Follow-up discussion on AA around one's retirement date

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
Uncorrelated
Posts: 1050
Joined: Sun Oct 13, 2019 3:16 pm

Re: Follow-up discussion on AA around one's retirement date

Post by Uncorrelated »

palanzo wrote: Tue May 12, 2020 10:21 pm
David Jay wrote: Tue May 12, 2020 10:17 pm
Triple digit golfer wrote: Tue May 12, 2020 9:06 pmAlso want to ask, doesn't how money is withdrawn affect the outcome? If one has a 60/40 portfolio and stocks decline and it goes to 55/45, wouldnt the first money be withdrawn from bonds until back to 60/40?
I don’t think where you withdraw the funds matters for SOR.

It comes down to the varying percentage of portfolio spending that the withdrawal represents. 50,000 is 2.6% of 1.9M, but it is only 2.4% of 2.05M (10% gain) and it is nearly 3.3% of 1.52M (20% loss). So one is spending a much higher percentage of their portfolio after a loss.
If one follows Humble Dollar advice and keeps 5 years of spending in cash then a 5 year SOR is less of an issue.
This rule of thumb allocates a larger percentage of your net worth to bonds if equities go down. But that is actually exactly the opposite of what you should do. If markets go up, there is less reason to take risk and you should be increasing the bond allocation (eventually this results in 100% annuitization). If markets go down, that necessitates taking more risk. The following graph shows the asset allocation that maximizes the probability of not running out of money, provided that you reset your asset allocation each year. The color scale is the percentage allocated to equities as opposed to bonds.

Image

I don't recommend this asset allocation because the underlying goal (maximizing the probability of success) is deeply flawed. Read Irlam's work for simulations involving more probable goals.

Advice involving cash cushions is dangerous. It sounds intuitive but will usually do more harm than good when applied in practice.
User avatar
Uncorrelated
Posts: 1050
Joined: Sun Oct 13, 2019 3:16 pm

Re: Follow-up discussion on AA around the retirement years

Post by Uncorrelated »

David Jay wrote: Tue May 12, 2020 7:17 pm
Uncorrelated wrote: Tue May 12, 2020 11:08 amresearch by Gordon Irman has shown that the optimal glidepath for accumulation can be approximated with very simple rules. The asset allocation (stock allocation = 50% * ( 1 + future income / current net worth)) brings you within 5% of the optimal possible asset allocation.
I have had a chance to "play" with this formula this evening. It is interesting, the formula suggests that I should have been over 100% stocks throughout accumulation, as my net worth has never been greater than my future SS income (i.e. future income / current net worth > 1). I don't think I am a unique case, I think a lot of moderate income accumulators who apply this formula will end up with very high equity allocations.

I'm one of those individuals who can "go Spock" and over-ride the emotions created by market swings. I went through the flash crash, the dot com bubble and the financial crisis holding 100% equity. But the vast majority of accumulators are not able to do that.

So what does that say about the real-world applicability of this rule of thumb? It appears one can just discard it and operate on the basis of behavioral finance. That would likely be to select the highest AA that one can hold without selling out equity positions in a downturn. Which I have advocated for a long time - at least since 2008 after watching my sister sell all of her stock holdings at the very bottom of the financial crisis.

I will look at the decumulation formula tomorrow.
For accumulation, it's reasonable to divert from the optimal path due to individual risk tolerance and balancing short-term needs with long term needs. Any form of short term security compromises long term goals and/or pushes back your retirement date. That's a trade-off that only you can make.

Individuals who are less versed in finance should probably use a target date retirement fund instead of building their own asset allocation. I trust that vanguard has done well to make their target date retirement fund hit an acceptable trade off between financial security and behavioral problems. Certainly I think they did a better job than the rules of thumb proposed in these topics.
User avatar
KEotSK66
Posts: 364
Joined: Wed Mar 11, 2020 7:03 pm

*

Post by KEotSK66 »

1) it is about the percentage of the portfolio that is withdrawn for expenses

this is only true in trivial cases, like when the draw is too much for the portfolio to realistically support or when the draw is so small the portfolio will always support it

