How to use duration matching in a stock/bond portfolio

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Fri Aug 05, 2022 12:46 pm But this seems counterintuitive ... I'm always selling the longer, more volatile bonds.

Perhaps one of the duration matching experts could confirm what is supposed to happen.
I suspect it seems counterintuitive because of the way the conventional, total return approach to portfolio management tends to focus on price volatility as "risk" instead of focusing on uncertainty about consumption.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

Parkinglotracer wrote: Fri Aug 05, 2022 12:29 pm Duration implies taking time into account … obviously one has has to make some assumptions as to when one needs their money in order to accomplish this. As to what affect it will have on their portfolio performance no one knows because one can’t predict what interest rates will do or won’t do - but if you don’t duration match you are leaving yourself open to having to get your bond money right after interest rates boom and your bond or bond funds take a hit.

Many would agree usually it is not a big hit; this year with bnd down 8% it might brother you to sell your bond or bond fund.

Tsp G fund, i bonds, CDs, MYGAs, ladder of treasury securities are alternatives to use that will require a little more work than bnd with a 7-8 yr duration.
One point I was trying to make in the OP is that while I do, in fact, know when I need money from my portfolio (every single year!) with a mix of stocks and bonds, using a total return approach I do not know when I will sell bonds, or how many of them.

I might sell bonds from the portfolio for several years in a row. Or maybe, the stock market goes on a big big run and I don't sell any bonds for a year or two. Maybe I'm using a variable withdrawal approach (i.e. not 4%, adjusted for inflation every year) and I pull less money from the portfolio in year X versus year X - 1.

This year, with both stocks and bonds down, I'd absolutely be selling bonds (using almost any harvesting method I suspect) despite the fact said bonds are down.

Total return management is different from a bunch of bonds purchased to fund specific future liabilities, where amount and timing is precisely defined.
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Fri Aug 05, 2022 1:17 pm
One point I was trying to make in the OP is that while I do, in fact, know when I need money from my portfolio (every single year!) with a mix of stocks and bonds, using a total return approach I do not know when I will sell bonds, or how many of them.
If you are maintaining a relatively constant asset allocation and using rebalancing to do that (both things I think you mentioned in the OP), you still DO actually know the duration you would want to target (if you chose to do so).

In other words, your investment horizon doesn't change just because you are using a "total return approach".
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Fri Aug 05, 2022 1:11 pm
TN_Boy wrote: Fri Aug 05, 2022 12:46 pm But this seems counterintuitive ... I'm always selling the longer, more volatile bonds.

Perhaps one of the duration matching experts could confirm what is supposed to happen.
I suspect it seems counterintuitive because of the way the conventional, total return approach to portfolio management tends to focus on price volatility as "risk" instead of focusing on uncertainty about consumption.
A total return approach does focus on price volatility, and I think that is appropriate. Everybody has to understand that there is risk attempting to pull a constant amount of dollars (e.g. the 4% rule) from an inherently variable fund source like a stock/bond portfolio. Which is why people talk about variable withdrawal methods! But that is absolutely okay. People also understand that historically, you've been able to get more money out of a stock/bond portfolio than a bond-only portfolio.

I guess subconsciously I think in terms of a ladder, so the spending comes from matured bonds (the ones that obviously been getting lower and lower in duration) and always selling the long bonds seems .... wrong somehow. Especially should I retire into a time of rising interest rates and now I'm selling the bonds which have been hit the hardest in that scenario .... I bought high and sold low.
Last edited by TN_Boy on Fri Aug 05, 2022 1:30 pm, edited 1 time in total.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Fri Aug 05, 2022 1:21 pm
TN_Boy wrote: Fri Aug 05, 2022 1:17 pm
One point I was trying to make in the OP is that while I do, in fact, know when I need money from my portfolio (every single year!) with a mix of stocks and bonds, using a total return approach I do not know when I will sell bonds, or how many of them.
If you are maintaining a relatively constant asset allocation and using rebalancing to do that (both things I think you mentioned in the OP), you still DO actually know the duration you would want to target (if you chose to do so).

In other words, your investment horizon doesn't change just because you are using a "total return approach".
No the investment horizon doesn't change but I don't know exactly when I'll sell the bonds, which is different from an LMP portfolio. Whereupon I circle back to my question, how does duration matching make things better for me?
alluringreality
Posts: 787
Joined: Tue Nov 12, 2019 10:59 am

Re: How to use duration matching in a stock/bond portfolio

Post by alluringreality »

TN_Boy wrote: Wed Aug 03, 2022 1:18 pm How am I better off selecting a duration than simply using an intermediate term bond fund like BND?
My impression is that for many investors duration matching may generally tend to encourage longer-term bond holdings compared to using an intermediate term bond fund. If future rates are roughly flat or decrease, such encouragement might result in a higher return from bonds. If rates increase then you get into the breakeven consideration from Situation 1 in the following link.
https://www.fool.com/knowledge-center/h ... -on-b.aspx

For roughly the last 40 years investors in the USA have enjoyed limited inflation, and longer-term rates have generally fallen. It's difficult to say if the future might end up being more inflationary, which could result in higher rates. Part of my preference for savings bonds is that the redemption period is variable, and I'm not convinced that interest rate risk benefits my primary considerations. Since longer-term TIPS currently offer higher rates than Series I savings bonds, the idea of duration matching seems potentially worth considering to me. Ultimately the strategy that ends up paying more depends on an unknown future, which is fairly common with many investing decisions.
Last edited by alluringreality on Fri Aug 05, 2022 2:54 pm, edited 1 time in total.
30% Savings Bonds, 45% US Indexes, 25% Ex-US Indexes - Buy & Hold
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Fri Aug 05, 2022 1:27 pm A total return approach does focus on price volatility, and I think that is appropriate.
My point was mainly that focusing on price volatility is contrary to the underlying goals of retirement investors. This is why an adherence to total return approach sometimes leads to suboptimal decisions.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Fri Aug 05, 2022 1:28 pm No the investment horizon doesn't change but I don't know exactly when I'll sell the bonds, which is different from an LMP portfolio. Whereupon I circle back to my question, how does duration matching make things better for me?
The word "exactly" is doing all the work in this statement.

I repeat: even with a fixed asset allocation and periodic rebalancing, your investment horizon is not different from that of an otherwise identical investor using an LMP approach. Therefore, your average bond duration shouldn't be either (if you want to avoid taking on uncompensated interest rate risk).

Very few investors know - when they start their retirements - EXACTLY when they are going to die. That shouldn't, and usually doesn't, keep them from financial planning.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
rossington
Posts: 1137
Joined: Fri Jun 07, 2019 2:00 am
Location: Florida

Re: How to use duration matching in a stock/bond portfolio

Post by rossington »

TN_Boy wrote: Fri Aug 05, 2022 1:28 pm
vineviz wrote: Fri Aug 05, 2022 1:21 pm
TN_Boy wrote: Fri Aug 05, 2022 1:17 pm
One point I was trying to make in the OP is that while I do, in fact, know when I need money from my portfolio (every single year!) with a mix of stocks and bonds, using a total return approach I do not know when I will sell bonds, or how many of them.
If you are maintaining a relatively constant asset allocation and using rebalancing to do that (both things I think you mentioned in the OP), you still DO actually know the duration you would want to target (if you chose to do so).

In other words, your investment horizon doesn't change just because you are using a "total return approach".
No the investment horizon doesn't change but I don't know exactly when I'll sell the bonds, which is different from an LMP portfolio. Whereupon I circle back to my question, how does duration matching make things better for me?
The thing I am grappling with is if one is using some type of total return approach and does take a shot at duration matching the bonds then how do you make it all work? You're withdrawing and rebalancing between stocks and bonds each year. But if you decide to duration match the bonds annually then you're "forced" to sell from bonds no matter what to maintain the declining duration. So it seems that you are selling bonds every year. But what if you have to make a choice to either use the bond money to rebalance stocks back up or use it for spending, OR both at the same time, in which case it seems you could come up short either in spending or AA. Not to mention maintaining the proper allocation to the shorter bond fund. This could happen multiple times over the investment horizon.
"Success is going from failure to failure without loss of enthusiasm." Winston Churchill.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Fri Aug 05, 2022 2:18 pm
TN_Boy wrote: Fri Aug 05, 2022 1:28 pm No the investment horizon doesn't change but I don't know exactly when I'll sell the bonds, which is different from an LMP portfolio. Whereupon I circle back to my question, how does duration matching make things better for me?
The word "exactly" is doing all the work in this statement.

