Long Bonds in retirement - curiosity...

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

abc132 wrote: Tue Jun 08, 2021 2:50 am "Withdrawing in proportion to what AA? " would be my response to what you suggest - that is not known in advance for anyone that uses probability modelling and has a diverse set of potential outcomes.
It doesn't matter WHAT the AA is at any particular point in time in this context: all that matters is that portfolio withdrawals occur proportionately from the portfolio at the time they are made.
abc132 wrote: Tue Jun 08, 2021 2:50 am
vineviz wrote: Sun Jun 06, 2021 7:37 am It's irrelevant because rebalancing has nothing to do with estimating the investor's time horizon OR calculating their portfolio's duration.
I was hoping you would think about providing something that illustrates your statement, or provides some support for it.
I did that when I pointed out that your investment horizon is calculated based only on portfolio withdrawals (i.e. consumption). The timing and size of the consumption are the only inputs into the calculation (apart from the discount rate used to convert them to present value).

Definitionally, rebalancing is NOT consumption ergo it is irrelevant.
abc132 wrote: Tue Jun 08, 2021 2:50 am
vineviz wrote: Sun Jun 06, 2021 7:37 am You just wrote this: "I'm not convinced we can plan how we plan to spend bonds in the future very accurately." I don't know how to interpret that statement except as an expression of a "belief that uncertainty about the investment horizon makes duration matching impossible ".

If you have a different belief then I invite you to clarify what you meant.
I think many Bogleheads have conservative enough plans that they may find their "survival portfolio" with 2% SWR and lots of bonds quickly turns into a legacy portfolio without the need for any bonds. Those who have done probabilistic modelling can see how quickly and easily this happens. Or you can just look at the last 12 years of history --> I've seen a lot of posts of people that surely thought they wanted bonds, but have now dumped them for other vehicles. The data I see says many people are unlikely to know what they want very far into the future, whether it is AA, spend rate, presence of bonds, etc.
Then my interpretation seems to be correct after all.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
abc132
Posts: 572
Joined: Thu Oct 18, 2018 1:11 am

Re: Long Bonds in retirement - curiosity...

Post by abc132 »

vineviz wrote: Tue Jun 08, 2021 6:18 am
abc132 wrote: Tue Jun 08, 2021 2:50 am "Withdrawing in proportion to what AA? " would be my response to what you suggest - that is not known in advance for anyone that uses probability modelling and has a diverse set of potential outcomes.
It doesn't matter WHAT the AA is at any particular point in time in this context: all that matters is that portfolio withdrawals occur proportionately from the portfolio at the time they are made.
I will walk you through the logic step by step.
1. Someone doing probability analysis does not have a fixed AA.
2. Someone changing their AA over time will need to make withdrawals non proportionally, often all of one asset.
3. Someone withdrawing will do what you suggested for duration, withdraw in a non proportional manner to achieve AA
4. Even the amount of withdrawals can be highly variable - in 2021 someone can be spending 3x what they thought they woudl be in 2009.


abc132 wrote: Tue Jun 08, 2021 2:50 am
vineviz wrote: Sun Jun 06, 2021 7:37 am It's irrelevant because rebalancing has nothing to do with estimating the investor's time horizon OR calculating their portfolio's duration.
I was hoping you would think about providing something that illustrates your statement, or provides some support for it.
vineviz wrote: Tue Jun 08, 2021 6:18 am I did that when I pointed out that your investment horizon is calculated based only on portfolio withdrawals (i.e. consumption). The timing and size of the consumption are the only inputs into the calculation (apart from the discount rate used to convert them to present value).

Definitionally, rebalancing is NOT consumption ergo it is irrelevant.
See above. This didn't help because your assumptions are not ture.



abc132 wrote: Tue Jun 08, 2021 2:50 am Then my interpretation seems to be correct after all.
I obviously disagree with your statement of my beliefs, and having asked you not to do this, I see it has had no effect.

I mentioned earlier the reasons I thought a discussion was going to be impossible, and I think we have confirmed it here.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

abc132 wrote: Tue Jun 08, 2021 1:21 pm
vineviz wrote: Tue Jun 08, 2021 6:18 am
abc132 wrote: Tue Jun 08, 2021 2:50 am "Withdrawing in proportion to what AA? " would be my response to what you suggest - that is not known in advance for anyone that uses probability modelling and has a diverse set of potential outcomes.
It doesn't matter WHAT the AA is at any particular point in time in this context: all that matters is that portfolio withdrawals occur proportionately from the portfolio at the time they are made.
I will walk you through the logic step by step.
1. Someone doing probability analysis does not have a fixed AA.
2. Someone changing their AA over time will need to make withdrawals non proportionally, often all of one asset.
3. Someone withdrawing will do what you suggested for duration, withdraw in a non proportional manner to achieve AA
4. Even the amount of withdrawals can be highly variable - in 2021 someone can be spending 3x what they thought they woudl be in 2009.
I’m not sure why you listed out those steps. None of them change the facts, which are that none of this so-called “logic” has any relationship to the topic.

It’s really simple: at any point in time the way to minimize your interest rate risk is to match your bond duration as closely as possible to your best estimate of your investment horizon.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
abc132
Posts: 572
Joined: Thu Oct 18, 2018 1:11 am

Re: Long Bonds in retirement - curiosity...

Post by abc132 »

vineviz wrote: Tue Jun 08, 2021 1:43 pm
abc132 wrote: Tue Jun 08, 2021 1:21 pm
vineviz wrote: Tue Jun 08, 2021 6:18 am
abc132 wrote: Tue Jun 08, 2021 2:50 am "Withdrawing in proportion to what AA? " would be my response to what you suggest - that is not known in advance for anyone that uses probability modelling and has a diverse set of potential outcomes.
It doesn't matter WHAT the AA is at any particular point in time in this context: all that matters is that portfolio withdrawals occur proportionately from the portfolio at the time they are made.
I will walk you through the logic step by step.
1. Someone doing probability analysis does not have a fixed AA.
2. Someone changing their AA over time will need to make withdrawals non proportionally, often all of one asset.
3. Someone withdrawing will do what you suggested for duration, withdraw in a non proportional manner to achieve AA
4. Even the amount of withdrawals can be highly variable - in 2021 someone can be spending 3x what they thought they woudl be in 2009.
I’m not sure why you listed out those steps. None of them change the facts, which are that none of this so-called “logic” has any relationship to the topic.

It’s really simple: at any point in time the way to minimize your interest rate risk is to match your bond duration as closely as possible to your best estimate of your investment horizon.
:sharebeer
international001
Posts: 2017
Joined: Thu Feb 15, 2018 7:31 pm

Re: Long Bonds in retirement - curiosity...

Post by international001 »

vineviz wrote: Mon Jun 07, 2021 8:37 pm
You wouldn't want a single 30-year bond in this case, but if you DID buy a single 30-year bond as the "40%" in your 60/40 portfolio then the bond would account for roughly half the overall volatility of the portfolio. I wouldn't call that insignificant.
In terms of volatility, consider

stocks volatility 15%, LT bond fund volatility 11%, one LT bond volatility 0%, LT bonds and stocks independent variable

60% stocks, 40% LT bond fund -> volatility 10% ( sqrt( (15%*60%)**2 + (11%*40%)**2 ))
67% stocks, 33% one LT bond -> volatility 10% (15%*65%)

So equal volatility, but for the second the expected return may be better. Am I looking at it the right way?
What it seems to me everything depends on the rate you locked when you buy the bond. You could be unlucky and buy it when yields are low.
So the risk should be considered before you buy that bond (the alternative being a bond fund that will allow you to diversify over multiple yields), not after you bought the bond

vineviz wrote: Mon Jun 07, 2021 8:37 pm And if you truly wanted to minimize the risk on the overall portfolio then a ladder of TIPS spanning the full 30 years would be the way to do that.
Why not one single 30 year TIP ?
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

international001 wrote: Tue Jun 08, 2021 5:37 pm So the risk should be considered before you buy that bond (the alternative being a bond fund that will allow you to diversify over multiple yields), not after you bought the bond
Interest rate risk isn't dependent on the level of yields, just on the match (or lack thereof) between the duration of your bonds and your investment horizons.