2) SOR is not about “selling stocks low”

sor risk is about selling low while drawing, too many shares with depressed nav have to be sold to make up the draw amount so the ability of the portfolio to support the retiree in the future is impaired

portfolio value and distributions are based on the number of shares. nav contributes to portfolio value but if the share base is too small due to selling low too many times an astronomical nav is required to get back on track

and you may get plenty of opportunities to sell low over the full length of retirement
"i just got fluctuated out of $1,500", jerry
User avatar
Topic Author
David Jay
Posts: 9388
Joined: Mon Mar 30, 2015 5:54 am
Location: Michigan

Re: *

Post by David Jay »

KEotSK66 wrote: Wed May 13, 2020 8:43 am 2) SOR is not about “selling stocks low”

sor risk is about selling low while drawing...
Help me understand what you are claiming. Are you saying that there is no Sequence of Returns phenomenon if I only spend the cash portion of a diversified portfolio when the market is down? I'm pretty sure the math doesn't support that position...

If that is not what you mean, please explain what you do mean.
Prediction is very difficult, especially about the future - Niels Bohr | To get the "risk premium", you really do have to take the risk - nisiprius
User avatar
Topic Author
David Jay
Posts: 9388
Joined: Mon Mar 30, 2015 5:54 am
Location: Michigan

Re: Follow-up discussion on AA around one's retirement date

Post by David Jay »

Uncorrelated:

Thanks for the links to Gordon Irlam's work. Lots if interesting stuff there. Here is an interesting graph from Irlam 2017:

Image

Kind of looks like a "bond tent" to my Mark 1 eyeball...
Prediction is very difficult, especially about the future - Niels Bohr | To get the "risk premium", you really do have to take the risk - nisiprius
BigJohn
Posts: 1895
Joined: Wed Apr 02, 2014 11:27 pm

Re: Follow-up discussion on AA around one's retirement date

Post by BigJohn »

FWIW, I retired at 58 and believe that SOOR is one of the more significant risk to my retirement portfolio. As a result I went to 65% bonds at retirement. The jury is still out for me on a glide path to a higher stock allocation. I made the decision back then to hold there for 10 years and then reevaluate. So 5 years in I remain at 65% bonds (rebalanced back to that in March). I’ll keep reading, learning and watching so ask me in another 5 years if I believe in a glide path or not.
palanzo
Posts: 1382
Joined: Thu Oct 10, 2019 4:28 pm

Re: Follow-up discussion on AA around one's retirement date

Post by palanzo »

Uncorrelated wrote: Wed May 13, 2020 12:26 am
palanzo wrote: Tue May 12, 2020 10:21 pm
David Jay wrote: Tue May 12, 2020 10:17 pm
Triple digit golfer wrote: Tue May 12, 2020 9:06 pmAlso want to ask, doesn't how money is withdrawn affect the outcome? If one has a 60/40 portfolio and stocks decline and it goes to 55/45, wouldnt the first money be withdrawn from bonds until back to 60/40?
I don’t think where you withdraw the funds matters for SOR.

It comes down to the varying percentage of portfolio spending that the withdrawal represents. 50,000 is 2.6% of 1.9M, but it is only 2.4% of 2.05M (10% gain) and it is nearly 3.3% of 1.52M (20% loss). So one is spending a much higher percentage of their portfolio after a loss.
If one follows Humble Dollar advice and keeps 5 years of spending in cash then a 5 year SOR is less of an issue.
This rule of thumb allocates a larger percentage of your net worth to bonds if equities go down. But that is actually exactly the opposite of what you should do. If markets go up, there is less reason to take risk and you should be increasing the bond allocation (eventually this results in 100% annuitization). If markets go down, that necessitates taking more risk. The following graph shows the asset allocation that maximizes the probability of not running out of money, provided that you reset your asset allocation each year. The color scale is the percentage allocated to equities as opposed to bonds.

Image

I don't recommend this asset allocation because the underlying goal (maximizing the probability of success) is deeply flawed. Read Irlam's work for simulations involving more probable goals.