I repeat: even with a fixed asset allocation and periodic rebalancing, your investment horizon is not different from that of an otherwise identical investor using an LMP approach. Therefore, your average bond duration shouldn't be either (if you want to avoid taking on uncompensated interest rate risk).

Very few investors know - when they start their retirements - EXACTLY when they are going to die. That shouldn't, and usually doesn't, keep them from financial planning.
It's possible I'm confused about exactly what interest rate risk I'm not taking on via duration matching.

Longer bonds fluctuate in value more than shorter duration bonds, so when I sell bonds in early retirement, in a period of rising interest rates, I'm not going to be happy. It would appear that my interest rate risk depends mostly on whether rates rise or fall during retirement, and when they rise and fall.

But again, it seems likely I'm not understand the "interest rate" risk here that duration matching is mitigating.
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

rossington wrote: Fri Aug 05, 2022 3:28 pm The thing I am grappling with is if one is using some type of total return approach and does take a shot at duration matching the bonds then how do you make it all work? You're withdrawing and rebalancing between stocks and bonds each year. But if you decide to duration match the bonds annually then you're "forced" to sell from bonds no matter what to maintain the declining duration.
If you want to manage your interest rate risk, then at any point in time you have two separate decisions to make about your asset allocation.

A) What ratio of stocks and bonds do you want to hold?
B) What is the average duration of the bonds (if any) should you hold?

Duration matching doesn't necessarily mean you are "forced" to sell from bonds. Many people decide to do that, because of the funds they choose, but it is a decision (or maybe one consequence of a decision) and not something that is forced on them.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Fri Aug 05, 2022 4:24 pm
rossington wrote: Fri Aug 05, 2022 3:28 pm The thing I am grappling with is if one is using some type of total return approach and does take a shot at duration matching the bonds then how do you make it all work? You're withdrawing and rebalancing between stocks and bonds each year. But if you decide to duration match the bonds annually then you're "forced" to sell from bonds no matter what to maintain the declining duration.
If you want to manage your interest rate risk, then at any point in time you have two separate decisions to make about your asset allocation.

A) What ratio of stocks and bonds do you want to hold?
B) What is the average duration of the bonds (if any) should you hold?

Duration matching doesn't necessarily mean you are "forced" to sell from bonds. Many people decide to do that, because of the funds they choose, but it is a decision (or maybe one consequence of a decision) and not something that is forced on them.
A point I was trying to make is that with a total return oriented portfolio (very common) and many harvesting methods (i.e. what assets to sell for withdrawals) you will, in fact, have to sell bonds at a time when bonds are depressed. Or toss your harvesting method out the window. Your decision on whether to sell bonds is based upon more than just the bond side of the portfolio.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: How to use duration matching in a stock/bond portfolio

Post by willthrill81 »

TN_Boy wrote: Fri Aug 05, 2022 4:14 pm But again, it seems likely I'm not understand the "interest rate" risk here that duration matching is mitigating.
Have you read the post below of vineviz's in another thread discussing interest rate risk?

viewtopic.php?p=4755489#p4755489

Part of the problem with discussing interest rate risk is that most posters and even most 'professional' sites only discuss it in relation to the principal value of one's bonds, but that's only market price risk. It ignores reinvestment risk (i.e., the risk of future yields being lower than today's) altogether.
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Fri Aug 05, 2022 4:34 pm Your decision on whether to sell bonds is based upon more than just the bond side of the portfolio.
Right. That' s what I was trying to say.

The decision to sell bonds is based on other factors unrelated to the decision to duration-match or not.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Parkinglotracer
Posts: 1374
Joined: Fri Dec 20, 2019 3:49 am
Location: Upstate NY

Re: How to use duration matching in a stock/bond portfolio

Post by Parkinglotracer »

TN_Boy wrote: Fri Aug 05, 2022 1:17 pm
Parkinglotracer wrote: Fri Aug 05, 2022 12:29 pm Duration implies taking time into account … obviously one has has to make some assumptions as to when one needs their money in order to accomplish this. As to what affect it will have on their portfolio performance no one knows because one can’t predict what interest rates will do or won’t do - but if you don’t duration match you are leaving yourself open to having to get your bond money right after interest rates boom and your bond or bond funds take a hit.

Many would agree usually it is not a big hit; this year with bnd down 8% it might brother you to sell your bond or bond fund.

Tsp G fund, i bonds, CDs, MYGAs, ladder of treasury securities are alternatives to use that will require a little more work than bnd with a 7-8 yr duration.
One point I was trying to make in the OP is that while I do, in fact, know when I need money from my portfolio (every single year!) with a mix of stocks and bonds, using a total return approach I do not know when I will sell bonds, or how many of them.

I might sell bonds from the portfolio for several years in a row. Or maybe, the stock market goes on a big big run and I don't sell any bonds for a year or two. Maybe I'm using a variable withdrawal approach (i.e. not 4%, adjusted for inflation every year) and I pull less money from the portfolio in year X versus year X - 1.

This year, with both stocks and bonds down, I'd absolutely be selling bonds (using almost any harvesting method I suspect) despite the fact said bonds are down.

Total return management is different from a bunch of bonds purchased to fund specific future liabilities, where amount and timing is precisely defined.

That makes sense; we can’t predict everything - I get it. It is tough to manage. I have the 1/3 rd of my portfolio in fixed income all w/ a duration of three years or less as described previously. I have prob given up some return in past few years but I avoided most of the recent bond retreat as interest rates went up. I don’t have the answer but I don’t like to own long duration bonds / bond funds while rates are so low and one doesn’t seem to be compensated at the long end of the curve for the interest rate risk that can rear its head on occasion. Of course no one knows what rates will do. Bond funds are yielding more now that interest rates have gone up a bit.
alluringreality
Posts: 787
Joined: Tue Nov 12, 2019 10:59 am

Re: How to use duration matching in a stock/bond portfolio

Post by alluringreality »

One of the top Google results on duration matching links to the following series of five posts. The parts I've skimmed are generally consistent with my understanding. In the conclusion the point is made that duration matching aims at a balance between price risk and reinvestment risk, and if someone has a different intent a separate strategy may be appropriate.

https://occaminvesting.co.uk/duration-m ... roduction/
https://occaminvesting.co.uk/why-you-sh ... erm-bonds/
https://occaminvesting.co.uk/duration-m ... -practice/
https://occaminvesting.co.uk/against-duration-matching/
https://occaminvesting.co.uk/should-you ... -matching/
30% Savings Bonds, 45% US Indexes, 25% Ex-US Indexes - Buy & Hold
rossington
Posts: 1137
Joined: Fri Jun 07, 2019 2:00 am
Location: Florida

Re: How to use duration matching in a stock/bond portfolio

Post by rossington »

alluringreality wrote: Fri Aug 05, 2022 4:50 pm One of the top Google results on duration matching links to the following series of five posts. The parts I've skimmed are generally consistent with my understanding. In the conclusion the point is made that duration matching aims at a balance between price risk and reinvestment risk, and if someone has a different intent a separate strategy may be appropriate.

https://occaminvesting.co.uk/duration-m ... roduction/
https://occaminvesting.co.uk/why-you-sh ... erm-bonds/
https://occaminvesting.co.uk/duration-m ... -practice/
https://occaminvesting.co.uk/against-duration-matching/
https://occaminvesting.co.uk/should-you ... -matching/
Did you notice in #3 the reference to this: "The following examples are taken from this excellent Bogleheads thread": viewtopic.php?t=340252 ?
"Success is going from failure to failure without loss of enthusiasm." Winston Churchill.
rossington
Posts: 1137
Joined: Fri Jun 07, 2019 2:00 am
Location: Florida

Re: How to use duration matching in a stock/bond portfolio

Post by rossington »

vineviz wrote: Fri Aug 05, 2022 4:24 pm
rossington wrote: Fri Aug 05, 2022 3:28 pm The thing I am grappling with is if one is using some type of total return approach and does take a shot at duration matching the bonds then how do you make it all work? You're withdrawing and rebalancing between stocks and bonds each year. But if you decide to duration match the bonds annually then you're "forced" to sell from bonds no matter what to maintain the declining duration.
If you want to manage your interest rate risk, then at any point in time you have two separate decisions to make about your asset allocation.

