Author Topic: Changing asset allocation as FIRE approaches  (Read 9895 times)

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Changing asset allocation as FIRE approaches
« on: September 20, 2016, 12:08:15 AM »
Is there any literature/thought/discussion out there on how someone should adjust their asset allocation as their FIRE date nears?  I'm probably something like a year out from FI, and am currently very heavily loaded toward equities (~90%).  When I run cFireSim calculations, I find that success probabilities go up even as my AA approaches 50/50. 

This agrees with intuition, and there's maybe something to be said for adopting some loss aversion when you think you're a year or so out: I decrease the risk of catastrophic loss near the date, at the expense of upside in the distant future.  So long as that lost upside all represents "winning harder" and not money that I might need for longevity insurance, that seems like a fine tradeoff to be making.  It also sets up nicely for a rising equity glide path, which is hard to do from a low-bond starting point.

So there are a lot of points in favor... but it sure isn't "staying the course" based on where I am now.  So I want to make sure I'm not just rationalizing myself into a big portfolio change that's actually just me getting jittery about markets.

Retire-Canada

  • Walrus Stache
  • *******
  • Posts: 7828
Re: Changing asset allocation as FIRE approaches
« Reply #1 on: September 20, 2016, 07:00:39 AM »
cFIREsim results:

100/0 = 94.8%
90/10 = 96.6%
80/20 = 96.6%
70/30 = 95.7%
60/40 = 94.8%
50/50 = 93.1%
40/60 = 89.7%

Using the default settings at cFIREsim and just adjusting the ratio of stocks to bonds I get this ^^^. You'll have to give us the inputs you are using so to understand the results you are getting.

Retire-Canada

  • Walrus Stache
  • *******
  • Posts: 7828
Re: Changing asset allocation as FIRE approaches
« Reply #2 on: September 20, 2016, 08:02:51 AM »
Again take it or leave it.  I tend to want to problematize shit.

OT - but I figure anything over 90% is so solid that it's not worth worrying about. I was just curious how the OP got results that I couldn't understand based on my knowledge of how cFIREsim works - like higher success at 50/50 than 90/10.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #3 on: September 20, 2016, 10:24:28 AM »
There's some good discussion in this thread on essentially the same topic:  "What is the appropriate time horizon to use for investing prior to FIRE?"

DrF

  • Bristles
  • ***
  • Posts: 464
Re: Changing asset allocation as FIRE approaches
« Reply #4 on: September 20, 2016, 10:36:54 AM »
I studied up on this a fair amount, and sequence of return risk is huge the few years just before reaching FI and the ~5-10 years after FI. I wrote something up in a previous thread. http://forum.mrmoneymustache.com/investor-alley/100-stocks-in-the-accumulation-phase/msg1177820/#msg1177820

Basically, if the market takes a 2-3 year slide in the 2-3 years before you reach FI (say down ~3-5% per year for a max drawdown of 10%) you'd have to work extra if you were 100% stocks. If you switch your AA to something much more conservative, you'd maybe have to work an additional few months or quite possibly no additional.

Example: You have $800k invested in 100% S&P500, and you need $1MM invested to give you $40k per year for FIRE. Let's say you can add $50k per year from job earnings. We'll make it easy and make the market give us equal % returns over 2 years to give us $1MM total (which would be 5.6% per year). Seems feasible right?

Conversely, the market starts to go down as the fed increases lending rates and the global economy doesn't do quite as well with higher borrowing rates. In the first year the S&P500 drops 5%. Now you have $810k invested (= $800k original minus $40k loss plus $50k earnings addition). Then the year after that, the market drops another 5%, which gives you $819.5k. At this point, you might say "screw it, the market has gone down the past 2 years and is ready to go back up so I'll just retire on a ~4.9% WR". Or, more likely, you'll work an additional 1-2 years to get to $1MM.

Alternatively, you could go 50/50 stocks/bonds over the next 2 years and hope to break even (0% returns over the 2 year period). Which would give you $900k invested ($100k addition from earnings over 2 years), which would be a ~4.4% WR. Or, you could work another 6 months (maxing out all tax deferred accounts during this time) get up to ~$950k just with new additions, FIRE, and see where the market goes from there.