And "diversify over multiple yields" isn't an actual thing, at least not in the financial sense of the word "diversify".
international001 wrote: Tue Jun 08, 2021 5:37 pm
vineviz wrote: Mon Jun 07, 2021 8:37 pm And if you truly wanted to minimize the risk on the overall portfolio then a ladder of TIPS spanning the full 30 years would be the way to do that.
Why not one single 30 year TIP ?
Because you presumably have relatively even consumption in each year of the 30 years, not just a little bit in years 1 to 29 and then a huge amount in year 30 (which would be the way the coupon payments and principal payment of a single bond would arrive).

The least risky approach would be a ladder of TIPS, with some bonds maturing each year. That way the natural cashflows from the bonds would match your expected consumption.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
dbr
Posts: 35993
Joined: Sun Mar 04, 2007 9:50 am

Re: Long Bonds in retirement - curiosity...

Post by dbr »

vineviz wrote: Tue Jun 08, 2021 8:11 pm

Because you presumably have relatively even consumption in each year of the 30 years, not just a little bit in years 1 to 29 and then a huge amount in year 30 (which would be the way the coupon payments and principal payment of a single bond would arrive).

The least risky approach would be a ladder of TIPS, with some bonds maturing each year. That way the natural cashflows from the bonds would match your expected consumption.
This is probably the ultimate pure example of fixed income being exactly that. The idea is nearly a no-brainer from a certain point of view.

I personally tend to think of this Liability Matching Portfolio, as it has been called, as not a portfolio, not even an investment at all. The ultimate version of such a thing would be an inflation indexed SPIA.
NiceUnparticularMan
Posts: 2305
Joined: Sat Mar 11, 2017 7:51 am

Re: Long Bonds in retirement - curiosity...

Post by NiceUnparticularMan »

vineviz wrote: Sun Jun 06, 2021 7:37 am The consistent thread here seems to be a belief that uncertainty about the investment horizon makes duration matching impossible to do, but it doesn't seem to me to be the general case that investors have that much uncertainty. People routine form estimates about things like future income, longevity, retirement consumption, tax rates, etc. about which they are uncertain but it doesn't prevent us from doing any financial planning which involves those factors. Why should we consider uncertainty to a particularly acute problem only in this discussion?
Uncertainty about these other issues is, or should be, considered in financial planning as well.

Take longevity. People are rightly quite uncertain about that, and at least the people who take that risk seriously account for longevity risk in their planning. Fortunately Social Security provides some longevity insurance, and some people have pensions as well, but people routinely consider other forms of longevity insurance (like SPIAs). And if not, they adopt lower SWRs as a way of self-insuring against longevity risk.

What people (hopefully) do not do is look up their actuarial results and then just assume those will be their results. That information can still be useful in terms of guiding some of their planning, but their planning should also reflect the large possibility their lifespan will be considerably shorter or longer.

As I have noted before, this is an example of how people should think in terms of probability distributions, and not substitute point estimates, as doing that assumes away important risks.
NiceUnparticularMan
Posts: 2305
Joined: Sat Mar 11, 2017 7:51 am

Re: Long Bonds in retirement - curiosity...

Post by NiceUnparticularMan »

dbr wrote: Tue Jun 08, 2021 8:24 pm
vineviz wrote: Tue Jun 08, 2021 8:11 pm

Because you presumably have relatively even consumption in each year of the 30 years, not just a little bit in years 1 to 29 and then a huge amount in year 30 (which would be the way the coupon payments and principal payment of a single bond would arrive).

The least risky approach would be a ladder of TIPS, with some bonds maturing each year. That way the natural cashflows from the bonds would match your expected consumption.
This is probably the ultimate pure example of fixed income being exactly that. The idea is nearly a no-brainer from a certain point of view.

I personally tend to think of this Liability Matching Portfolio, as it has been called, as not a portfolio, not even an investment at all. The ultimate version of such a thing would be an inflation indexed SPIA.
Yeah, I think it is useful sometimes to separate out two different concepts--investment for the purpose of wealth-building, and saving for the purpose of deferring spending.

Liability-matching is a form of saving in the latter sense, because you assume known future spending (liabilities), and you are going to take some money you could spend now, but instead buy something which will fund that future spending instead.

One downside to liability-matching is just that if you do it in a really low-risk way, you might not build much wealth along the way, and indeed might lose wealth in real terms (see current TIPS rates). That's not necessarily a problem as long as you don't need any more wealth, and in fact can lose a little over time, and still meet your financial goals. But it is one common consequence of such an approach.

The other downside we keep discussing is that the specific timing of when a given family might want to spend money is not typically all that certain. I know another poster believes this uncertainty can be reduced to point estimates anyway. I believe it really should be kept as probability distributions and not collapsed into point estimates. And once you think in terms of probability distributions, saving for deferred spending becomes a significantly more complicated analytic tasks.

Now, the solution I am seeing some people here taking to these issues these days is just to take only a portion of their wealth and stick it in a TIPS ladder (maybe only until they anticipate taking Social Security), and have another portion of their wealth invested in ways consistent with uncertainty about the timing of future spending and also the possibility of additional wealth building. And that's a perfectly sensible approach, assuming you have enough wealth to fund all that adequately!

Which a lot of people here do, after a long period of good returns on typical Bogleheads-style accumulation plans. But, the next cohorts might not be so fortunate with what happens during their accumulation phases, at which point some of these tradeoffs might not be solvable for so many people with a "why not both?" approach.
abc132
Posts: 572
Joined: Thu Oct 18, 2018 1:11 am

Re: Long Bonds in retirement - curiosity...

Post by abc132 »

NiceUnparticularMan wrote: Wed Jun 09, 2021 5:41 am
...

As I have noted before, this is an example of how people should think in terms of probability distributions, and not substitute point estimates, as doing that assumes away important risks.
Yes, I fully agree.

This is how decision making should be done.

If we argue for altering spending based on anything that changes with portfolio value or life expectancy (VPW, CAPE10, etc), that should be in the calculations to meet this expected spending. This is a function of market performance, and not known in advance. It is trivial to handle through probability calculations. The range of outcomes will demonstrate how accurately we might expect to be able to predict spending.

Using consistent assumptions for how we spend and how we match spending is important, otherwise it will be a case of garbage-in garbage-out.

Probability informs decisions, it does not prevent them.

The level of uncertainty is irrelevant to good decision making, as it informs rather than prevents the decision.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

NiceUnparticularMan wrote: Wed Jun 09, 2021 5:41 am As I have noted before, this is an example of how people should think in terms of probability distributions, and not substitute point estimates, as doing that assumes away important risks.
Thinking "in terms of probability distributions" is not incompatible with using point estimates in your modeling. At typical individual investor shouldn't need (and doesn't need) a sophisticated decision tree or Monte Carlo simulation in order to figure out whether they should use Vanguard Short-Term Treasury ETF (VGSH) or Vanguard Long-Term Treasury ETF (VGLT) for their bond allocation.