A) What ratio of stocks and bonds do you want to hold?
B) What is the average duration of the bonds (if any) should you hold?

Duration matching doesn't necessarily mean you are "forced" to sell from bonds. Many people decide to do that, because of the funds they choose, but it is a decision (or maybe one consequence of a decision) and not something that is forced on them.
But...what else can one do besides selling the longer duration bonds each year to glide down to the shorter duration over the desired investment horizon? As I mentioned above this could potentially open up a can of investment worms.
"Success is going from failure to failure without loss of enthusiasm." Winston Churchill.
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

rossington wrote: Sat Aug 06, 2022 3:37 am But...what else can one do besides selling the longer duration bonds each year to glide down to the shorter duration over the desired investment horizon? As I mentioned above this could potentially open up a can of investment worms.
I don't know what "can of investment worms" you think it opens, but there are options: target maturity bond funds, non-rolling bond ladders, income annuities, etc.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
dbr
Posts: 41845
Joined: Sun Mar 04, 2007 9:50 am

Re: How to use duration matching in a stock/bond portfolio

Post by dbr »

vineviz wrote: Sat Aug 06, 2022 7:48 am
rossington wrote: Sat Aug 06, 2022 3:37 am But...what else can one do besides selling the longer duration bonds each year to glide down to the shorter duration over the desired investment horizon? As I mentioned above this could potentially open up a can of investment worms.
I don't know what "can of investment worms" you think it opens, but there are options: target maturity bond funds, non-rolling bond ladders, income annuities, etc.
In practice those kinds of things may be too much overlooked. But if one goes back to an old model of Social Security and pensions, that is exactly what people had. I am still impressed that years ago the sophisticated Financial Engines planning tool did no retirement investment planning because it assumed one used the assets to buy an inflation indexed annuity. Am I just imagining that or was it really so?
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

willthrill81 wrote: Fri Aug 05, 2022 4:42 pm
TN_Boy wrote: Fri Aug 05, 2022 4:14 pm But again, it seems likely I'm not understand the "interest rate" risk here that duration matching is mitigating.
Have you read the post below of vineviz's in another thread discussing interest rate risk?

viewtopic.php?p=4755489#p4755489

Part of the problem with discussing interest rate risk is that most posters and even most 'professional' sites only discuss it in relation to the principal value of one's bonds, but that's only market price risk. It ignores reinvestment risk (i.e., the risk of future yields being lower than today's) altogether.
But I care about the principal value of my bonds, and I think you should too, if you are working with the scenario described in the OP. When harvesting from the investment portfolio, I am selling a non-fixed $$ amount of bonds on a non-fixed schedule. I do not care about reinvestment risk directly. I care about the total return of the bond portion of the portfolio and its correlation with the stock portion of the portfolio. If you can show me that longer duration bonds achieve those goals then sure, that makes sense. But I confess to not seeing that argument made.

There is also the risk of future yields being higher than today's yields, in which case buying longer bonds is not an obvious win.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Sat Aug 06, 2022 7:48 am
rossington wrote: Sat Aug 06, 2022 3:37 am But...what else can one do besides selling the longer duration bonds each year to glide down to the shorter duration over the desired investment horizon? As I mentioned above this could potentially open up a can of investment worms.
I don't know what "can of investment worms" you think it opens, but there are options: target maturity bond funds, non-rolling bond ladders, income annuities, etc.
But the question on the table is how duration matching is used with the total return portfolio and rebalancing described in the OP.

Income annuities are not relevant to the OP scenario, so help me out on what to do with the target maturity bond fund or non-rolling bond ladder, where I'm selling a variable amount of the bonds each year.

How is my result likely to differ in those cases versus simply having an intermediate bond fund?
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: How to use duration matching in a stock/bond portfolio

Post by willthrill81 »

TN_Boy wrote: Sat Aug 06, 2022 9:08 am
willthrill81 wrote: Fri Aug 05, 2022 4:42 pm
TN_Boy wrote: Fri Aug 05, 2022 4:14 pm But again, it seems likely I'm not understand the "interest rate" risk here that duration matching is mitigating.
Have you read the post below of vineviz's in another thread discussing interest rate risk?

viewtopic.php?p=4755489#p4755489

Part of the problem with discussing interest rate risk is that most posters and even most 'professional' sites only discuss it in relation to the principal value of one's bonds, but that's only market price risk. It ignores reinvestment risk (i.e., the risk of future yields being lower than today's) altogether.
But I care about the principal value of my bonds, and I think you should too, if you are working with the scenario described in the OP.
Principal value is certainly meaningful to all fixed income investors, but the future yield on that principal is meaningful too.
TN_Boy wrote: Sat Aug 06, 2022 9:08 am When harvesting from the investment portfolio, I am selling a non-fixed $$ amount of bonds on a non-fixed schedule. I do not care about reinvestment risk directly. I care about the total return of the bond portion of the portfolio and its correlation with the stock portion of the portfolio. If you can show me that longer duration bonds achieve those goals then sure, that makes sense. But I confess to not seeing that argument made.
This is an interesting point, and I too would like to see an analysis of it.
TN_Boy wrote: Sat Aug 06, 2022 9:08 am There is also the risk of future yields being higher than today's yields, in which case buying longer bonds is not an obvious win.
Of course. Something we refer to as a 'risk' could turn out to help. This includes things like interest rate risk, currency risk, sequence of returns risk, etc. But the point is that the presence of such a risk means that it could hurt us.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

willthrill81 wrote: Sat Aug 06, 2022 9:22 am
TN_Boy wrote: Sat Aug 06, 2022 9:08 am
willthrill81 wrote: Fri Aug 05, 2022 4:42 pm
TN_Boy wrote: Fri Aug 05, 2022 4:14 pm But again, it seems likely I'm not understand the "interest rate" risk here that duration matching is mitigating.
Have you read the post below of vineviz's in another thread discussing interest rate risk?

viewtopic.php?p=4755489#p4755489

Part of the problem with discussing interest rate risk is that most posters and even most 'professional' sites only discuss it in relation to the principal value of one's bonds, but that's only market price risk. It ignores reinvestment risk (i.e., the risk of future yields being lower than today's) altogether.
But I care about the principal value of my bonds, and I think you should too, if you are working with the scenario described in the OP.
Principal value is certainly meaningful to all fixed income investors, but the future yield on that principal is meaningful too.
TN_Boy wrote: Sat Aug 06, 2022 9:08 am When harvesting from the investment portfolio, I am selling a non-fixed $$ amount of bonds on a non-fixed schedule. I do not care about reinvestment risk directly. I care about the total return of the bond portion of the portfolio and its correlation with the stock portion of the portfolio. If you can show me that longer duration bonds achieve those goals then sure, that makes sense. But I confess to not seeing that argument made.
This is an interesting point, and I too would like to see an analysis of it.
TN_Boy wrote: Sat Aug 06, 2022 9:08 am There is also the risk of future yields being higher than today's yields, in which case buying longer bonds is not an obvious win.
Of course. Something we refer to as a 'risk' could turn out to help. This includes things like interest rate risk, currency risk, sequence of returns risk, etc. But the point is that the presence of such a risk means that it could hurt us.
Right, but I'm saying that if you don't know which way interest rates will trend over your investment horizon, that buying longer bonds is as likely to hurt as help.

