sounds like, for what you're trying to do, you would always have "next year" be 1966. Since you're trying to account for "the worst 30 years", that will always be 19661996. So if you retired next year in 2018 and it turned out to be an amazing year, for your spreadsheet you'd still use 1966 for "next year".OkieIndexer wrote: ↑Wed Oct 04, 2017 5:12 pmIt's often said that 1966 was the worst year to retire (or maybe 1929 or 2000, but let's say 1966 for this question). I've set up a spreadsheet that calculates how much a retiree would have had each year with a 4% initial withdrawal rate, starting with Year 1 of retirement being 1966 with each year's investment returns and inflation rates. But since we are trying to account for "bad case scenario going forward," do you have to start Year 2, Year 3, Year 4, etc. of retirement as 1966 again? If Year 1 of retirement actually ended up being a good year for stocks (unlike 1966), should you start Year 2 as being "1966" on the spreadsheet? Or is year 2 supposed to be "1967" (which was a great year for stocks)? I hope I'm making sense.

 Posts: 318
 Joined: Sun Aug 23, 2009 1:10 pm
Re: Sequence of returns and retirement
Yeah, that was my intent, but the VPW spreadsheet posted above seems like a better idea.AlohaJoe wrote: ↑Wed Oct 04, 2017 7:18 pmIt sounds like, for what you're trying to do, you would always have "next year" be 1966. Since you're trying to account for "the worst 30 years", that will always be 19661996. So if you retired next year in 2018 and it turned out to be an amazing year, for your spreadsheet you'd still use 1966 for "next year".OkieIndexer wrote: ↑Wed Oct 04, 2017 5:12 pmIt's often said that 1966 was the worst year to retire (or maybe 1929 or 2000, but let's say 1966 for this question). I've set up a spreadsheet that calculates how much a retiree would have had each year with a 4% initial withdrawal rate, starting with Year 1 of retirement being 1966 with each year's investment returns and inflation rates. But since we are trying to account for "bad case scenario going forward," do you have to start Year 2, Year 3, Year 4, etc. of retirement as 1966 again? If Year 1 of retirement actually ended up being a good year for stocks (unlike 1966), should you start Year 2 as being "1966" on the spreadsheet? Or is year 2 supposed to be "1967" (which was a great year for stocks)? I hope I'm making sense.
Also, if I did Monte Carlo analysis, should you use 19662016 returns and inflation for the inputs (starting with 1966 because that was when the "worst case" in U.S. data started), or use the full 18712016 period?
Re: Sequence of returns and retirement
You can use whatever you want, whatever you think it is appropriate to guesstimate the future. I've seen both approaches used (and many others). Usually people want "the worst case" so pick whichever one looks worst to you (lowest return and highest standard deviation).OkieIndexer wrote: ↑Wed Oct 04, 2017 8:53 pmAlso, if I did Monte Carlo analysis, should you use 19662016 returns and inflation for the inputs (starting with 1966 because that was when the "worst case" in U.S. data started), or use the full 18712016 period?
Re: Sequence of returns and retirement
No, you use 1967 for the next year and so on. So far as the assets available to the program, the output you want is already done in FireCalc, for one example. You can download those results as a spreadsheet. I imagine other retirement planners would give you similar data. To a certain extent you are reinventing the wheel. In case you want to study some very specialized feature you might have to create your own model, but there are a lot out there already. Retirement scenarios are a pretty well plowed field at this point.