The numbers don't lie, if you are near FIRE, it is in your best interest to have your AA be closer to 50/50 than 100/0. If you are already FI, and are still working OMY, then I'd say keep your AA where you have it. Anyone planning a 3% WR with FIRE could easily keep their allocation closer to 100% stocks.
« Last Edit: September 20, 2016, 10:42:13 AM by DrF »

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Re: Changing asset allocation as FIRE approaches
« Reply #5 on: September 20, 2016, 10:37:53 AM »
I was just curious how the OP got results that I couldn't understand based on my knowledge of how cFIREsim works - like higher success at 50/50 than 90/10.

I have it set up to simulate a situation where we downsize out of our high COL home once the kids are out of the house and we aren't benefiting from the local school system.  This unlocks a lot of equity, which turns into investment wealth at that time.  There's also a drop in living expenses around that time, for the same reason.  (I was careful not to double count --when your property taxes run more than most people's rent, both can happen at once. :))

You can think of it as having a big chunk of my 'stache' locked away for ~15 years.  Maybe a third or more, depending on how you count the lifestyle cutbacks.  It means less is there now, esp. relative to the higher immediate drawdown rate, so it's intrinsically more sensitive to early downswings than normal.  All my worst portfolio failures are really early.  That's probably atypical, too.

Thanks for posting that data and asking that question.  It's really putting a finger on what's going on here.  My investment plan was originally built on more typical situations, but modelling my own that way was hard, so a lot of conservatism had to be employed.  Then I find CFireSim, it allows me to model my situation more accurately, and that accuracy shaves a year or two off without even revisiting the AA question.  Not revisiting the AA question earlier was an error.  Since I'm learning about a real error I made in the past, a big adjustment now could make sense.

I've got enough in tax shelters that I could go 50/50 without a tax event or even transaction costs, but I'm still not sure how fast an adjustment is appropriate.  Though I can't think of a persuasive reason against moving quick other than if I move too quick, my analysis might wind up being wrong again, and then I'll be locked out of rebalancing for a while.  I think I'm convinced enough to rebalance into 75/25 immediately.

Thanks again! 

Scandium

  • Handlebar Stache
  • *****
  • Posts: 2320
  • Location: EastCoast
Re: Changing asset allocation as FIRE approaches
« Reply #6 on: September 20, 2016, 10:39:53 AM »
Again take it or leave it.  I tend to want to problematize shit.

OT - but I figure anything over 90% is so solid that it's not worth worrying about.

Is that an argument for sticking with 90% stocks since you're 90% safe? Or an argument for adding bonds since you'd get lower volatility/return but still 90%+ safe?
« Last Edit: September 20, 2016, 11:08:17 AM by Scandium »

Retire-Canada

  • Walrus Stache
  • *******
  • Posts: 7828
Re: Changing asset allocation as FIRE approaches
« Reply #7 on: September 20, 2016, 11:01:13 AM »
OT - but I figure anything over 90% is so solid that it's not worth worrying about.

I that an argument for sticking with 90% stocks since you're 90% safe? Or an argument for adding bonds since you'd get lower volatility/return but still 90%+ safe?

The short answer is I think that there are so many FIRE failure modes not related to how much money is in your portfolio that it's pointless to spend time trying to theoretically optimize the situation in a simulator beyond a certain arbitrary point.

The slightly longer answer is I would suggest you select a mix of stocks and bonds that you can live with and stay the course during market run ups and crashes. I've seen compelling arguments made for both a high equity ratio and a more balanced approach. In the heat of the battle without the benefit of hindsight you will have to trust your asset allocation and stick with your plan. Having a plan you believe in and trust is more important than the having the most optimized AA [by some objective measure] unless the two happen to be the same AA.