There are various back-tests that can (and have been) run comparing portfolio survival with different duration bonds, though I believe those studies are holding bond duration constant, versus decreasing to match retirement horizon. And longer treasuries tend to do better, because they usually have a lower correlation with stocks. Except they don't do better in a period of rising interest rates .... so if you go with duration matching in a time of rising rates, and are selling the longer bonds for portfolio withdrawals early, then you wind up, as time goes on, holding shorter duration bonds now that interest rates are higher. I.e. the worst of all possible worlds.

You just don't know and the simple argument that duration matching is good even when bond sales are potentially irregular in both time and amount is just not getting through to my simple brain.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: How to use duration matching in a stock/bond portfolio

Post by willthrill81 »

TN_Boy wrote: Sat Aug 06, 2022 9:44 am Right, but I'm saying that if you don't know which way interest rates will trend over your investment horizon, that buying longer bonds is as likely to hurt as help.
Again, that's only true of market price risk. It's not true of reinvestment risk.
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Sat Aug 06, 2022 9:12 am
vineviz wrote: Sat Aug 06, 2022 7:48 am
rossington wrote: Sat Aug 06, 2022 3:37 am But...what else can one do besides selling the longer duration bonds each year to glide down to the shorter duration over the desired investment horizon? As I mentioned above this could potentially open up a can of investment worms.
I don't know what "can of investment worms" you think it opens, but there are options: target maturity bond funds, non-rolling bond ladders, income annuities, etc.
But the question on the table is how duration matching is used with the total return portfolio and rebalancing described in the OP.
And I think that question has been answered already.

There is nothing different about duration matching in a "total return" portfolio versus any other type of portfolio. The circumstances described in the OP don't change the process or the target.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Sat Aug 06, 2022 9:44 am You just don't know and the simple argument that duration matching is good even when bond sales are potentially irregular in both time and amount is just not getting through to my simple brain.
I'm not sure anyone is suggesting (or has suggested) that duration matching is "good". It's simply a tool that solves a problem (interest rate risk).

An electric drill isn't "good" or "bad", but if you need to create a bunch of holes then it sure can be handy to know how to use one.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

willthrill81 wrote: Sat Aug 06, 2022 9:48 am
TN_Boy wrote: Sat Aug 06, 2022 9:44 am Right, but I'm saying that if you don't know which way interest rates will trend over your investment horizon, that buying longer bonds is as likely to hurt as help.
Again, that's only true of market price risk. It's not true of reinvestment risk.
Yes. Are you saying that reinvestment risk is more important/critical than market price risk in the OP scenario?
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Sat Aug 06, 2022 9:54 am
TN_Boy wrote: Sat Aug 06, 2022 9:44 am You just don't know and the simple argument that duration matching is good even when bond sales are potentially irregular in both time and amount is just not getting through to my simple brain.
I'm not sure anyone is suggesting (or has suggested) that duration matching is "good". It's simply a tool that solves a problem (interest rate risk).

An electric drill isn't "good" or "bad", but if you need to create a bunch of holes then it sure can be handy to know how to use one.
Wait, no, you have asserted that duration matching is good. Here is what you said early in the thread:
I'd argue that the primary benefit of duration matching is one of risk reduction. Duration matching provides more upfront certainty about the level of portfolio duration, but the truth is that an investor is set on rebalancing and has a significant equity exposure (e.g. 50% in your example) is giving up some of the certainty anyway. For such a retiree, the difference might not be very great.
Risk reduction is a good thing. Though you observe that in the OP scenario, maybe it is not a large effect.

I got nothing against duration matching. Whether it is valuable or "good" in the OP scenario is my question. There are plenty of scenarios where I can see it makes perfect sense.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: How to use duration matching in a stock/bond portfolio

Post by willthrill81 »

TN_Boy wrote: Sat Aug 06, 2022 9:57 am
willthrill81 wrote: Sat Aug 06, 2022 9:48 am
TN_Boy wrote: Sat Aug 06, 2022 9:44 am Right, but I'm saying that if you don't know which way interest rates will trend over your investment horizon, that buying longer bonds is as likely to hurt as help.
Again, that's only true of market price risk. It's not true of reinvestment risk.
Yes. Are you saying that reinvestment risk is more important/critical than market price risk in the OP scenario?
For a long-term investor, yes.

I'm quoting vineviz from one of the threads I linked to above.
vineviz wrote: Wed Sep 18, 2019 2:46 pm This is because interest rate risk actually has two components. One of them is price risk (the effect that changes in interest rates have on the price of assets) and the the other one is reinvestment risk (the effect that changes in interest rates have on rate at which income can be reinvested).

Each component works in an opposing direction, and there is a length of time at which the two precisely balance out to zero. This point is defined as the Macualey duration.

Image
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Sat Aug 06, 2022 10:01 am Risk reduction is a good thing.
Those are your words, not mine.

I certainly don't believe that risk reduction is always good. But when it is, I want to know how to do it.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: How to use duration matching in a stock/bond portfolio

Post by willthrill81 »

vineviz wrote: Sat Aug 06, 2022 9:49 am
TN_Boy wrote: Sat Aug 06, 2022 9:12 am
vineviz wrote: Sat Aug 06, 2022 7:48 am
rossington wrote: Sat Aug 06, 2022 3:37 am But...what else can one do besides selling the longer duration bonds each year to glide down to the shorter duration over the desired investment horizon? As I mentioned above this could potentially open up a can of investment worms.
I don't know what "can of investment worms" you think it opens, but there are options: target maturity bond funds, non-rolling bond ladders, income annuities, etc.
But the question on the table is how duration matching is used with the total return portfolio and rebalancing described in the OP.
And I think that question has been answered already.

There is nothing different about duration matching in a "total return" portfolio versus any other type of portfolio. The circumstances described in the OP don't change the process or the target.
I believe that a demonstration with real world data of how duration matching in a rebalanced portfolio of stocks and bonds reduced interest rate risk would be very helpful.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

willthrill81 wrote: Sat Aug 06, 2022 10:11 am
vineviz wrote: Sat Aug 06, 2022 9:49 am
TN_Boy wrote: Sat Aug 06, 2022 9:12 am
vineviz wrote: Sat Aug 06, 2022 7:48 am
rossington wrote: Sat Aug 06, 2022 3:37 am But...what else can one do besides selling the longer duration bonds each year to glide down to the shorter duration over the desired investment horizon? As I mentioned above this could potentially open up a can of investment worms.
I don't know what "can of investment worms" you think it opens, but there are options: target maturity bond funds, non-rolling bond ladders, income annuities, etc.
But the question on the table is how duration matching is used with the total return portfolio and rebalancing described in the OP.
And I think that question has been answered already.

There is nothing different about duration matching in a "total return" portfolio versus any other type of portfolio. The circumstances described in the OP don't change the process or the target.
I believe that a demonstration with real world data of how duration matching in a rebalanced portfolio of stocks and bonds reduced interest rate risk would be very helpful.
I believe I am not the only person wishing for such a demonstration.

But I would tweak that request. A total return investor cares about the portfolio. So the demonstration would be how the portfolio risk is reduced or return is enhanced. I believe BHers are constantly telling investors to focus on the portfolio behavior, right?
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: How to use duration matching in a stock/bond portfolio

Post by willthrill81 »

TN_Boy wrote: Sat Aug 06, 2022 10:24 am But I would tweak that request. A total return investor cares about the portfolio. So the demonstration would be how the portfolio risk is reduced or return is enhanced. I believe BHers are constantly telling investors to focus on the portfolio behavior, right?
Almost eliminating the interest rate risk of one's fixed income holdings should, all else held equal, reduce portfolio risk.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Sat Aug 06, 2022 10:06 am
TN_Boy wrote: Sat Aug 06, 2022 10:01 am Risk reduction is a good thing.
Those are your words, not mine.