Saying that you estimate your investment horizon to be about 40 years (for instance) doesn't necessarily mean you're "assuming away" any risks or uncertainty, since the 40 year estimate might very well reflect a 50% chance of the actual horizon being 60 years and a 50% chance of the actual horizon being 20 years. The uncertainty is an input into the point estimate.

And because investors are typically highly loss averse AND because longevity is one of the most powerful sources of uncertainty in retirement planning, the planned investment horizon is effectively already incorporating the distribution of probabilities. For instance, planning a retirement that lasts until age 95 is not planning for the median outcome in terms of longevity but for something closer to a 20th percentile outcome for most people. And people with highly volatiles earned income will have larger emergency funds which, typically being invested in cash or short-term instruments, effectively lowers the duration of their overall household portfolio. Etc.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
NiceUnparticularMan
Posts: 2305
Joined: Sat Mar 11, 2017 7:51 am

Re: Long Bonds in retirement - curiosity...

Post by NiceUnparticularMan »

vineviz wrote: Thu Jun 10, 2021 8:57 am
NiceUnparticularMan wrote: Wed Jun 09, 2021 5:41 am As I have noted before, this is an example of how people should think in terms of probability distributions, and not substitute point estimates, as doing that assumes away important risks.
Thinking "in terms of probability distributions" is not incompatible with using point estimates in your modeling. At typical individual investor shouldn't need (and doesn't need) a sophisticated decision tree or Monte Carlo simulation in order to figure out whether they should use Vanguard Short-Term Treasury ETF (VGSH) or Vanguard Long-Term Treasury ETF (VGLT) for their bond allocation.

Saying that you estimate your investment horizon to be about 40 years (for instance) doesn't necessarily mean you're "assuming away" any risks or uncertainty, since the 40 year estimate might very well reflect a 50% chance of the actual horizon being 60 years and a 50% chance of the actual horizon being 20 years. The uncertainty is an input into the point estimate.
So one of the immediate implications of think in terms of probability distributions over spending is you can no longer assume things like that you will hold a given bond to maturity, at which point you will have spent all the proceeds.

This is important because obviously liquidating a bond position well before you originally expected can in fact have some pretty significant consequences. And more subtly, but just as important, there can be significant opportunity costs to not actually needing the money back as soon as you expected.

So, these are not equivalent assumptions:

(1) I will need this money back in X years;

(2) There is a 1/3rd chance I will need this money back in X years, but also a 1/3rd chance I will need it back much sooner than X years, and a 1/3rd chance I will need it back much later than X years.

Now of course if you were, say, an insurance company and covering a large number of such personal investors, then it might make sense to think that enough people in the second position start to resemble just the first position (although that depends on assumptions that are not always valid).

But if you are just one individual planning for one family, you have to account for what happens in those alternative scenarios. And in many cases, it will make sense to give up some expected return in the middle scenario in order to provide against the other scenarios.
And because investors are typically highly loss averse AND because longevity is one of the most powerful sources of uncertainty in retirement planning, the planned investment horizon is effectively already incorporating the distribution of probabilities. For instance, planning a retirement that lasts until age 95 is not planning for the median outcome in terms of longevity but for something closer to a 20th percentile outcome for most people. And people with highly volatiles earned income will have larger emergency funds which, typically being invested in cash or short-term instruments, effectively lowers the duration of their overall household portfolio. Etc.
Right, in practice people here are already providing for these various alternate scenarios--as they should!

But that doesn't mean that it is easy to know exactly how to do it.

And unfortunately, I think there is a bit of a distortion going on here in that due to the recent pleasant history of typical "Boglehead" plans, a lot of people here are now in a position where they are wealthy enough that it is relatively easy for them to self-insure. Meaning they have accumulated so much wealth they can put a lot in SPIAs (or already have pensions or can defer Social Security and so on), AND put a bunch more in TIPS ladders that will cover a lot of expected spending even though current TIPS rates are low, AND still have a bunch left over for liquid short-term funds, AND still have a bunch left over for long-term at-risk investments.

But it might not always be so easy. If there is an extend period where things are less pleasant for typical Boglehead plans, then people might be facing harder choices about what to do, how much insurance of what kind to buy, and so on.

So, I don't think it is a good idea to assume these are actually always going to be easy issues for everyone. They are relatively easy if you are really wealthy, but not everyone is there yet, and it might be a lot harder to get there in the next long period.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

NiceUnparticularMan wrote: Thu Jun 10, 2021 9:42 am So, these are not equivalent assumptions:

(1) I will need this money back in X years;

(2) There is a 1/3rd chance I will need this money back in X years, but also a 1/3rd chance I will need it back much sooner than X years, and a 1/3rd chance I will need it back much later than X years.
If you're thinking probabilistically, as I thought we agreed that people should do, they are indeed equivalent statements.

A wager which has a 10 % chance of paying you $90 has the same expected value as a wager which has a 90% chance of paying you $10. If you're being rational, neither wager has more value than the other: you'd pay the same price to receive either wager.

NiceUnparticularMan wrote: Thu Jun 10, 2021 9:42 am But if you are just one individual planning for one family, you have to account for what happens in those alternative scenarios. And in many cases, it will make sense to give up some expected return in the middle scenario in order to provide against the other scenarios.
It's quite obvious what bond duration an investor should choose if they have 100% probability of needing the money in 10 years.

Can you explain why the investor would choose a DIFFERENT bond duration if they have a 1/3 chance of needing the money in 5 years, a 1/3 chance of needing the money in 10 years, and a 1/3 chance of needing the money in 15 years?

I've never seen an economic model which gives any answer other than 10 years as the appropriate duration for the second scenario, but if you've got one in mind I'd love to know about it.

If any of this discussion is to amount to anything other than hand-waving, there's got to be some way to explain how to get a different SOLUTION to the two scenarios you presented above. Because if they have the same solution then what's the value in taking such pains to distinguish them?
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
NiceUnparticularMan
Posts: 2305
Joined: Sat Mar 11, 2017 7:51 am

Re: Long Bonds in retirement - curiosity...

Post by NiceUnparticularMan »

vineviz wrote: Thu Jun 10, 2021 1:37 pm
NiceUnparticularMan wrote: Thu Jun 10, 2021 9:42 am So, these are not equivalent assumptions:

(1) I will need this money back in X years;

(2) There is a 1/3rd chance I will need this money back in X years, but also a 1/3rd chance I will need it back much sooner than X years, and a 1/3rd chance I will need it back much later than X years.
If you're thinking probabilistically, as I thought we agreed that people should do, they are indeed equivalent statements.

A wager which has a 10 % chance of paying you $90 has the same expected value as a wager which has a 90% chance of paying you $10. If you're being rational, neither wager has more value than the other: you'd pay the same price to receive either wager.
So that is true if you are dealing with low-stakes wagers. In cases like that, you can assume a flat $ to utility relationship as the stakes are low.

OK, but take your numbers and now make it $90K and $10K, and assume you need $10K for a life-saving surgery or you will die.

In that case, a wager with a 90% chance of getting you $10K is in fact a better bet than a wager with a 10% chance of paying you $90K. Because it improves your chances of surviving from 10% to 90%.

What has happened in this case is we have made the stakes higher and tied them to a utility discontinuity.

Of course if you thought of expected value in terms of utility, you would get the right answer. We can assume surviving is worth an arbitrarily large amount of dollars--say $10 million. A 90% chance of $10,000,000 is worth more than a 10% chance of $10,080,000.

