Do long-term bonds belong in one's portfolio?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
hdas
Posts: 1327
Joined: Thu Jun 11, 2015 8:24 am

Re: Do long-term bonds belong in one's portfolio?

Post by hdas » Tue Dec 11, 2018 10:58 pm

ThrustVectoring wrote:
Tue Dec 11, 2018 1:55 pm
What about holding the same amount of stocks and more bonds of a shorter duration? Leverage is extraordinarily cheap in the highly liquid treasury futures market, so it's pretty straightforward to grab the DV01 of the cash bond you're considering buying and the DV01 of a shorter-term futures contract and buy those up until you've gotten your desired interest rate sensitivity. It's no more risky (by the measure of dollar change per basis point), and you get to use whatever part of the yield curve is most attractive.
What is keeping you from implementing this ASAP?
For me, I I need to know all the ins and outs of the theoretical implementation and then I have to look at the empirical evidence. 2y futures have been around since 1990 I believe, so good enough to get a sense.

Is the 2yr future still 20% First In First Out (FIFO)/80% Pro-Rata? How would that work with a 4 lot?

Thanks.
"whenever there is a randomized way of doing something, then there is a nonrandomized way that delivers better performance but requires more thought" ET Jaynes

JackoC
Posts: 1009
Joined: Sun Aug 12, 2018 11:14 am

Re: Do long-term bonds belong in one's portfolio?

Post by JackoC » Wed Dec 12, 2018 10:17 am

vineviz wrote:
Sun Dec 09, 2018 7:50 pm

1. I think anyone who uses the ACM model, and especially anyone who defends it, should know what the 95% confidence interval is for that model's estimate of the the term premium. Do you know it is?


2. The OIS reflects real investors putting real money into an estimate of what short-term rates will look like over a given period of time. It seems pretty evident that proponents of the ACM model should have a pretty good idea of WHY that model is generating estimates that are 60bps above what the market is pricing in. I mean, if your model of market expectations is 20% away from a direct observation of market expectations, that's an important piece of information.

3. I'm pretty sure the chief investment officer and head macroeconomist at Guggenheim know that the Fed is merely estimating, not setting, the term premium. Their point is that the estimate produced by the Fed, using the ACM model, is probably wildly inaccurate. For what it is worth, the Fed's response to that criticism was akin to "you might be right".
1. If we had the 95% confidence interval and we accepted *that* as accurate, what exactly would we do with it, assume the real term premium was at the 95%-tile highest level to justify long term bonds? :D
Term structure models in general show a much lower, very possibly negative, term premium now. Analysis backing the use of long term bonds is based in large part on an era where the same models show a high, in 80's-90's *very* high term premium. This is the basic point about term structure models. If you have to argue hard to establish the term premium isn't actually negative, *maybe* it's zero, not a great time for long term bonds.

2. Again, this is a basic misunderstanding by the Guggenheim article writer. The swap yield curve contains an unobservable term premium just like the treasury curve, and very likely virtually exactly the same magnitude of term premium*. Real investors in both market are putting real money into an estimate of what short term rates will be over a given period of time. Investors in treasuries with a 10 yr horizon who choose the 10 yr note over the 5 yr note are making an implicit judgement what the 5 yr rate 5 yrs forward will be, *adjusted for their risk preference* for the risk of waiting to see what the 5yr rate will be in 5yrs (buy a 5yr note now, buy another in 5yrs time) v locking it in now (buy a 10 yr note). That's what the term premium means. It means you can't directly observe the market's true expectation of future short(er) rates because today's forward rate for a given period contains a term risk premium. Same on the market swap curve, which is held in tight agreement with the market treasury curve by arbitrage*.

3. No, the whole thing they are arguing against is the claim the Federal Reserve Open Market Committee (not the research branch at the NY Fed who wrote that model) has created a negative term premium with QE. The two big holes in their argument are
a) they have not in fact provided evidence that the ACM model gives wildly inaccurate results. The key argument they make in that regard, 'look at the swap curve' is not only wrong, but remarkably amateurish.
b) to the extent central bank policies are creating a negative term premium, it doesn't have to be the Federal Reserve (Open Market Comm) doing it, or doing it alone. It could also be the very low term rate policies (result, anyway) of the ECB, BOJ, etc.