MustacheAndaHalf

  • Magnum Stache
  • ******
  • Posts: 2942
Re: Changing asset allocation as FIRE approaches
« Reply #8 on: September 20, 2016, 11:05:31 AM »
... we downsize out of our high COL home once the kids are out of the house and we aren't benefiting from the local school system.  This unlocks a lot of equity, which turns into investment wealth at that time.
...
You can think of it as having a big chunk of my 'stache' locked away for ~15 years.  Maybe a third or more, depending on how you count the lifestyle cutbacks.
You could also view that as 1/3rd of your retirement being at risk to the local real estate market.  It might be safer to plan on your house value taking a correction, so that maybe you work a little longer to ensure having enough.  And then if there's no correction, you can revise plans.

Take a peek at different companies "target retirement" date funds, too.  You can see how Vanguard or Fidelity glides towards the retirement allocation that way.  Just another source of information.

DrF

  • Bristles
  • ***
  • Posts: 464
Re: Changing asset allocation as FIRE approaches
« Reply #9 on: September 20, 2016, 11:11:18 AM »
Having a 90-99.99% success rate with 90-100% stocks means that there is a low likelyhood (.01-10%) of catastrophic failure. Having a 90-99.99% success rate with 40-60% stocks means that there is a low likelyhood (.01-10%) of a gradual failure.

I can think of lot's of ways to modify my retirement to prevent gradual failure. Less number of ways to prevent catastrophic failure.

DrF

  • Bristles
  • ***
  • Posts: 464
Re: Changing asset allocation as FIRE approaches
« Reply #10 on: September 20, 2016, 12:33:35 PM »
Perhaps.  But is that still true in a world where the yield on many stocks is higher than the yield on most bonds?  And with higher duration risk?  Or maybe the question is whether your WR is higher or lower than the blended yield on your bonds?  My only point is that the relationships between and among asset classes change in different economic environments.  Was the 70's a gradual or catastrophic failure for nominal bondholders?

The 70's were irrational. The fed exercised poor monetary policy that couldn't control inflation. http://www.federalreserve.gov/pubs/ifdp/2004/799/ifdp799.htm
Bond yields went crazy high, crushing the underlying nominal value of the bond. But, if you were to hold the bond to duration - you would have received a considerably great coupon yield compared to today.
Even if you had kept your money in CDs, you would have done quite well. http://mortgage-x.com/general/indexes/codi_history.asp
The chart on that link is the average annualized yield based on the trailing 12 months for 3mo CDs. Basically, you would have realized a minimum of ~4.5% APY and maximum of ~16.69% APY if you had only invested in CDs from 1967 through year end 1990. What a sweet time compared to current yields. http://www.bankrate.com/finance/cd-rates-history-0112.aspx

Also, bonds tend to not be correlated with the movement direction of stocks. http://awealthofcommonsense.com/2015/03/the-blueprint-for-a-bond-bear-market/
Going forward, if you think of bonds as increasing yield rather than price appreciation, you'll probably be OK.

DrF

  • Bristles
  • ***
  • Posts: 464
Re: Changing asset allocation as FIRE approaches
« Reply #11 on: September 20, 2016, 02:53:34 PM »
My investment strategy would be considered "alternative" here in the forums, but yes if you are holding treasuries I would suggest they be all short term. Long term treasuries seem to be in a near bubble state (last one holding them gets the rotten egg). During bubble times it seems like people always say "Shucks, I'll just sell 'em when they start going down...some other sucker will take the loss". I've been hearing things along those lines recently. I recommend ~50% of your bond holdings be short term. I like BSV, mix of short term government and corporate. Then I'd allocate the other ~50% of my bond holdings to long corporate, something like VCLT, or LQD. One could also throw in some mortgage backed securities (VMBS) and change their bond AA around a bit to taste.

You don't hold bonds for growth, you hold bonds for low volatility and low correlation with stock price movement.