These are important principles to keep in mind when dealing with high stakes financial bets. If you do those calculations in terms of dollars rather than utility, you can easily get the wrong answers.

And that sort of issue will often arise for real world families. In many cases, having $N less to spend in one time will subtract more utility than adding another $N to spend in a different time will add utility.

NiceUnparticularMan wrote: Thu Jun 10, 2021 9:42 am But if you are just one individual planning for one family, you have to account for what happens in those alternative scenarios. And in many cases, it will make sense to give up some expected return in the middle scenario in order to provide against the other scenarios.
It's quite obvious what bond duration an investor should choose if they have 100% probability of needing the money in 10 years.

Can you explain why the investor would choose a DIFFERENT bond duration if they have a 1/3 chance of needing the money in 5 years, a 1/3 chance of needing the money in 10 years, and a 1/3 chance of needing the money in 15 years?
Sure, we're almost there already. Assume "need" in this cases means you need it for a life-saving surgery. If you choose a bond duration of 10 years, you might not have enough money to save your life in the event you need that money to pay for the life-saving surgery in 5 years.
abc132
Posts: 572
Joined: Thu Oct 18, 2018 1:11 am

Re: Long Bonds in retirement - curiosity...

Post by abc132 »

vineviz wrote: Thu Jun 10, 2021 1:37 pm
If any of this discussion is to amount to anything other than hand-waving, there's got to be some way to explain how to get a different SOLUTION to the two scenarios you presented above. Because if they have the same solution then what's the value in taking such pains to distinguish them?
There are many ways to do it, but one way is to compare two specific scenarios.

Let's take a really simple example to illustrate the point:

The OP at Age 61 can either
1) take Vineviz's suggestion of 14 year duration (check out Vineviz "first 20 years in LTT thread" for details).
2) compare alternatives

What if we ensure vs longevity not by duration matching to age 95 but by taking on an age based annuity at age 80?

This is a plausible scenario that could have better portfolio characteristics.

The desired duration is now 11.4 years, which is oddly similar to what I suggested earlier in the thread.

This is one example why someone may rationally prefer a somewhat shorter duration.

Extending that to someone at age 69, we are now comparing medium vs long duration, depending on if we plan to insure against longevity through an age-based annuity. This a clear example that "medium" vs "long" is not always going to be an appropriate differentiator using the method Vineviz has proposed.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

NiceUnparticularMan wrote: Thu Jun 10, 2021 2:09 pm Sure, we're almost there already. Assume "need" in this cases means you need it for a life-saving surgery. If you choose a bond duration of 10 years, you might not have enough money to save your life in the event you need that money to pay for the life-saving surgery in 5 years.
You’re not drawing closer to the solution. You’re confusing yourself by changing the scenario such that it no longer resembles the one we started with.

Let’s say you knew for certain that you will eventually need this surgery and you know for sure it will cost $100 but you don’t know when you will need it.

It could be 5 years, it could be 10 years, or it could be 15 years.

You have $90 now. 5-year bonds yield 0.5%. 10-year bonds yield 1%. 15 -year bonds yield 1.5%.

What process do you use to choose how to allocate your $90?
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

abc132 wrote: Thu Jun 10, 2021 2:16 pm
What if we ensure vs longevity not by duration matching to age 95 but by taking on an age based annuity at age 80?

This is a plausible scenario that could have better portfolio characteristics.

The desired duration is now 11.4 years, which is oddly similar to what I suggested earlier in the thread.
That’s not the desired duration if you intend to purchase the annuity at age 80.

It is the desired duration only after you purchase a deferred annuity at age 61 which starts paying out at age 80, and in this case it is the desired duration precisely because it own matches your newly reduced investment horizon.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
abc132
Posts: 572
Joined: Thu Oct 18, 2018 1:11 am

Re: Long Bonds in retirement - curiosity...

Post by abc132 »

vineviz wrote: Thu Jun 10, 2021 2:51 pm
abc132 wrote: Thu Jun 10, 2021 2:16 pm
What if we ensure vs longevity not by duration matching to age 95 but by taking on an age based annuity at age 80?

This is a plausible scenario that could have better portfolio characteristics.

The desired duration is now 11.4 years, which is oddly similar to what I suggested earlier in the thread.
That’s not the desired duration if you intend to purchase the annuity at age 80.

It is the desired duration only after you purchase a deferred annuity at age 61 which starts paying out at age 80, and in this case it is the desired duration precisely because it own matches your newly reduced investment horizon.
If I thought I was going to need something in 14 years yesterday, and today I find out I need it in 11 years, then I needed that exact same thing in 11 years and 1 day yesterday. You are saying the new estimate changes according to duration matching (correctly), but in fact there was a mismatch between when the portfolio was used and when it was originally expected to be used with the original 14 year duration.

One would have been better off assuming 11 years than 14 years, even if that violates your concept (technical definition) of duration matching. The investor does not want to duration match as you are doing, they would rather have predicted the true duration of expenses a day earlier.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

abc132 wrote: Thu Jun 10, 2021 5:21 pm
If I thought I was going to need something in 14 years yesterday, and today I find out I need it in 11 years, then I needed that exact same thing in 11 years and 1 day yesterday.
If yesterday you thought there was a 100% probability that your investment horizon was 14 years and today you think there’s a 100% probability that your investment horizon is 11 years then you have either weird luck or an inability to make accurate forecasts.

Life is uncertain, but we have to plan anyway. The only way to do that is to make the best estimates you can using the information available to you at the time you make them.

When you get new information, you update your estimates.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
international001
Posts: 2017
Joined: Thu Feb 15, 2018 7:31 pm

Re: Long Bonds in retirement - curiosity...

Post by international001 »

dbr wrote: Tue Jun 08, 2021 8:24 pm
vineviz wrote: Tue Jun 08, 2021 8:11 pm

Because you presumably have relatively even consumption in each year of the 30 years, not just a little bit in years 1 to 29 and then a huge amount in year 30 (which would be the way the coupon payments and principal payment of a single bond would arrive).

The least risky approach would be a ladder of TIPS, with some bonds maturing each year. That way the natural cashflows from the bonds would match your expected consumption.
This is probably the ultimate pure example of fixed income being exactly that. The idea is nearly a no-brainer from a certain point of view.

I personally tend to think of this Liability Matching Portfolio, as it has been called, as not a portfolio, not even an investment at all. The ultimate version of such a thing would be an inflation indexed SPIA.
Ooops my bad, I didn't understand how TIPs worked
So what would you suggest if you plan to leave for 30 years? One TIP of 30 year duration, One TIP of 29 year duration, ... , and one TIP of 1 year duration?
dbr
Posts: 35993
Joined: Sun Mar 04, 2007 9:50 am

Re: Long Bonds in retirement - curiosity...

Post by dbr »

international001 wrote: Thu Jun 10, 2021 5:49 pm
dbr wrote: Tue Jun 08, 2021 8:24 pm
vineviz wrote: Tue Jun 08, 2021 8:11 pm

Because you presumably have relatively even consumption in each year of the 30 years, not just a little bit in years 1 to 29 and then a huge amount in year 30 (which would be the way the coupon payments and principal payment of a single bond would arrive).

The least risky approach would be a ladder of TIPS, with some bonds maturing each year. That way the natural cashflows from the bonds would match your expected consumption.
This is probably the ultimate pure example of fixed income being exactly that. The idea is nearly a no-brainer from a certain point of view.