Point 2 seems to actually be the most important point of contention. Besides circular arguing about burdens of proof (the burden is on you to prove it to 95%!, no *you* have to be 95% sure! :happy ) the main potentially valid point offered by G against the term structure model results is 'look at this market measure which contradicts it!' But that is bogus. Swap forward rates contain the unobservable term premium as much as treasury rates do, so prove nothing about the output of a term term structure model.

*only differing by any separate term premium of swap spreads, and only variation between those curves is via those low and stable (in recent years) spreads.

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Wed Dec 12, 2018 10:58 am

JackoC wrote:
Wed Dec 12, 2018 10:17 am
1. If we had the 95% confidence interval and we accepted *that* as accurate, what exactly would we do with it, assume the real term premium was at the 95%-tile highest level to justify long term bonds? :D
For starters, we'd have some idea of how much or little regard we should have for people who rely on the ACM model to say things like "the term premium is negative."
JackoC wrote:
Wed Dec 12, 2018 10:17 am
2. Again, this is a basic misunderstanding by the Guggenheim article writer.
You're far too quick to assume that Guggenheim's economists don't understand economics, IMHO.

Rather than be so confident that you're smarter than they are, why no focus on what they actually are pointing out: the ACM model is producing estimates of the risk-neutral yield that is dramatically out of line with other (arguably more accurate) estimates for that parameter. This isn't something Guggenheim just made up: it's a common, and totally legitimate, criticism of the ACM model. Citigroup pointed out the same thing in 2016.
Most term premium models (and market participants) agree that term premia have fallen over the last few decades (Figure 24). The factors behind the decline are controversial but likely include declining volatility in inflation as well as a global glut of savings. But while models agree qualitatively, quantitatively they can give quite different estimates of current term premium and any given model will have wide confidence intervals around the point estimate for the term premium.
Circling back to the Financial Times, they've provided this chart which shows the sometime-absurd outputs of the ACM model over history.

Image
In general, 10-year interest rates have been higher than 5-year interest rates, although there were long periods when investors were essentially indifferent between the two time horizons. Longer-term rates were rarely much lower than medium-term rates, and when they were, it was during the sharp and deep recessions of the 1970s and early 1980s.

Weirdly, the NY Fed model suggests this has occurred despite traders’ expectations that real rates would fall and inflation would slow down for much of the past 55 years, which is why the teal line is negative for two-thirds of its history. (The current period is an unusual exception.)

Offsetting this apparent belief that short-term interest rates would fall over time, or at least almost never go up, was a modest risk premium — a premium that stayed confined within a range of 1 percentage point between the early 1970s and the early 2000s despite massive changes in the trend of inflation and several significant changes in the policy regime.

Is it reasonable to think people in the late 1970s “expected” short-term interest rates to be lower in the future than in the present, even as the country was suffering from a worsening inflationary spiral? Is it reasonable to think people in the mid-1990s “expected” short rates to fall over time even when the economy was booming and before the productivity boom had become fully evident?
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

JackoC
Posts: 1009
Joined: Sun Aug 12, 2018 11:14 am

Re: Do long-term bonds belong in one's portfolio?

Post by JackoC » Wed Dec 12, 2018 11:34 am

vineviz wrote:
Wed Dec 12, 2018 10:58 am
JackoC wrote:
Wed Dec 12, 2018 10:17 am
1. If we had the 95% confidence interval and we accepted *that* as accurate, what exactly would we do with it, assume the real term premium was at the 95%-tile highest level to justify long term bonds? :D
For starters, we'd have some idea of how much or little regard we should have for people who rely on the ACM model to say things like "the term premium is negative."
JackoC wrote:
Wed Dec 12, 2018 10:17 am
2. Again, this is a basic misunderstanding by the Guggenheim article writer.
You're far too quick to assume that Guggenheim's economists don't understand economics, IMHO.
1. The ACM model estimates the term premium is negative. Quote where I said I absolutely know this is true. However again, it would now take a big error to make it zero, and even if it's zero when the 1961-current average output was 1.6%, what does that say about evaluating long term bond investing in terms of historicals, which a bunch of posts had done prior to the start of this phase of the discussion? This is only natural to do, history is what we have. But I'd say at least as much caution is due in evaluating long term bond investing now based on history (and some of the historicals given in this thread were from 1980's when it was *really* high, per TS models in general) as quibbling about confidence intervals of term structure models.