Advice involving cash cushions is dangerous. It sounds intuitive but will usually do more harm than good when applied in practice.
What approach would you recommend in retirement if not a cash cushion?
User avatar
Uncorrelated
Posts: 1050
Joined: Sun Oct 13, 2019 3:16 pm

Re: Follow-up discussion on AA around one's retirement date

Post by Uncorrelated »

palanzo wrote: Wed May 13, 2020 3:48 pm
What approach would you recommend in retirement if not a cash cushion?
I recommend reading the papers from https://www.aacalc.com/about titled "Floor and Upside Investing in Retirement with ..." and go from there. Asset allocation in retirement is a very complicated problem. Depending on your bequest motives, spending needs and net worth, it's possible to argue for a wide range of possible approaches ranging from full annuitization to 100% stocks. For individuals without bequest motives, it appears annuities are almost always preferred to bonds.

One thing his methods will never tell you to do is to hold a cash cushion. A cash cushion is a specific type of solution (a path dependent solution) that is suboptimal in the mathematical framework that he is using.
User avatar
Topic Author
David Jay
Posts: 9388
Joined: Mon Mar 30, 2015 5:54 am
Location: Michigan

Re: Follow-up discussion on AA around one's retirement date

Post by David Jay »

rbaldini wrote: Tue May 12, 2020 10:58 am
David Jay wrote: Tue May 12, 2020 10:55 am
rbaldini wrote: Tue May 12, 2020 10:40 am Correct me if I'm wrong, but sequence of returns risk always exists after retirement. It's not like it goes away after the first x years. It is always the case that losing a bunch now is worse than later. I suppose one difference is that you are x years older, and therefore your long term needs are reduced (in other words, you are closer to death), so maybe there isn't the same need to guard against that risk?
Sustainable withdrawal rates are strongly affected by the early years. Compounding effects occur in both positive and negative directions. A major downturn in first 5 years after retirement has a much greater effect on portfolio longevity than in - say - the 15th year of retirement.
Let's say you are in the 10th year of retirement *right now*. Is it not the case that "a major downturn" right now is much worse that a major downturn in 10 more years? (Replace 10 with any number you want).

In other words, *at every moment in retirement*, it is always worse to lose a lot in the near term than in the far term. In other words, there is always the same sequence of return risk, because the constant withdrawal rate reduces your ability to make up the difference later. If this requires you to be conservative at year 1 of your retirement, should it not require the same every year after?
I was less than satisfied with my first answer to your question:
What you say is logically true in isolation, but if you have 2 million at year 10 due to good sequence of returns in those first 10 years, it is very different from having only 500K due to a bad sequence of returns in those first 10 years.
So I decided to explore this a bit more. I have come to the conclusion that SOR impact is dramatically influenced by the distance (in number of years) between the sequence inversion points, let me show my work:

I used a similar arrangement as in my previous example bu this time I show the SOR difference between a 30 year sequence of returns inversion and a two year sequence of returns inversion:

Example (using real dollars):
$1.0M portfolio
$50K annual withdrawal for living expenses (for replication, expenses are withdrawn after portfolio return each year)

30 years
29 up years, 8% annual gain each year
1 down year, 30% loss

Position the down year at the end of the sequence (29 up years, then 1 down year), remaining portfolio after year 30: $ 2,833,285

Position the down year at the beginning of the sequence (1 down year, then 29 up years), remaining portfolio after year 30: $857,932

Over a 30 year span, the sequence of returns difference for inverting the sequence of a single down year is 330%. Note that the portfolio never recovers to it's initial value after a single initial 30% drop when using a 5% withdrawal.

2 years
1 up year, 8% gain
1 down year, 30% loss

Position the down year at the end of the sequence (1 up year, then 1 down year), remaining portfolio after year 2: $ 671,000

Position the down year at the beginning of the sequence (1 down year, then 1 up year), remaining portfolio after year 2: $ 652,000

Over a 2 year span, the sequence of returns difference for inverting the sequence of the single down year is 2.9%

In summary, the more years in the inverted sequence, the larger the effect on portfolio size. I keep coming back to this, it is the withdrawals that drive sequence of return risk. If there were no withdrawals then the commutative property of multiplication would apply and the results for the two calculations for each term would be identical (x*y*z =z*y*x).

So, in answer to the question initially posed: No, the sequence of returns risk over the final 20 years of retirement will be significantly less than the sequence of returns risk over the full 30 years of retirement.
Prediction is very difficult, especially about the future - Niels Bohr | To get the "risk premium", you really do have to take the risk - nisiprius
Post Reply