I personally tend to think of this Liability Matching Portfolio, as it has been called, as not a portfolio, not even an investment at all. The ultimate version of such a thing would be an inflation indexed SPIA.
Ooops my bad, I didn't understand how TIPs worked
So what would you suggest if you plan to leave for 30 years? One TIP of 30 year duration, One TIP of 29 year duration, ... , and one TIP of 1 year duration?
I don't suggest it. I mention that it is suggested by people including Bill Bernstein. There are a number of threads on the topic of how to actually do this because TIPS are not available that hit all 30 years needed. But, yes, you have the idea correctly.
abc132
Posts: 572
Joined: Thu Oct 18, 2018 1:11 am

Re: Long Bonds in retirement - curiosity...

Post by abc132 »

vineviz wrote: Thu Jun 10, 2021 5:47 pm
abc132 wrote: Thu Jun 10, 2021 5:21 pm
If I thought I was going to need something in 14 years yesterday, and today I find out I need it in 11 years, then I needed that exact same thing in 11 years and 1 day yesterday.
If yesterday you thought there was a 100% probability that your investment horizon was 14 years and today you think there’s a 100% probability that your investment horizon is 11 years then you have either weird luck or an inability to make accurate forecasts.

Life is uncertain, but we have to plan anyway. The only way to do that is to make the best estimates you can using the information available to you at the time you make them.

When you get new information, you update your estimates.
The best planning includes the possibility of likely and unlikely actions before they happen.

Someone considering the annuity would be better of somewhere between 11 and 14, and certainly doesn't need 100% probability of anything. 0.001% probability works just fine in planning.

This is a clear example of why we might rationally reduce duration from what you suggest, and that effort to duration match is limited by the degree to which we can predict future expenses.
Lee_WSP
Posts: 4822
Joined: Fri Apr 19, 2019 5:15 pm
Location: Arizona

Re: Long Bonds in retirement - curiosity...

Post by Lee_WSP »

vineviz wrote: Thu Jun 10, 2021 1:37 pm Can you explain why the investor would choose a DIFFERENT bond duration if they have a 1/3 chance of needing the money in 5 years, a 1/3 chance of needing the money in 10 years, and a 1/3 chance of needing the money in 15 years?

I've never seen an economic model which gives any answer other than 10 years as the appropriate duration for the second scenario, but if you've got one in mind I'd love to know about it.
I feel like I should know this answer, but I'm drawing a blank at the moment.

If the rates for each duration are different, don't you get a different result if you bought 1/3 of each duration if buying a bond fund as opposed to individual bonds held to maturity?

Edit: PV has a semi-answer.

https://www.portfoliovisualizer.com/bac ... tion3_2=34

Looks similar, but slightly different. The maturities probably don't quite average out to 10 is my best guess.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

abc132 wrote: Thu Jun 10, 2021 5:54 pm This is a clear example of why we might rationally reduce duration from what you suggest, and that effort to duration match is limited by the degree to which we can predict future expenses.
No, this is a clear example of EXACTLY how duration matching works in practice and EXACTLY how I describe it working.

If there is a 33% probability your investment horizon is 11 years and a 67% probability your investment horizon is 14 years, then your bond duration should be 13 years to match the weighted average of those two possibilities.

Either way, even if you don't feel comfortable putting probabilities on those two outcomes then at least the framing of the question tells you that either 11 years or 14 years is MUCH closer to the risk-free solution than the 6 year duration of total bond market of the zero duration of cash.

It's not nearly as hard as you trying to portray it to be.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
SteadyOne
Posts: 436
Joined: Fri Mar 22, 2019 5:26 pm

Re: Long Bonds in retirement - curiosity...

Post by SteadyOne »

vineviz wrote: Tue Jun 08, 2021 8:11 pm
international001 wrote: Tue Jun 08, 2021 5:37 pm So the risk should be considered before you buy that bond (the alternative being a bond fund that will allow you to diversify over multiple yields), not after you bought the bond
Interest rate risk isn't dependent on the level of yields, just on the match (or lack thereof) between the duration of your bonds and your investment horizons.

And "diversify over multiple yields" isn't an actual thing, at least not in the financial sense of the word "diversify".
international001 wrote: Tue Jun 08, 2021 5:37 pm
vineviz wrote: Mon Jun 07, 2021 8:37 pm And if you truly wanted to minimize the risk on the overall portfolio then a ladder of TIPS spanning the full 30 years would be the way to do that.
Why not one single 30 year TIP ?
Because you presumably have relatively even consumption in each year of the 30 years, not just a little bit in years 1 to 29 and then a huge amount in year 30 (which would be the way the coupon payments and principal payment of a single bond would arrive).

The least risky approach would be a ladder of TIPS, with some bonds maturing each year. That way the natural cashflows from the bonds would match your expected consumption.
Would a TIPS fund be a reasonable choice here?
“Every de­duc­tion is al­lowed as a mat­ter of leg­isla­tive grace.” US Federal Court
dbr
Posts: 35993
Joined: Sun Mar 04, 2007 9:50 am

Re: Long Bonds in retirement - curiosity...

Post by dbr »

SteadyOne wrote: Thu Jun 10, 2021 7:29 pm
vineviz wrote: Tue Jun 08, 2021 8:11 pm
international001 wrote: Tue Jun 08, 2021 5:37 pm So the risk should be considered before you buy that bond (the alternative being a bond fund that will allow you to diversify over multiple yields), not after you bought the bond
Interest rate risk isn't dependent on the level of yields, just on the match (or lack thereof) between the duration of your bonds and your investment horizons.

And "diversify over multiple yields" isn't an actual thing, at least not in the financial sense of the word "diversify".
international001 wrote: Tue Jun 08, 2021 5:37 pm
vineviz wrote: Mon Jun 07, 2021 8:37 pm And if you truly wanted to minimize the risk on the overall portfolio then a ladder of TIPS spanning the full 30 years would be the way to do that.
Why not one single 30 year TIP ?
Because you presumably have relatively even consumption in each year of the 30 years, not just a little bit in years 1 to 29 and then a huge amount in year 30 (which would be the way the coupon payments and principal payment of a single bond would arrive).

The least risky approach would be a ladder of TIPS, with some bonds maturing each year. That way the natural cashflows from the bonds would match your expected consumption.
Would a TIPS fund be a reasonable choice here?
No A TIPS fund does not deliver a cash flow matched to when you want to spend the money. The idea of the ladder is that the ladder is completely consumed by each bond being spent as it matures. Since you know the inflation indexed value you will have at maturity it is a predictable stream of real income. In effect it is an inflation indexed annuity without pooled longevity risk, a big difference, by the way. Note that at negative real interest rates or even zero a TIPS ladder is expensive to acquire. These are not panaceas to avoid low interest rates. SPIA payouts are also low when interest rates are low.

LMPs are not about just dumping a bunch of assets into lower risk investments.
abc132
Posts: 572
Joined: Thu Oct 18, 2018 1:11 am

Re: Long Bonds in retirement - curiosity...

Post by abc132 »

vineviz wrote: Thu Jun 10, 2021 6:56 pm
abc132 wrote: Thu Jun 10, 2021 5:54 pm This is a clear example of why we might rationally reduce duration from what you suggest, and that effort to duration match is limited by the degree to which we can predict future expenses.
No, this is a clear example of EXACTLY how duration matching works in practice and EXACTLY how I describe it working.

If there is a 33% probability your investment horizon is 11 years and a 67% probability your investment horizon is 14 years, then your bond duration should be 13 years to match the weighted average of those two possibilities.