I certainly don't believe that risk reduction is always good. But when it is, I want to know how to do it.
I want to know if duration matching in a stock/bond portfolio as described in the OP results in portfolio risk reduction, or perhaps enhanced returns with similar risk. If it doesn't then I cannot understand why I should be interested in it for that scenario. That's the point of this thread.

Perhaps the title is misleading - it says "how to use" and you have provided an answer. But that answer does not include an explanation for how the portfolio is better off. Just that duration now matches the investing horizon. But I do not feel that you have shown that said matching is a good portfolio thing.

If you think so, and that I'm simply not following, fine, you can leave it at that. I don't want to waste your time.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

willthrill81 wrote: Sat Aug 06, 2022 10:26 am
TN_Boy wrote: Sat Aug 06, 2022 10:24 am But I would tweak that request. A total return investor cares about the portfolio. So the demonstration would be how the portfolio risk is reduced or return is enhanced. I believe BHers are constantly telling investors to focus on the portfolio behavior, right?
Almost eliminating the interest rate risk of one's fixed income holdings should, all else held equal, reduce portfolio risk.
Perhaps my refusal to grasp that reinvestment risk is a lot more important than risk to principal is holding me back ... that and believing that reducing the interest rate "risk" is only good when rates are steady or falling.
dbr
Posts: 41845
Joined: Sun Mar 04, 2007 9:50 am

Re: How to use duration matching in a stock/bond portfolio

Post by dbr »

TN_Boy wrote: Sat Aug 06, 2022 10:36 am
willthrill81 wrote: Sat Aug 06, 2022 10:26 am
TN_Boy wrote: Sat Aug 06, 2022 10:24 am But I would tweak that request. A total return investor cares about the portfolio. So the demonstration would be how the portfolio risk is reduced or return is enhanced. I believe BHers are constantly telling investors to focus on the portfolio behavior, right?
Almost eliminating the interest rate risk of one's fixed income holdings should, all else held equal, reduce portfolio risk.
Perhaps my refusal to grasp that reinvestment risk is a lot more important than risk to principal is holding me back ... that and believing that reducing the interest rate "risk" is only good when rates are steady or falling.
Well, an example of reinvestment risk is people relying in CDs at 5% for income and finding they can only renew at 1%. Yet CDs are supposed to be risk free. But if volatility of return is the definition of risk then a change from 5% to 1% is pretty volatile if that 5% is what one is living on. That is one example how short fixed income can be really dangerous. If those investors had bought long bonds for a long retirement things would have worked better. It also would not be dangerous to just withdraw from the money saved and let the interest be gravy. But that takes huge wealth relative to spending, especially allowing for inflation.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

dbr wrote: Sat Aug 06, 2022 10:51 am
TN_Boy wrote: Sat Aug 06, 2022 10:36 am
willthrill81 wrote: Sat Aug 06, 2022 10:26 am
TN_Boy wrote: Sat Aug 06, 2022 10:24 am But I would tweak that request. A total return investor cares about the portfolio. So the demonstration would be how the portfolio risk is reduced or return is enhanced. I believe BHers are constantly telling investors to focus on the portfolio behavior, right?
Almost eliminating the interest rate risk of one's fixed income holdings should, all else held equal, reduce portfolio risk.
Perhaps my refusal to grasp that reinvestment risk is a lot more important than risk to principal is holding me back ... that and believing that reducing the interest rate "risk" is only good when rates are steady or falling.
Well, an example of reinvestment risk is people relying in CDs at 5% for income and finding they can only renew at 1%. Yet CDs are supposed to be risk free. But if volatility of return is the definition of risk then a change from 5% to 1% is pretty volatile if that 5% is what one is living on. That is one example how short fixed income can be really dangerous. If those investors had bought long bonds for a long retirement things would have worked better. It also would not be dangerous to just withdraw from the money saved and let the interest be gravy. But that takes huge wealth relative to spending, especially allowing for inflation.
Buying long is absolutely better if rates hold steady or decline. And if you are happy with current rates then yes, go long and stay there. Especially if you can let the fixed income mature and not sell at arbitrary times.

I still think the point rossington brought up is important. I have my duration matching bond funds -- one long, one short, the weighted average duration being say 15 years. When I sell bonds from the portfolio, I will be selling the longer duration bonds, as I reduce the duration (or as I rebalance). Such a portfolio seems to me to be extremely vulnerable to rising interest rates early in retirement, as I sell the longer bonds, quite possibly at a big loss, and wind up with shorter bonds after a period of rising rates. I don't want a portfolio that is highly vulnerable to rising interest rates, though you are probably going to tell me it's not really more vulnerable to rising rates :-)
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

willthrill81 wrote: Sat Aug 06, 2022 10:11 am I believe that a demonstration with real world data of how duration matching in a rebalanced portfolio of stocks and bonds reduced interest rate risk would be very helpful.
It's hard to illustrate the reduction in risk given the conditions established in the OP.

I'm afraid that the key point about duration matching, which is the goal is to reduce uncertainty of consumption rather than to increase returns, might get lost in the discussion, but here's one example.

It starts with a 1992 retiree, and compares two investors with a 50% stock and 50% bond allocation (with rebalancing). One maintains their bond allocation as a steady holding with a roughly 4 year duration, the other employs duration-matching (starting with an average duration of about 15 years and steadily declining to an average duration of 0 years). Withdrawals are taken at the start of each year and are adjusted for inflations.

Image

This is a retirement cohort that had the opportunity to "lock in" Treasury yields of 7 to 8% (and therefore a SWR of about 6.2% on their bond portfolio) through duration matching. A choice to play it "safe" by holding total bond market or a four-year duration portfolio of Treasuries cost them about 100 bps in SWR on their bond holdings). This is, as you have said before, an example of the risk showing up.

Obviously the 50% or 60% allocated to stocks would have mitigated that advantage. Also, obviously, it's a risk and so the bet might have turned out differently.

Some people in 1992 might have preferred a certain 6% withdrawal rate to a chance of getting something between 4% and 8%, whereas others might have preferred the riskier bet. Duration matching is a tool for the former.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Sat Aug 06, 2022 1:06 pm
willthrill81 wrote: Sat Aug 06, 2022 10:11 am I believe that a demonstration with real world data of how duration matching in a rebalanced portfolio of stocks and bonds reduced interest rate risk would be very helpful.
It's hard to illustrate the reduction in risk given the conditions established in the OP.

I'm afraid that the key point about duration matching, which is the goal is to reduce uncertainty of consumption rather than to increase returns, might get lost in the discussion, but here's one example.

It starts with a 1992 retiree, and compares two investors with a 50% stock and 50% bond allocation (with rebalancing). One maintains their bond allocation as a steady holding with a roughly 4 year duration, the other employs duration-matching (starting with an average duration of about 15 years and steadily declining to an average duration of 0 years). Withdrawals are taken at the start of each year and are adjusted for inflations.

Image

This is a retirement cohort that had the opportunity to "lock in" Treasury yields of 7 to 8% (and therefore a SWR of about 6.2% on their bond portfolio) through duration matching. A choice to play it "safe" by holding total bond market or a four-year duration portfolio of Treasuries cost them about 100 bps in SWR on their bond holdings). This is, as you have said before, an example of the risk showing up.

Obviously the 50% or 60% allocated to stocks would have mitigated that advantage. Also, obviously, it's a risk and so the bet might have turned out differently.

Some people in 1992 might have preferred a certain 6% withdrawal rate to a chance of getting something between 4% and 8%, whereas others might have preferred the riskier bet. Duration matching is a tool for the former.
vineviz,

Thanks for doing that. The difference in ending portfolio value is striking.