And again: who determines which side has to be 95% sure they're 'right'? This comes up all the time in disputes about statistical studies or models and often turns comical eventually, as it will get here if you persist any longer in 'well you presented it so you have to prove it's right'.

2. The Guggenheim writer does not understand fixed income markets if they point at one market yield curve (the OIS swap curve) now and say it proves something about the term premium embedded in another market yield curve now, the treasury curve. That is an amateur misunderstanding. It's not about who is smart. It's not about who understands 'economics'. It's simply about the strongest potential argument they offer against the ACM output: 'look at this market variable, the result of real investors putting up real money, and it directly contradicts ACM output!'. No it doesn't. The swap curve also reflects a *not directly observable* term premium, like the one in the treasury curve. The swap curve doesn't tell you anything about the accuracy of ACM. That's an important mistake in their analysis.

Given that the other arguments are in the 'this doesn't look right' category, which is a lot less solid. Especially if we don't move the goal posts to 'the ACM model is exactly right all the time', from, 'ACM shows a negative term premium now v *ACM* showing a 1.6% avg term premium 1961-2018, how if at all should this factor into our view of longer term bonds?'. The supposed slam dunk argument that the ACM output is 'wildly inaccurate' now is the swap curve shape. That argument is simply wrong. Whereas, the argument that the term premium-free yield curve output of ACM was too often inverted in the past might be a valid criticism, but if so would indicate the model tends to *understate* true market expectations of future short rates. If true, that would make long bonds look even worse now.

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Wed Dec 12, 2018 12:18 pm

JackoC wrote:
Wed Dec 12, 2018 11:34 am
1. The ACM model estimates the term premium is negative. Quote where I said I absolutely know this is true. However again, it would now take a big error to make it zero, and even if it's zero when the 1961-current average output was 1.6%, what does that say about evaluating long term bond investing in terms of historicals, which a bunch of posts had done prior to the start of this phase of the discussion?
There is a huge problem relying on a model that provides a point estimate but provides no confidence interval at all: we have no internal indication that the difference between the current estimate and zero is statistically significant.

Heck, the model doesn't even tell us whether the difference between the current estimate and and the previous maximum estimate is statistically significant.

This is a problem that can't just be waved away, in part because just as the current term premium estimate is likely too low by a lot it is also true that many of the previous estimates were too high by a lot. Did you read the things I quoted?
JackoC wrote:
Wed Dec 12, 2018 11:34 am
And again: who determines which side has to be 95% sure they're 'right'? This comes up all the time in disputes about statistical studies or models and often turns comical eventually, as it will get here if you persist any longer in 'well you presented it so you have to prove it's right'.
There are inarguable burdens placed upon a model that is proposed to represent the world. If the term premium is being estimated by ACM with a 95% confidence interval of +/- 200 bps then what, precisely, is the value of that term premium estimate?
JackoC wrote:
Wed Dec 12, 2018 11:34 am
The Guggenheim writer does not understand fixed income markets if they point at one market yield curve (the OIS swap curve) now and say it proves something about the term premium embedded in another market yield curve now, the treasury curve. That is an amateur misunderstanding.
You seem determined to maintain that these PhD economists at the top of their field are guilty of "an amateur misunderstanding", even when I've pointed out that you've plainly misrepresented what at the economists at Guggenheim and Citigroup have written.

Can you not slow down for a second to consider MAYBE you are misinterpreting them somehow? That maybe the considered option of multiple professional economists MIGHT have some merit.

Or do you genuinely believe that an unvalidated model is producing better estimates of the expected real interest rate than the explicitly stated expectations of economists, bond managers, and options traders?
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

JackoC
Posts: 1009
Joined: Sun Aug 12, 2018 11:14 am

Re: Do long-term bonds belong in one's portfolio?