Either way, even if you don't feel comfortable putting probabilities on those two outcomes then at least the framing of the question tells you that either 11 years or 14 years is MUCH closer to the risk-free solution than the 6 year duration of total bond market of the zero duration of cash.
So now you are using probability based spending and duration and not constant consumption, which is in agreement with what I have said all along.

If you agree we can use probability and find a shorter duration than your age based calculation, please just agree.

Please don't argue against using probability to find a shorter duration while using probability and getting a shorter duration.

One possibility reduced your duration by 1 year. Considering ALL of the reasons we might reasonably need early spending, it is not hard to see why the reduction should be more than 1 year.

If you find it unnecessary, I take no issue with that being your opinion. I will note that the difference in duration calculations becomes bigger (more than 3 years) as age increases past 60, so I think by age 70 your age-based constant-consumption method can be worse than taking no action for someone with intermediate bonds. At age 20-50's the difference in duration calculation is not going to make much of a difference, and your advice in your age based thread is excellent for most of these people just getting into bonds, assuming they understand and accept the additional price fluctuations.

For the record, I argued 11.3 vs 14, not 6 vs 14.

In my personal opinion (not a fact), being 3 over isn't enough of a difference as compared to 5 under to warrant the change you previously suggested (duration=14 years), especially given the preference for simplicity (not wanting a spouse to have to duration match, and a desire for a more simple portfolio, not a more complex portfolio).
vineviz wrote: Thu Jun 10, 2021 6:56 pm It's not nearly as hard as you trying to portray it to be.
I've said it is simple all along. Subtract 3 from your 14 year duration and call it 11, and you are good.

I believe that using probability is very simple, and that duration matching is possible, despite misstatements to the contrary.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

dbr wrote: Thu Jun 10, 2021 7:39 pm
SteadyOne wrote: Thu Jun 10, 2021 7:29 pm Would a TIPS fund be a reasonable choice here?
No A TIPS fund does not deliver a cash flow matched to when you want to spend the money. The idea of the ladder is that the ladder is completely consumed by each bond being spent as it matures. Since you know the inflation indexed value you will have at maturity it is a predictable stream of real income. In effect it is an inflation indexed annuity without pooled longevity risk, a big difference, by the way. Note that at negative real interest rates or even zero a TIPS ladder is expensive to acquire. These are not panaceas to avoid low interest rates. SPIA payouts are also low when interest rates are low.
Slow down, please.

A TIPS fund is a VERY reasonable alternative to building a custom ladder of individual TIPS.

Using a fund doesn't perfectly replicate the cash flow characteristics of the ladder, of course, but as long as the duration of the fund matches the duration of the ladder the two approaches will support the same cash flows with the same amount of risk.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

abc132 wrote: Thu Jun 10, 2021 10:16 pm If you agree we can use probability and find a shorter duration than your age based calculation, please just agree.
I've consistently said that estimating the investment horizon directly based on expected cash flows (including the use of probability weighting) can result in a more accurate estimate than using age-based rules of thumb. In fact I even created an entire thread about it in February which describes HOW to do it.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
dbr
Posts: 35993
Joined: Sun Mar 04, 2007 9:50 am

Re: Long Bonds in retirement - curiosity...

Post by dbr »

vineviz wrote: Fri Jun 11, 2021 12:22 am
dbr wrote: Thu Jun 10, 2021 7:39 pm
SteadyOne wrote: Thu Jun 10, 2021 7:29 pm Would a TIPS fund be a reasonable choice here?
No A TIPS fund does not deliver a cash flow matched to when you want to spend the money. The idea of the ladder is that the ladder is completely consumed by each bond being spent as it matures. Since you know the inflation indexed value you will have at maturity it is a predictable stream of real income. In effect it is an inflation indexed annuity without pooled longevity risk, a big difference, by the way. Note that at negative real interest rates or even zero a TIPS ladder is expensive to acquire. These are not panaceas to avoid low interest rates. SPIA payouts are also low when interest rates are low.
Slow down, please.

A TIPS fund is a VERY reasonable alternative to building a custom ladder of individual TIPS.

Using a fund doesn't perfectly replicate the cash flow characteristics of the ladder, of course, but as long as the duration of the fund matches the duration of the ladder the two approaches will support the same cash flows with the same amount of risk.
Presumably you would have to use more than one fund as the duration is declining through the course of the retirement.
NiceUnparticularMan
Posts: 2305
Joined: Sat Mar 11, 2017 7:51 am

Re: Long Bonds in retirement - curiosity...

Post by NiceUnparticularMan »

vineviz wrote: Thu Jun 10, 2021 2:44 pm
NiceUnparticularMan wrote: Thu Jun 10, 2021 2:09 pm Sure, we're almost there already. Assume "need" in this cases means you need it for a life-saving surgery. If you choose a bond duration of 10 years, you might not have enough money to save your life in the event you need that money to pay for the life-saving surgery in 5 years.
You’re not drawing closer to the solution. You’re confusing yourself by changing the scenario such that it no longer resembles the one we started with.
To be precise, I am noting that real world families do not have a simple relationship between marginal dollars spent and utility over time, such as transferring dollars, even in real terms, between one period and another period can make a significant difference in utility.
Let’s say you knew for certain that you will eventually need this surgery and you know for sure it will cost $100 but you don’t know when you will need it.

It could be 5 years, it could be 10 years, or it could be 15 years.

You have $90 now. 5-year bonds yield 0.5%. 10-year bonds yield 1%. 15 -year bonds yield 1.5%.

What process do you use to choose how to allocate your $90?
Well, to start, ideally I would have a lot more than $90 to save for this purpose, because surgery costs are not fixed in nominal terms, and in fact medical costs tend to go up more than CPI.

So, just funding the 5-year scenario is already in significant jeopardy of failing. Just for that scenario, I would want to put more than $100 in 5-year TIPS, more both because 5-year TIPS already have negative real returns just as to CPI (I think like -1.75% or so--yikes!), and then even more because of how medical costs tend to inflate.

And that is just for the 5-year scenario! Assuming I get past 5 years, I now have to also provide for 10 years. If I knew it was 10 years, I could just do the same sort of calculation, now using something like -0.85% instead and then accounting for medical cost inflating faster. Instead, I have to figure out what I have left in real dollars by that point after round 1. Then I will have to estimate reinvesting that in another 5-year TIPS in round 2. But at what rates? I don't know. And then I will have to do this all AGAIN to get to 15 years.

And yes, that is all very complicated. But, it is the right way to think of it! If you instead just treat the problem as if you know it will be 10 years, you are going to be at significant risk of failing at 5 years (due to premature selling), or 15 years (due to more time for inflation).
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

dbr wrote: Fri Jun 11, 2021 7:21 am
vineviz wrote: Fri Jun 11, 2021 12:22 am
dbr wrote: Thu Jun 10, 2021 7:39 pm
SteadyOne wrote: Thu Jun 10, 2021 7:29 pm Would a TIPS fund be a reasonable choice here?
No A TIPS fund does not deliver a cash flow matched to when you want to spend the money. The idea of the ladder is that the ladder is completely consumed by each bond being spent as it matures. Since you know the inflation indexed value you will have at maturity it is a predictable stream of real income. In effect it is an inflation indexed annuity without pooled longevity risk, a big difference, by the way. Note that at negative real interest rates or even zero a TIPS ladder is expensive to acquire. These are not panaceas to avoid low interest rates. SPIA payouts are also low when interest rates are low.
Slow down, please.