That said, I had a couple of obvious comments/questions about this example:

1) It's a period of falling interest rates. Certainly going long is better and that was predictable.
2) Related to 1), how would a *constant* bond duration of 15 years fared versus decreasing duration?
3) Four years is rather short. I'd think comparing to something like six or seven years would be better.
4) How about a comparison during a period of rising rates?

Is there a straightforward way for us to recreate this type of scenario (and varying the different variables)? Rather than asking you to do work? Can you fine-tune portfolio visualizer to do this (specifically the decreasing duration?).

Also, I'm not trying to be argumentative, but with respect to this:
It's hard to illustrate the reduction in risk given the conditions established in the OP.
I believe the OP conditions are the default for most retirees. We can quibble over harvesting methods, and even fixed versus variable withdrawals but basically, people have income streams from pensions and SS, and they supplement that with withdrawals from the portfolio that is something like what is described. It's not some off-the-wall scenario.

(Hmm. Here's a thought. Some posters with large portfolios say they will live off equity dividends and bond interest only. It seems to me that such an approach would be amenable to duration matching as I understand it. Am I right? Or, since they are not selling anything, should they just go long and stay long with the bonds?)
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

TN_Boy wrote: Sat Aug 06, 2022 1:34 pm 1) It's a period of falling interest rates. Certainly going long is better and that was predictable.
It wasn't predictable in 1992, if that's what you mean. Market timing with bonds isn't any easier than market timing with stocks.
TN_Boy wrote: Sat Aug 06, 2022 1:34 pm 2) Related to 1), how would a *constant* bond duration of 15 years fared versus decreasing duration?
A lot better, clearly. But keeping duration>investment horizon would have been just as risky as keeping duration<investment horizon.
TN_Boy wrote: Sat Aug 06, 2022 1:34 pm 3) Four years is rather short. I'd think comparing to something like six or seven years would be better.
Maybe, but four years was approximately the duration of Total Bond Market in 1992. And many investors keep significant amounts of cash (HYSA, T-bills, TSP G fund, stable value, CDs) in their portfolios along with TBM. I think it's a fair comparison, but maybe not ....
TN_Boy wrote: Sat Aug 06, 2022 1:34 pm
4) How about a comparison during a period of rising rates?
A sustained period of increasing bond yields would certainly make duration<investment horizon the "winner".

But again, the purpose of duration matching isn't to produce a "winner" in terms of market timing.

The purpose is to eliminate the risk so that you can more accurately predict your retirement income.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Sat Aug 06, 2022 1:49 pm
TN_Boy wrote: Sat Aug 06, 2022 1:34 pm 1) It's a period of falling interest rates. Certainly going long is better and that was predictable.
It wasn't predictable in 1992, if that's what you mean. Market timing with bonds isn't any easier than market timing with stocks.
TN_Boy wrote: Sat Aug 06, 2022 1:34 pm 2) Related to 1), how would a *constant* bond duration of 15 years fared versus decreasing duration?
A lot better, clearly. But keeping duration>investment horizon would have been just as risky as keeping duration<investment horizon.
TN_Boy wrote: Sat Aug 06, 2022 1:34 pm 3) Four years is rather short. I'd think comparing to something like six or seven years would be better.
Maybe, but four years was approximately the duration of Total Bond Market in 1992. And many investors keep significant amounts of cash (HYSA, T-bills, TSP G fund, stable value, CDs) in their portfolios along with TBM. I think it's a fair comparison, but maybe not ....
TN_Boy wrote: Sat Aug 06, 2022 1:34 pm
4) How about a comparison during a period of rising rates?
A sustained period of increasing bond yields would certainly make duration<investment horizon the "winner".

But again, the purpose of duration matching isn't to produce a "winner" in terms of market timing.

The purpose is to eliminate the risk so that you can more accurately predict your retirement income.
1) I just meant that knowing now interest rates fell during that time period, I would have predicted that longer bonds did well.
2) This is actually a slightly different point, but it's not clear to me why my risk is so much higher keeping the bond duration longer near the end of my estimated retirement horizon. I still think portfolio, not just the bonds. And a detail, I can assume a 30 year horizon starting at age 60, but if I get there and am still around at 30 years my horizon could still be 5 years or more. Probably not, but certainly possible.
3) Fair point on BND being around 4 years in 92, which I didn't know. I would have compared longer treasuries with intermediate treasuries to keep the credit risk identical and probably used more like six years. At least five.
4) Right, so if bond rates rise with longer duration bonds I do worse. If the converse is true I do better. And I can't predict interest rates. Which is why a lot of people obviously go with intermediate duration.

I will look more at portfolio visualizer and see if there is a way to see at how different scenarios played out using that tool. Is that what you used? There is a dynamic allocation tab.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Sat Aug 06, 2022 1:06 pm
willthrill81 wrote: Sat Aug 06, 2022 10:11 am I believe that a demonstration with real world data of how duration matching in a rebalanced portfolio of stocks and bonds reduced interest rate risk would be very helpful.
It's hard to illustrate the reduction in risk given the conditions established in the OP.

I'm afraid that the key point about duration matching, which is the goal is to reduce uncertainty of consumption rather than to increase returns, might get lost in the discussion, but here's one example.

It starts with a 1992 retiree, and compares two investors with a 50% stock and 50% bond allocation (with rebalancing). One maintains their bond allocation as a steady holding with a roughly 4 year duration, the other employs duration-matching (starting with an average duration of about 15 years and steadily declining to an average duration of 0 years). Withdrawals are taken at the start of each year and are adjusted for inflations.

Image

This is a retirement cohort that had the opportunity to "lock in" Treasury yields of 7 to 8% (and therefore a SWR of about 6.2% on their bond portfolio) through duration matching. A choice to play it "safe" by holding total bond market or a four-year duration portfolio of Treasuries cost them about 100 bps in SWR on their bond holdings). This is, as you have said before, an example of the risk showing up.

Obviously the 50% or 60% allocated to stocks would have mitigated that advantage. Also, obviously, it's a risk and so the bet might have turned out differently.

Some people in 1992 might have preferred a certain 6% withdrawal rate to a chance of getting something between 4% and 8%, whereas others might have preferred the riskier bet. Duration matching is a tool for the former.
Vineviz,

Having a little trouble duplicating some of your results. I'm starting by using portfolio visualizer and static allocations. There appears to be a way to specify a shifting allocation (i.e. for the duration matching bond test) but I wanted to start with the simple case.

So my test is 1992 retiree, portfolio through end of 2021, with 50% in Vanguard's VFINX (S&P 500) and then either short, intermediate, or long term treasury funds. As expected the longer the duration, the better the portfolio did. Bond durations held steady during the test. An inflation adjusted 4% is withdrawn every year.

But the inflation adjusted portfolio end values don't match yours. All three were up, in inflation adjusted terms. Here is the link:

https://www.portfoliovisualizer.com/bac ... tion4_1=50

You have to click the button on the portfolio growth chart to get inflation adjusted results. I mention as since I want to duplicate your results and try some other combinations, I'm immediately hitting a difference. Your graph shows both investors have less in real terms than they started with, while I see all three investors having more real dollars.

After playing a bit more with this, I'll try the dynamic allocation tab, feeding it a changing mix of bond durations (you specify with a csv file). Vineviz (or anybody else) let me know if I've mishandled portfolio visualizer.
rossington
Posts: 1137
Joined: Fri Jun 07, 2019 2:00 am
Location: Florida

Re: How to use duration matching in a stock/bond portfolio

Post by rossington »

TN_Boy wrote: Sun Aug 07, 2022 12:15 pm
vineviz wrote: Sat Aug 06, 2022 1:06 pm
willthrill81 wrote: Sat Aug 06, 2022 10:11 am I believe that a demonstration with real world data of how duration matching in a rebalanced portfolio of stocks and bonds reduced interest rate risk would be very helpful.
It's hard to illustrate the reduction in risk given the conditions established in the OP.