Here's another nice write up showing intermediate bonds have less volatility and near equal total return since 1976. In today's market, I'd go with even shorter duration bonds for the foreseeable future.
http://awealthofcommonsense.com/2015/05/are-long-term-bonds-worth-the-risk/

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Re: Changing asset allocation as FIRE approaches
« Reply #12 on: September 20, 2016, 08:35:58 PM »
Thanks for the link, Brooklynguy.  Found exactly what I was originally looking for over there:
  https://www.kitces.com/blog/retirement-date-risk-how-sequence-of-returns-risk-impacts-a-pre-retirement-accumulator/

I like the symmetry between those arguments and the rising equity glidepath.

Agreed that bonds right now are best kept short term.  Could probably even simulate it in cFiresim using a mix of bonds and cash.

Anyway, I have a some analysis to do and trades to queue up.  Thanks for all the commentary.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #13 on: September 21, 2016, 08:39:30 AM »
I like the symmetry between those arguments and the rising equity glidepath.

Yeah, though as I said in the other thread, I'm not a big fan of either of those approaches because they're not really mitigating sequence of return risk so much as just shifting it around.  It might make more sense in your atypical case (with the big chunk of locked away stash), but then again you're also intrinsically more sensitive than usual to late downswings (that coincide with the deployment of those locked away funds into the market).

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Re: Changing asset allocation as FIRE approaches
« Reply #14 on: September 21, 2016, 11:14:06 PM »
I like the symmetry between those arguments and the rising equity glidepath.

Yeah, though as I said in the other thread, I'm not a big fan of either of those approaches because they're not really mitigating sequence of return risk so much as just shifting it around.  It might make more sense in your atypical case (with the big chunk of locked away stash), but then again you're also intrinsically more sensitive than usual to late downswings (that coincide with the deployment of those locked away funds into the market).

I think I disagree with that characterization.

Part of what I like about glidepaths is that they patch up some unsatisfying properties of standard modern portfolio theory.  Like how it calls for bonds even over extremely long periods of time.  Over sufficiently long periods of time, stocks are always just better, right?

Yet MPT approaches have some decent arguments (and results) that suggest that over some time horizons, bonds are good in at least some way.  (Think: efficient frontiers, which have modest but real effects.)  The explanation is in how rebalancing causes you to buy-low/sell-high, but it looks like a an inefficient way of leveraging the anticorrelation between asset classes.  You're never buying-low or selling-high for very much of your portfolio at once.

Glidepaths reconcile all of this.  You're using bonds for risk mitigation, but only over the short-medium periods of time when stocks aren't very good.  The function of your bonds is simply to make sure your stocks have enough time in market to hit critical mass.  This is a really targeted/specialized use of both asset classes, right?  You're letting both of them shine over the time periods they're best at, while relieving them of the time horizons over which they underperform.

I wish CFireSim supported glidepaths.  I'd bet a nickel that you could parameterize them in a way that always outperforms the classical approach.

Goldielocks

  • Walrus Stache
  • *******
  • Posts: 6821
  • Location: BC
Re: Changing asset allocation as FIRE approaches
« Reply #15 on: September 21, 2016, 11:57:39 PM »
Approaching FIRE, I would look at having 3-10 years in a conservative portfolio, and the rest as equities or ??

Ideally a 10 year bond ladder would be great.   The market can be flat or down (against inflation) for up to 10 years, so that would ensure I never withdraw when it is down...   

In reality, it would be 10 year bond ladder for only half my spend, with the plan to cut back if I needed on spending.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #16 on: September 22, 2016, 08:36:02 AM »
I wish CFireSim supported glidepaths.  I'd bet a nickel that you could parameterize them in a way that always outperforms the classical approach.

Maybe.  It would've been nice if the Pfau/Kitces research on rising equity glidepaths included backtests using actual historical data.

What worries me about scaling back the initial equity allocation in the context of super-long retirement periods is the opposite of what you described in the OP:  that it might come at the cost of foregoing potentially necessary longevity insurance rather than unnecessary upside.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #17 on: September 22, 2016, 09:34:46 AM »
Can't you approximate a glide in cFIREsim by including a one time change in AA at year 10? 

Probably not a great approximation, because it doesn't fully capture the effect of dollar cost averaging into the stock market at cheaper and cheaper valuations in a poor early returns sequencing scenario.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #18 on: September 22, 2016, 10:22:08 AM »
I agree that it would not be perfect but I wonder about how much breakage there would be.  Maybe I need to understand the strategy better. 