A TIPS fund is a VERY reasonable alternative to building a custom ladder of individual TIPS.

Using a fund doesn't perfectly replicate the cash flow characteristics of the ladder, of course, but as long as the duration of the fund matches the duration of the ladder the two approaches will support the same cash flows with the same amount of risk.
Presumably you would have to use more than one fund as the duration is declining through the course of the retirement.
That's right, probably a long-term TIPS fund and a short-term TIPS fund (though any high-quality short-term bond fund would work for the second fund).
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

NiceUnparticularMan wrote: Fri Jun 11, 2021 10:56 am To be precise, I am noting that real world families do not have a simple relationship between marginal dollars spent and utility over time, such as transferring dollars, even in real terms, between one period and another period can make a significant difference in utility.
All you're describing is the fundamental concepts of risk aversion and loss aversion. These already affect portfolio construction in that they affect the allocation between the risky assets and the risk-free asset. Risk aversion doesn't change what the risk-free asset is for an investor. [/quote]
NiceUnparticularMan wrote: Fri Jun 11, 2021 10:56 am Well, to start, ideally I would have a lot more than $90 to save for this purpose, because surgery costs are not fixed in nominal terms, and in fact medical costs tend to go up more than CPI.

So, just funding the 5-year scenario is already in significant jeopardy of failing. Just for that scenario, I would want to put more than $100 in 5-year TIPS, more both because 5-year TIPS already have negative real returns just as to CPI (I think like -1.75% or so--yikes!), and then even more because of how medical costs tend to inflate.

And that is just for the 5-year scenario! Assuming I get past 5 years, I now have to also provide for 10 years. If I knew it was 10 years, I could just do the same sort of calculation, now using something like -0.85% instead and then accounting for medical cost inflating faster. Instead, I have to figure out what I have left in real dollars by that point after round 1. Then I will have to estimate reinvesting that in another 5-year TIPS in round 2. But at what rates? I don't know. And then I will have to do this all AGAIN to get to 15 years.
So your proposed answer to the question involves creating resources that the question specifically said weren't there and STILL ends up with a conclusion of "I don't know how I'd actually solve the problem".
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
NiceUnparticularMan
Posts: 2305
Joined: Sat Mar 11, 2017 7:51 am

Re: Long Bonds in retirement - curiosity...

Post by NiceUnparticularMan »

vineviz wrote: Fri Jun 11, 2021 1:20 pm
NiceUnparticularMan wrote: Fri Jun 11, 2021 10:56 am To be precise, I am noting that real world families do not have a simple relationship between marginal dollars spent and utility over time, such as transferring dollars, even in real terms, between one period and another period can make a significant difference in utility.
All you're describing is the fundamental concepts of risk aversion and loss aversion. These already affect portfolio construction in that they affect the allocation between the risky assets and the risk-free asset. Risk aversion doesn't change what the risk-free asset is for an investor.
Oh no, those are quite different concepts, although people sometimes try to explain some observed behavior with reference to more than one.

But definitionally, risk aversion is just a preference for low uncertainty outcomes to high uncertainty outcomes, and loss aversion is just a preference for avoiding losses over acquiring gains.

Now, it is true that if, say, you have a decreasing marginal utility with dollars in some particular circumstances, it might LOOK a lot like loss aversion. Meaning you will value gaining $X less than losing $X, given that assumption of how your utility curve looks within the region of =/- $X. But that isn't really loss aversion per se, once you translate that into utility rather than dollars.

But in fact the relationship between utility and spending in dollars can vary greatly depending on circumstances, so it definitely doesn't always work like that. Indeed, sometimes that means someone should be risk SEEKING, not risk averse. Sometimes it means you should value dollar gains more than equivalent dollar losses. And so on.

So it is very important not to confuse these concepts.

The point I am making is more like this. Spending on something like a somewhat nicer car typically has less utility than spending on a college education (for yourself or family members). So, if transferring $X from Time A to Time B means you can't afford to pay for college for yourself or a family member at Time A, and only allows you to buy a somewhat nicer car at Time B, that is probably a bad tradeoff in utility terms.

Because of that, we can't then take a wager that has a 100% chance of paying $X at Time A and $0 at Time B as the same as a wager that has a 0% chance of paying $X at Time A and 100% of paying $X at Time B. And that then feeds through a lot more "expected value" calculations that assume away this problem of there not being a consistent relationship between utility and $ at different times.
NiceUnparticularMan wrote: Fri Jun 11, 2021 10:56 am Well, to start, ideally I would have a lot more than $90 to save for this purpose, because surgery costs are not fixed in nominal terms, and in fact medical costs tend to go up more than CPI.

So, just funding the 5-year scenario is already in significant jeopardy of failing. Just for that scenario, I would want to put more than $100 in 5-year TIPS, more both because 5-year TIPS already have negative real returns just as to CPI (I think like -1.75% or so--yikes!), and then even more because of how medical costs tend to inflate.

And that is just for the 5-year scenario! Assuming I get past 5 years, I now have to also provide for 10 years. If I knew it was 10 years, I could just do the same sort of calculation, now using something like -0.85% instead and then accounting for medical cost inflating faster. Instead, I have to figure out what I have left in real dollars by that point after round 1. Then I will have to estimate reinvesting that in another 5-year TIPS in round 2. But at what rates? I don't know. And then I will have to do this all AGAIN to get to 15 years.
So your proposed answer to the question involves creating resources that the question specifically said weren't there and STILL ends up with a conclusion of "I don't know how I'd actually solve the problem".
I mean, my general thesis is these are often complex problems for real families to solve, particularly when they only have limited resources to work with and are facing unavoidable tradeoffs between different sorts of risks.

So, yes, I used this example to support that thesis. I did not propose a simple solution, because I don't think the real world versions of these problems actually have simple solutions.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

NiceUnparticularMan wrote: Fri Jun 11, 2021 1:37 pm Oh no, those are quite different concepts, although people sometimes try to explain some observed behavior with reference to more than one.
You're mistaken. Loss aversion and risk aversion simply describe the shape of the utility function.

NiceUnparticularMan wrote: Fri Jun 11, 2021 1:37 pm I mean, my general thesis is these are often complex problems for real families to solve, particularly when they only have limited resources to work with and are facing unavoidable tradeoffs between different sorts of risks.

So, yes, I used this example to support that thesis. I did not propose a simple solution, because I don't think the real world versions of these problems actually have simple solutions.
So the big insight here is that real life is complex and uncertain? And you have no suggestions for resolving that complexity and uncertainty other than to use the tools we were already using?

Awesome.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
NiceUnparticularMan
Posts: 2305
Joined: Sat Mar 11, 2017 7:51 am

Re: Long Bonds in retirement - curiosity...

Post by NiceUnparticularMan »

vineviz wrote: Fri Jun 11, 2021 2:13 pm
NiceUnparticularMan wrote: Fri Jun 11, 2021 1:37 pm Oh no, those are quite different concepts, although people sometimes try to explain some observed behavior with reference to more than one.
You're mistaken. Loss aversion and risk aversion simply describe the shape of the utility function.
Hmm. When I studied those concepts in graduate and professional school, and in various reading I have done since, that was not true. Perhaps that is in part because I was typically studying them from the perspective of what is sometimes known as behavioral economics, where it is not assumed people always behave as rational utility maximizers, such that people could be risk averse or loss averse for psychological reasons, and not because it reflected a rational choice.