I'm afraid that the key point about duration matching, which is the goal is to reduce uncertainty of consumption rather than to increase returns, might get lost in the discussion, but here's one example.

It starts with a 1992 retiree, and compares two investors with a 50% stock and 50% bond allocation (with rebalancing). One maintains their bond allocation as a steady holding with a roughly 4 year duration, the other employs duration-matching (starting with an average duration of about 15 years and steadily declining to an average duration of 0 years). Withdrawals are taken at the start of each year and are adjusted for inflations.

Image

This is a retirement cohort that had the opportunity to "lock in" Treasury yields of 7 to 8% (and therefore a SWR of about 6.2% on their bond portfolio) through duration matching. A choice to play it "safe" by holding total bond market or a four-year duration portfolio of Treasuries cost them about 100 bps in SWR on their bond holdings). This is, as you have said before, an example of the risk showing up.

Obviously the 50% or 60% allocated to stocks would have mitigated that advantage. Also, obviously, it's a risk and so the bet might have turned out differently.

Some people in 1992 might have preferred a certain 6% withdrawal rate to a chance of getting something between 4% and 8%, whereas others might have preferred the riskier bet. Duration matching is a tool for the former.
Vineviz,

Having a little trouble duplicating some of your results. I'm starting by using portfolio visualizer and static allocations. There appears to be a way to specify a shifting allocation (i.e. for the duration matching bond test) but I wanted to start with the simple case.

So my test is 1992 retiree, portfolio through end of 2021, with 50% in Vanguard's VFINX (S&P 500) and then either short, intermediate, or long term treasury funds. As expected the longer the duration, the better the portfolio did. Bond durations held steady during the test. An inflation adjusted 4% is withdrawn every year.

But the inflation adjusted portfolio end values don't match yours. All three were up, in inflation adjusted terms. Here is the link:

https://www.portfoliovisualizer.com/bac ... tion4_1=50

You have to click the button on the portfolio growth chart to get inflation adjusted results. I mention as since I want to duplicate your results and try some other combinations, I'm immediately hitting a difference. Your graph shows both investors have less in real terms than they started with, while I see all three investors having more real dollars.

After playing a bit more with this, I'll try the dynamic allocation tab, feeding it a changing mix of bond durations (you specify with a csv file). Vineviz (or anybody else) let me know if I've mishandled portfolio visualizer.
It looks like you would need 2 separate PV screens to do this. One to combine VFINX with a bond fund of a constant 4% duration. And a separate screen to combine VFINX with TWO bond funds (one long and one short(er) that are matched to have an average duration of 15 years. So you would need to calculate the percentages to allocate to each fund and then enter them into PV. But the problem is you need to recalculate these percentages for every year of the investment horizon. That's a lot of work ( I don't have a clue how Vineviz came up with that graph). * Your inflation adjusted data looks way better anyway as it is presented*. I'm leaning to the conclusion that duration matching for a total return stock/bond portfolio just does not apply. I still have not received a clear answer as to where the annual withdrawal actually comes from in this situation as I have posted earlier in the thread. It seems to me that the concept really applies to a 100% bond allocation.
"Success is going from failure to failure without loss of enthusiasm." Winston Churchill.
alluringreality
Posts: 787
Joined: Tue Nov 12, 2019 10:59 am

Re: How to use duration matching in a stock/bond portfolio

Post by alluringreality »

rossington wrote: Sat Aug 06, 2022 2:08 am
alluringreality wrote: Fri Aug 05, 2022 4:50 pm One of the top Google results on duration matching links to the following series of five posts. The parts I've skimmed are generally consistent with my understanding. In the conclusion the point is made that duration matching aims at a balance between price risk and reinvestment risk, and if someone has a different intent a separate strategy may be appropriate.

https://occaminvesting.co.uk/duration-m ... roduction/
https://occaminvesting.co.uk/why-you-sh ... erm-bonds/
https://occaminvesting.co.uk/duration-m ... -practice/
https://occaminvesting.co.uk/against-duration-matching/
https://occaminvesting.co.uk/should-you ... -matching/
Did you notice in #3 the reference to this: "The following examples are taken from this excellent Bogleheads thread": viewtopic.php?t=340252 ?
I know the first post noted the series was based on info from vineviz. Generally I thought the other parts of the series had a bit more third-party commentary mixed in with the original material than that one. I had read the original posts and only skimmed the seres, so I'm not sure if anywhere they brought up how most bond funds are basically rolling ladders. If someone is planning to fund yearly expenses, personally I tend to think non-rolling ladders might make practical sense, but I'm aware of potential limitations. Such a plan of using individual bonds to make a non-rolling ladder of bonds probably is most practical for government bonds. Aside from Bulletshares or iShares iBonds, I haven't personally ran across many diversified choices for other categories, and those choices seem fairly limited in relation to my own horizon.
30% Savings Bonds, 45% US Indexes, 25% Ex-US Indexes - Buy & Hold
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

rossington wrote: Mon Aug 08, 2022 4:25 am
TN_Boy wrote: Sun Aug 07, 2022 12:15 pm
stuff deleted ...

But the inflation adjusted portfolio end values don't match yours. All three were up, in inflation adjusted terms. Here is the link:

https://www.portfoliovisualizer.com/bac ... tion4_1=50

You have to click the button on the portfolio growth chart to get inflation adjusted results. I mention as since I want to duplicate your results and try some other combinations, I'm immediately hitting a difference. Your graph shows both investors have less in real terms than they started with, while I see all three investors having more real dollars.

After playing a bit more with this, I'll try the dynamic allocation tab, feeding it a changing mix of bond durations (you specify with a csv file). Vineviz (or anybody else) let me know if I've mishandled portfolio visualizer.
It looks like you would need 2 separate PV screens to do this. One to combine VFINX with a bond fund of a constant 4% duration. And a separate screen to combine VFINX with TWO bond funds (one long and one short(er) that are matched to have an average duration of 15 years. So you would need to calculate the percentages to allocate to each fund and then enter them into PV. But the problem is you need to recalculate these percentages for every year of the investment horizon. That's a lot of work ( I don't have a clue how Vineviz came up with that graph). * Your inflation adjusted data looks way better anyway as it is presented*. I'm leaning to the conclusion that duration matching for a total return stock/bond portfolio just does not apply. I still have not received a clear answer as to where the annual withdrawal actually comes from in this situation as I have posted earlier in the thread. It seems to me that the concept really applies to a 100% bond allocation.
Yes, I don't understand how Vineviz's charts show an investor starting in 1992 winding up with less real dollars, the portfolio visualizer (PV) certainly showed the investor up in real terms with a 4% inflation adjusted withdrawal (unless I'm not understanding that tool correctly ...). At least with a 4% inflation adjusted withdrawal, life should be good.

I was able to take a shot at the duration matching last night. There is a way to backtest dynamic allocations, though I found it via searching and not a link on the site. To use the dynamic allocation, you have to have a login, so I signed up for the 2 week free trial. You supply a csv file showing (in my case) the asset allocation each year. So using a combination of vanguard bond funds, I changed the mix per Vineviz's formula each year. I assumed an investment horizon of 34 years (I had to tweak things just a little because I needed to use two vanguard funds, the long treasures, vsutx, with a duration of 17 right now, and short treasuries, vfisx, with a duration a little over 2 years). You start out with all your bonds in vsutx in 1992, and all your bonds in vfisx in 2022, changing the percentage each year. The csv file tells PV what the mix is every year.

I can't create a link to that output, probably because that is a paid feature (I'm not going to bother to save it, upload, etc).

From the first test (see link) I came up with a terminal value of $255,540 in real terms for the portfolio using a static allocation of intermediate term treasuries, with my best shot at the duration matching I came up with a terminal value of $269,212. (Let's call the static portfolio A and the dynamic portfolio B)*.