If it's just a bond ladder, then wouldn't the ladder be used for living expenses with no rebalancing, so you wouldn't be getting bang for buck in that way?  If that's the strat you could use the average bond position coded as cash and the average locked in rate in the first leg.

If on the other hand the glidepath includes rebalancing (not a bond ladder) wouldn't cFIREsim capture that by rebalancing to one AA in the first leg and a different AA in the second leg?  It would just be more of a jump than a glide in terms of the change... 

The original primary research on the rising equity glidepath strategy is available here, and Kitces did a good summary and analysis here.  In that research, they used a straight line allocation adjustment over the entirety of the retirement period, so a "jump" doesn't simulate their proposed "glide" very well.

However, I just re-read the cFIREsim FAQ, and this is what it has to say about the "Keep Allocation Constant" option:

Quote from: cFIREsim FAQ
Keep Allocation Constant: Choosing "Yes" means that your asset allocation will remain the same throughout the simulation. Choosing "No", you can set what is commonly called a Glide Path allocation change. You will choose a target allocation that is different from your current one (Ex. You want to go from 80%/20% stocks/bonds to 60%/40% stocks and bonds in retirement). You will then choose a "Start Year" and "End Year" to execute this allocation change. Your allocation will slowly change to the target allocation over the course of those years.

So it looks like it is possible to test glidepath strategies using cFIREsim after all.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #19 on: September 23, 2016, 11:34:30 AM »
coolio, so does someone want to run a horse race?

I figured msilenus would do so, since he expressed a wish for cFIREsim to have glidepath-testing capability (and a willingness to stake a nickel on the outcome).

I tried a few test runs; based on the wonky results, it seems that cFIREsim may not be correctly executing the glidepath feature -- for example, the slightly rising equity glidepath that shifts from 75/25 stocks/bonds to 80/20 stocks/bonds over the full 30 year period produces a significantly lower success rate than a static allocation that corresponds to either of those endpoints, which seems to make no sense.

But the old cFIREsim site produces different results, which are more in line with expectations.  I didn't spend too much time tesing, but I couldn't find any glides that improve the results over static stock-heavy allocations.

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Re: Changing asset allocation as FIRE approaches
« Reply #20 on: September 23, 2016, 07:04:38 PM »
I'll play around with it when I get a moment.  I just haven't had one of those the last few days.  Part of what I'm realizing is that you have to be careful about how you define what you're looking for.  Ie: What do you compare a 60/40 ->100/0 5-year glidepath to, in terms of an MPT?  80/20?  90/10?

My thought right now is that we have to go the other way: consider an 80/20 or 75/25 MPT.  What sorts of glidepaths outperform this on all of: { the median ending balance, lowest 25%, lowest dip, success rate }?  (If any.)  The reasoning behind the focus on the bottom half is that anything above the median is just winning harder and running up the score.  Even the median might be misguided, but I figure that if you get that too, then that says something pretty strong.

Another option is to pick a target success rate and see what kinds of glidepaths+SWRs and what kind of MPT AAs+SWRs deliver it.  Ie: if you can get a 4.5% SWR with a certain glidepath configuration, but the highest you can get is 4% with MPT (across all asset allocations for both), then glidepath wins.  (I don't like this idea because it smells super over-fitted, but I'm tossing it out there.)

Thanks for finding the options!

Cottonswab

  • Stubble
  • **
  • Posts: 177
  • Age: 34
  • Location: Boulder, CO
  • Occasional Advice Dispensary
    • My Journal
Re: Changing asset allocation as FIRE approaches
« Reply #21 on: September 24, 2016, 12:56:22 AM »
There is no ideal asset allocation for all times.  The expected return on investment for stocks and bonds is always changing.  The current interest rates for bonds are at historic lows, with negative real returns expected. 