In any event, as noted previously, it is definitely not true that the relevant utility curves of families should always be shaped the same way over the same relevant ranges of dollars at different times. So, even assuming you attribute all loss aversion and risk aversion to the shape of that curve, at various points with respect to various decisions families should be gain seeking and risk seeking instead.
NiceUnparticularMan wrote: Fri Jun 11, 2021 1:37 pm I mean, my general thesis is these are often complex problems for real families to solve, particularly when they only have limited resources to work with and are facing unavoidable tradeoffs between different sorts of risks.

So, yes, I used this example to support that thesis. I did not propose a simple solution, because I don't think the real world versions of these problems actually have simple solutions.
So the big insight here is that real life is complex and uncertain?
I didn't promise surprising revelations. The fact family financial lives are various, complicated, uncertain, and changing over time is indeed not much of a revelation, nor is the fact that means financial planning can involve a lot of uncertain tradeoffs and that different instruments and different strategies may make sense for different families at different times.
And you have no suggestions for resolving that complexity and uncertainty other than to use the tools we were already using?
If by "the tools we were already using" you mean all the different instruments and strategies being discussed here at Bogleheads, that is undoubtedly true--I have my thoughts on specific subjects which I share, but I wouldn't claim to have some special way of thinking about these difficult problems that is better than what everyone else here thinks.

So generally for help in financial planning, I think people are well-advised to post their circumstances and questions, add information as reasonably requested, and then carefully consider what may be a range of different insights and suggestions.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

NiceUnparticularMan wrote: Fri Jun 11, 2021 2:33 pm So generally for help in financial planning, I think people are well-advised to post their circumstances and questions, add information as reasonably requested, and then carefully consider what may be a range of different insights and suggestions.
So after all this back and forth we are right back where we started: if the question someone asks is “what should my average bond duration be” we have no answer better than “a duration which matches your expected investment time horizon”.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
NiceUnparticularMan
Posts: 2305
Joined: Sat Mar 11, 2017 7:51 am

Re: Long Bonds in retirement - curiosity...

Post by NiceUnparticularMan »

vineviz wrote: Fri Jun 11, 2021 2:47 pm
NiceUnparticularMan wrote: Fri Jun 11, 2021 2:33 pm So generally for help in financial planning, I think people are well-advised to post their circumstances and questions, add information as reasonably requested, and then carefully consider what may be a range of different insights and suggestions.
So after all this back and forth we are right back where we started: if the question someone asks is “what should my average bond duration be” we have no answer better than “a duration which matches your expected investment time horizon”.
I agree we are back to the beginning. And I stand by my original response, which is investors should account for uncertainty over their possible future spending needs, that they should think in terms of probability distributions, and that they should not substitute point estimates for probability distributions as doing so assumes away potentially important risks.

As a result, I think the best initial response to your hypothetical question ("what should my average bond duration be?") should usually be something like this:

"So let's take a step back. Bonds are special-purpose risk management tools, and different bonds have different risk management attributes at different costs, including opportunity costs. You should buy specific bonds, or sometimes perhaps bond funds, which are rationally calculated to address your risk management goals at an acceptable cost. And then whatever bonds you hold at any given time will have some average duration, but that is just going to be an observation that you can make at the end of the process of figuring out what various specific bonds or bond funds it actually makes sense for you to hold at that time."

And yes, my answer is way more complicated than yours, and won't be actionable until that investor does a lot more thinking.

Which I see as appropriate, not a criticism.
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

NiceUnparticularMan wrote: Fri Jun 11, 2021 3:02 pm "So let's take a step back. Bonds are special-purpose risk management tools, and different bonds have different risk management attributes at different costs, including opportunity costs. You should buy specific bonds, or sometimes perhaps bond funds, which are rationally calculated to address your risk management goals at an acceptable cost. And then whatever bonds you hold at any given time will have some average duration, but that is just going to be an observation that you can make at the end of the process of figuring out what various specific bonds or bond funds it actually makes sense for you to hold at that time."

And yes, my answer is way more complicated than yours, and won't be actionable until that investor does a lot more thinking.
Your answer isn't more complicated than mine: it just has more words and less actionable content.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
international001
Posts: 2017
Joined: Thu Feb 15, 2018 7:31 pm

Re: Long Bonds in retirement - curiosity...

Post by international001 »

vineviz wrote: Fri Jun 11, 2021 2:13 pm You're mistaken. Loss aversion and risk aversion simply describe the shape of the utility function.
Are you talking about this: https://en.wikipedia.org/wiki/Indifference_curve ?
Escapevelocity
Posts: 606
Joined: Mon Feb 18, 2019 8:32 am

Re: Long Bonds in retirement - curiosity...

Post by Escapevelocity »

vineviz wrote: Mon Jun 07, 2021 4:32 pm
Escapevelocity wrote: Mon Jun 07, 2021 10:47 am I agree with the premise, but here is where my self-professed market timing attitude on rates comes into play. With bonds, I refuse to accede to the nobody knows nothing motto.
The reason that market timing doesn't work isn't that "nobody knows nothing". It's that you don't know anything useful about the future that isn't already incorporated into prices by the market.
Respectfully disagree. I believe that the interest rate assumptions baked into the current bond market prices are very likely wrong by a wide margin. Even Jamie Dimon has $500M in cash which is surely a market timing situation:
https://www.marketwatch.com/story/dimon ... =home-page
User avatar
vineviz
Posts: 10262
Joined: Tue May 15, 2018 1:55 pm

Re: Long Bonds in retirement - curiosity...

Post by vineviz »

Escapevelocity wrote: Mon Jun 14, 2021 3:08 pm I believe that the interest rate assumptions baked into the current bond market prices are very likely wrong by a wide margin. Even Jamie Dimon has $500M in cash which is surely a market timing situation:
https://www.marketwatch.com/story/dimon ... =home-page
I suspect you are far too confident in your abilities. I know you are too confident in Dimon's.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Escapevelocity
Posts: 606
Joined: Mon Feb 18, 2019 8:32 am

Re: Long Bonds in retirement - curiosity...

Post by Escapevelocity »

vineviz wrote: Tue Jun 15, 2021 7:42 am
Escapevelocity wrote: Mon Jun 14, 2021 3:08 pm I believe that the interest rate assumptions baked into the current bond market prices are very likely wrong by a wide margin. Even Jamie Dimon has $500M in cash which is surely a market timing situation:
https://www.marketwatch.com/story/dimon ... =home-page
I suspect you are far too confident in your abilities. I know you are too confident in Dimon's.
Probably so, but at least I'm honest with myself that I'm a market timer with fixed income. Always have been.
international001
Posts: 2017
Joined: Thu Feb 15, 2018 7:31 pm

Re: Long Bonds in retirement - curiosity...

Post by international001 »

vineviz wrote: Tue Jun 08, 2021 8:11 pm
And "diversify over multiple yields" isn't an actual thing, at least not in the financial sense of the word "diversify".
Let me try an example. Assume your horizon is 5 years. You can choose to either have a 5 year bond or multiple 1 year bonds. In normal market, 5 year bond pays a premium yield (assume 0.5%).

Assume this sequence of 1 year and 5 year bond yields
1 year: 2.0% 4.0% 5.0% 1.0% 3.0%
5 year: 2.5% 4.5% 5.5% 1.5% 3.5%

If you buy a 5 year bond, you are locked to the 2.5% yield
If you buy multiple 1 year bonds, you diversify across the yields variations over the years

So 5 year bond is better in average, but there is a risk (more variation for the final outcome) depending on which year you start investing it.

Am I looking at it the right way?
Post Reply