What does that tell us? Not a whole lot. The intermediate term treasury fund has a duration a little over 5 years right now (I assume that's about what it has been). The *average* duration for the dynamic test (as we move from all long bonds to all short bonds) is 9.5. So during a period of falling interest rates the portfolio with a longer average bond duration did better. Not a surprise! Basically during this timeframe, the longer the duration, the better off you were. Portfolio B did a bit better than A, but you can easily do better than B with a static allocation by simply lengthening the bond duration a bit.

What happens if interest rates are rising? I don't know. I suspect duration matching does worse than an intermediate bond fund. I don't have access to a dataset I can easily use allowing such portfolio visualizer like tests to be run. It may be possible, I just don't know how to do it without a lot more work.

I'm kinda left with this:

1) Despite my respect for the posters advocating duration matching I can't see that a strong case has been made for its use in the conditions described in the original post.

2) A stock bond portfolio is different from a bond-only portfolio because, again given the OP conditions, your sales of bonds can be irregular in both timing and amount.

3) An intermediate bond fund is a decent choice.

4) If you want to do duration matching it is quite simple with a mix of two bond funds. Thus if you find the idea compelling, there is no reason to avoid doing so.

5) I have no idea why my data and Vineviz's differ so much; my results showed that duration matching had a modest advantage over intermediate term treasuries from 1992 to 2021. His graph showed both solutions with the portfolio down in real terms and duration matching doubling the portfolio final value over using a 4 year bond duration.

* I would need to post more details of my work in the dynamic case for someone to verify that I did everything correctly, but I think I was in the ball park.
dbr
Posts: 41845
Joined: Sun Mar 04, 2007 9:50 am

Re: How to use duration matching in a stock/bond portfolio

Post by dbr »

I also respect the financial know how and (what to me are) interesting discussions of how investing works for different kinds of assets and different kinds of plans that we have here.

But there is a quarantine in place. My actual investments are just some simple-minded three fund portfolio with my own quirky choice of a bond fund and then I go away and just leave it alone. There is enough analysis to show the plan is going to suit the objectives well enough to stop thinking about it, or even enough to never have thought about it too much.

PS I am equally quarantined from things like buying I bonds, going out and buying individual Treasuries and TIPS, tilting to factors, holding years of cash in CDs, finding dividend paying stocks and a lot of other things that are interesting to understand and discuss.
Last edited by dbr on Mon Aug 08, 2022 9:27 am, edited 1 time in total.
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

dbr wrote: Mon Aug 08, 2022 7:57 am I also respect the financial know how and (what to me are) interesting discussions of how investing works for different kinds of assets and different kinds of plans that we have here.

But there is a quarantine in place. My actual investments are just some simple-minded three fund portfolio with my own quirky choice of a bond fund and then I go away and just leave it alone. There is enough analysis to show the plan is going to suit the objectives well enough to stop thinking about it, or even enough to never have thought about it too much.
What the experiments I did was make me believe that duration matching didn't do much over an intermediate bond fund during a period of falling interest rates. That's a pretty limited conclusion. It is of course possible I was not using portfolio visualizer correctly, though I'm sure someone will correct me if that is the case.

It's not that simple, of course; the timing of interest rate changes, stock market fluctuations, etc affect what assets are sold in what quantity, and when, thus the overall portfolio impact is hard to discern. But it seems 1992 to 2021 was a good set of conditions for a duration matching approach (just thinking about how interest rates would be interacting with bond duration) and if didn't help much then, then I'm not sure what conditions would cause it to help a lot.

That said, it did show a modest benefit over a five year static duration, and having a lower bond duration towards the end of retirement might (might) be of value. Then again, I care about overall portfolio behavior, not what the bonds are doing.

But it is easy to implement. Just rebalance the bond funds every year; the spreadsheet computations are trivial.
User avatar
vineviz
Posts: 12280
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: How to use duration matching in a stock/bond portfolio

Post by vineviz »

rossington wrote: Mon Aug 08, 2022 4:25 am It looks like you would need 2 separate PV screens to do this. One to combine VFINX with a bond fund of a constant 4% duration. And a separate screen to combine VFINX with TWO bond funds (one long and one short(er) that are matched to have an average duration of 15 years.
True, there is no one place in PortfolioVisualizer that allows a direct comparison of the two approaches. I downloaded the data and ran the comparison in Excel.
TN_Boy wrote: Mon Aug 08, 2022 8:38 am What the experiments I did was make me believe that duration matching didn't do much over an intermediate bond fund during a period of falling interest rates.
Leaving aside the subjective nature of "didn't do much", it definitely is true that we will observe a smaller difference in outcomes with a smaller difference in average duration.

On the other hand, a duration glide path isn't very difficult to implement and offers an easy (IMHO) of reducing the uncertainty around retirement consumption. In other words, it's a tool for increasing the likelihood of getting the withdrawal rate you were initially expecting. Whether that is something worth achieving will depend on the investor's aversion to (possibly) needing to reduce consumption later on.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Topic Author
TN_Boy
Posts: 2945
Joined: Sat Jan 17, 2009 12:51 pm

Re: How to use duration matching in a stock/bond portfolio

Post by TN_Boy »

vineviz wrote: Mon Aug 08, 2022 8:58 am
rossington wrote: Mon Aug 08, 2022 4:25 am It looks like you would need 2 separate PV screens to do this. One to combine VFINX with a bond fund of a constant 4% duration. And a separate screen to combine VFINX with TWO bond funds (one long and one short(er) that are matched to have an average duration of 15 years.
True, there is no one place in PortfolioVisualizer that allows a direct comparison of the two approaches. I downloaded the data and ran the comparison in Excel.
TN_Boy wrote: Mon Aug 08, 2022 8:38 am What the experiments I did was make me believe that duration matching didn't do much over an intermediate bond fund during a period of falling interest rates.
Leaving aside the subjective nature of "didn't do much", it definitely is true that we will observe a smaller difference in outcomes with a smaller difference in average duration.

On the other hand, a duration glide path isn't very difficult to implement and offers an easy (IMHO) of reducing the uncertainty around retirement consumption. In other words, it's a tool for increasing the likelihood of getting the withdrawal rate you were initially expecting. Whether that is something worth achieving will depend on the investor's aversion to (possibly) needing to reduce consumption later on.
As I noted above, there IS a place in portfolio visualizer to do the exact test required. It's harder to find and you need a login. This is the link:

https://www.portfoliovisualizer.com/bac ... allocation

For settings, click on "Compared Portfolio" then select static allocation and you can define a portfolio. But as I said, you have to create a login (14 day free trial!) to actually upload an allocation history and run a backtest.

Where I differ here is that depending upon the interplay of interest rates, bond duration, harvesting approach, etc it is unclear that duration matching the bonds reduces uncertainty. It might or might not. Someone retiring in 2021 might find that the damage done to long bonds early in retirement is a big deal as the future unfolds. If you are selling from the bond side early in retirement, selling long bonds hit hard by interest rate increases is just going to be painful. Our 1992 retiree would have been watching their stock side skyrocket early, so no need (in a rebalancing harvesting method) to sell bonds for several years.

It would be nice to have a large set of backtests to see how differing interest rate and market conditions have affected this choice in the past.

Here's another way to think about it. At one time, people advocated reducing stock percentage as retirement progressed - to "reduce risk." More recently, some people have argued for a rising percentage of equity (or a bond tent, etc) as retirement progresses -- lower equity percent earlier to avoid a bad return sequence if equities crash early. Then (yet more people) argue that a bond tent does nothing to improve survivability. And so forth. I think duration matching versus static intermediate bond fund is less important than any substantial change in equity percentage and the wonderfulness of both increasing equity percentage and duration matching depends upon the set of market conditions you hit. I confess, duration matching was actually less better than intermediate bonds than I expected*, given interest rates trends over the time period and the relatively short duration of the intermediate treasury fund.

I do agree implementation of duration matching is straightforward. If you like long bonds early, short bonds late, then I'd do that. And I can see why some people might like that.

* that IS assuming my piloting (settings and data) of portfolio visualizer's dynamic allocation test was correct.
Post Reply