For those looking to retire soon with high equity allocations, I would recommend projecting your expenses (planned and unexpected) for at least the next 5 years.  Project the passive income that you will receive.  Calculate the negative cash flow.  Convert enough stocks to short-term bonds to cover the negative cash flow.  Time the conversions to minimize taxes on capital gains.

As for the rest, I would calculate the expected ROI, based on current market conditions and then go for the highest ROI investments with "adequate" diversification of risk.  Don't use CFIREsim as the basis for your asset allocation. 

TomTX

  • Magnum Stache
  • ******
  • Posts: 4202
  • Location: Texas
Re: Changing asset allocation as FIRE approaches
« Reply #22 on: September 25, 2016, 05:06:22 AM »
  So I get a bit obsessed with vintage 1966 (one of the key failures in those cFIREsim runs) and ask myself what happened and how I would insulate.  That 1966 vintage is the stangflation example and would have been helped a lot by a little gold (e.g. by reallocating 5% of the portfolio from bonds to gold).  That's really interesting to me.

Gold in 1966 isn't worth considering because US citizens could not own gold bullion until 1974

Gold prices were artificially suppressed.

Globally, gold was about the cheapest it had been in 800 years, other than being briefly lower during/after WWI.

http://www.zerohedge.com/news/charting-price-gold-all-way-back-1265

Sure, if you buy a commodity at or near its 800 year low, it can boost returns.

mathjak107

  • Pencil Stache
  • ****
  • Posts: 558
Re: Changing asset allocation as FIRE approaches
« Reply #23 on: September 25, 2016, 05:34:28 AM »
if you bought an ounce of gold in 1975 ,the first year citizens could own gold here that was not a collectible  that same money put in a 1 month t-bill and rolled over is worth more today .

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Re: Changing asset allocation as FIRE approaches
« Reply #24 on: September 25, 2016, 07:19:33 PM »
I didn't spend too much time tesing, but I couldn't find any glides that improve the results over static stock-heavy allocations.

I did some playing around with this.  70/30 -> 100/0 over 10 years was the best glidepath I've found so far for what I imagine might be a "typical" situation.  Care to check me?

The intuition I'm developing is that you're going to give up median end portfolio with either glidepaths, or by increasing bond allocations.  Both of those strategies improve worst case performance, but the 70/30,10 glidepath seemed to be striking a better tradeoff than I could find via statically increasing bonds.  Ie: I could get better security at a higher median than I could with static allocations, but the median wasn't as good as with 100/0 static.

Caveat: I'm a little worried I'm overfitting.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #25 on: September 26, 2016, 07:15:02 AM »
I did some playing around with this.  70/30 -> 100/0 over 10 years was the best glidepath I've found so far for what I imagine might be a "typical" situation.  Care to check me?

I agree that this looks like a pretty attractive package of trade-offs over 100/0 static -- using all default cFIREsim parameter settings other than the specified AA and glidepath, it gives a meaningful (albeit not huge) improvement in downside performance for a relatively small sacrifice in median ending balance.  As long as you're otherwise inclined to stick with a stock-heavy AA, I wouldn't worry about overfitting -- this is just tweaking at the margins, shifting from one reasonable AA along the stock-heavy spectrum to another.

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Re: Changing asset allocation as FIRE approaches
« Reply #26 on: September 26, 2016, 09:50:46 AM »
Lazarus, I was mostly using the cFireSim defaults, except for AA.  So success rate tended to hover in the 95% range.  After I found that parameter set, I plugged it into my own (complex, as noted earlier) simulation, and found that it was about as good for as boosting my portfolio by 5% in terms of impact on success rate, relative to the 90/10 static I had before.  It was actually enough to bump my model up to 100% SR.  I should note that a static 70/30 was also enough to put it over 100%, but the static 70/30 had a much weaker median end-portfolio score.

Not gonna quit immediately for a slew of reasons, but a pretty cool result nonetheless.

When I get a moment (again) I'll see how it performs around a fixed 100% confidence point.  My intuition is that it should be the same because when I was doing the unfixed simulations, SR wound up hovering within a narrow range.  As Brooklynguy noted, the apparent benefit was meaningful, but small.

Brooklynguy, thanks!

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Re: Changing asset allocation as FIRE approaches
« Reply #27 on: September 27, 2016, 01:17:10 AM »
Got it.  Nice.  My worry with the defaults is that this uses 30 years as the total horizon.  So I can see the glidepath making a difference because you glide for the first third and then ride out the last two thirds.  In a 50 year horizon on the other hand drag factors hurt more (deductively) and initial glidepaths probably have to be longer to make an overall difference (intuitively).  So I worry a 50 year horizon plays differently than a 30.  Now the counter to that might be that some people think trying to predict between the 31st and 50th year is a bit of a trick in any case, but still..

The big problem with that kind of time horizon in in cFireSim is that it limits the dataset.  The latest 50 year window available starts in 1965 or 1966, so a lot of the stagflation years aren't in the simulation.  The longest horizon I use is the one that takes me out to max SS retirement age.  From there, the spectrum of end portfolios is what's interesting, keeping in mind that I get to add full SS benefits into that.

mathjak107

  • Pencil Stache
  • ****
  • Posts: 558
Re: Changing asset allocation as FIRE approaches
« Reply #28 on: September 27, 2016, 02:48:06 AM »
very very long time frames can be a problem with many of the calculators because the poster child of worst case scenario's is 1966 .  eliminate that starting year from most simulations  and you no longer are comparing apples to apples since that was the worst case scenario

mathjak107

  • Pencil Stache
  • ****
  • Posts: 558
Re: Changing asset allocation as FIRE approaches
« Reply #29 on: September 27, 2016, 07:32:36 AM »
yep ,  catching the time frame with out the effect on other time frames makes the data set very weak . all time frames are influenced by other time frames criss crossing it  at various points .

why this can be important is  if we take a time frame like 1982  when the greatest bull in history happened the reality was the time frame leading up sucked for most of us .

flat markets ,high inflation , no 401k's  and a poor economy had most of us unable to save much at all .

so all well and good the greatest bull market was starting but few of us had any money saved to take advantage . most of us were first starting from dollar 1 .

so it is important to get a good sampling of all kinds of other time frames interacting .
« Last Edit: September 27, 2016, 07:37:42 AM by mathjak107 »

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #30 on: September 27, 2016, 07:49:09 AM »
Given the mission (retire for 50 years), there just isn't enough data in cFIREsim to rely on it as an exclusive tool for testing portfolios.

There isn't enough data on 50+ year retirement periods anywhere, because there are just so few of them in the historical record (not to mention the staleness of those that do exist, the last of which started 50 years ago!).  One useful workaround is to extrapolate from shorter periods -- I'm a big fan of the analysis Go Curry Cracker did in his "Endowment Fund" post.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 40
Re: Changing asset allocation as FIRE approaches
« Reply #31 on: October 05, 2016, 08:18:32 AM »
Kitces put out another article on this topic today--"The Portfolio Size Effect and Using a Bond Tent to Navigate the Retirement Danger Zone"--arguing that the optimal glidepath for managing sequence of returns risk is a "V-shaped" equity allocation, dialing down equity exposure in the years leading up to retirement and in the early years following retirement, before returning to normal.

msilenus

  • Pencil Stache
  • ****
  • Posts: 517
Re: Changing asset allocation as FIRE approaches
« Reply #32 on: October 05, 2016, 11:50:24 AM »
Kitces put out another article on this topic today--"The Portfolio Size Effect and Using a Bond Tent to Navigate the Retirement Danger Zone"--arguing that the optimal glidepath for managing sequence of returns risk is a "V-shaped" equity allocation, dialing down equity exposure in the years leading up to retirement and in the early years following retirement, before returning to normal.

Yeah, it was kind of inevitable that something like this pop up.  It's the obvious synthesis of his two earlier posts.

I'm surprised at the parameterizations he's illustrating in that article.  He mentions that a lot of work needs to do on modelling and describing the space to find the optimal configurations, but even so, it's interesting that his straw values are so much different from what we seem to be finding.