Post by JackoC » Wed Dec 12, 2018 1:00 pm

vineviz wrote:
Wed Dec 12, 2018 12:18 pm

1.
This is a problem that can't just be waved away, in part because just as the current term premium estimate is likely too low by a lot it is also true that many of the previous estimates were too high by a lot. Did you read the things I quoted?


There are inarguable burdens placed upon a model that is proposed to represent the world. If the term premium is being estimated by ACM with a 95% confidence interval of +/- 200 bps then what, precisely, is the value of that term premium estimate?

2. You seem determined to maintain that these PhD economists at the top of their field are guilty of "an amateur misunderstanding", even when I've pointed out that you've plainly misrepresented what at the economists at Guggenheim and Citigroup have written.

Or do you genuinely believe that an unvalidated model is producing better estimates of the expected real interest rate than the explicitly stated expectations of economists, bond managers, and options traders?
1. I think you are now well into exaggeration here. There's no plausible basis to think the model results are random. Hence the argument that past term premium-free yield curves were too downsloping is the wrong away around in arguing the model is wrong now, in the other direction. To a degree where we'd disregard its results. Which is really the question about any financial model, not whether it's 'right', and not whether we could dismiss it if we assume the real result is at the 95%-tile confidence level in the direction we'd like it to be. :happy The weight of evidence in general IMO does not add up to saying this model gives a random result, or that it's errors are random. The huge move down in term premium from the period of 'history' at lot of people look at to evaluate bonds, the 1980's to now, is sobering for long term bond investing, IMO.

2. The Guggenheim point about the swap curve as given in your earlier post was this:

" The problem is that today the ACM model overstates expected short-term rates, falsely implying that term premiums are lower than they really are. Specifically, ACM indicates that the risk-neutral 10-year yield is currently 3.50 percent, considerably above our own forecasts of the average short-term rate that will prevail over the next decade. Other outside estimates support this view: the 10-year overnight index swap (OIS) rate, or the average fed funds rate priced in to a fixed-floating swap, is 2.90 percent."

That is wrong. The swap yield curve is *not* an "outside estimate" of the term premium. It's a market yield curve just like the treasury curve, in an arbitrage relationship with the treasury curve, and reflecting similar risk premia to the treasury curve except for the swap spread, which is not directly relevant to the term premium. There is IOW no outside estimate of the term premium being provided by 'options traders' or any other traders. And that's otherwise the potentially most solid argument against the ACM output now. Economists and bond managers are entitled to their *opinions* of the term premium, as are any investors, and it cannot be excluded that these prove more correct than model output. But it would be a whole other level of proof if there was a *market indicator* contradicting ACM on its face. There isn't, pretty much by definition, and that's a major hole in Guggenheim's argument. It is again not a matter of 'smart', nor 'Phd'. The above quote is just not a valid argument as written. You cannot tell anything about the term premium by simple inspection of the OIS swap curve, just like you can't tell it by simple inspection of the treasury curve. The quote seems in plain English to claim otherwise, can't see what I'm misinterpreting about it.

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Wed Dec 12, 2018 2:38 pm

JackoC wrote:
Wed Dec 12, 2018 1:00 pm
1. I think you are now well into exaggeration here. There's no plausible basis to think the model results are random.
I never used the word random.

I'm suggesting that the model persistently overestimates the variance in the risk-free neutral rate, resulting in systematically inaccurate estimates of the term premium. During periods of volatile inflation expectations (e.g. 1960s through 1980s) the term premium is overestimated during periods of stable inflation expectations (e.g. 1990s through today) the term premium is underestimated.

I'm also suggesting that estimates for the term premium have confidence intervals that are so wide as to give the estimates limited utility in forming monetary or investment policy.

JackoC wrote:
Wed Dec 12, 2018 1:00 pm
The swap yield curve is *not* an "outside estimate" of the term premium.
Guggenheim were quite obviously using the term "outside estimate" to refer to estimates produced outside of Guggenheim. They mention that their own internal estimates for forward short-term rates are lower than the ACM estimate and so are "outside estimates" produced by economists and investors not employed by Guggenheim.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

international001
Posts: 1200
Joined: Thu Feb 15, 2018 7:31 pm

Re: Do long-term bonds belong in one's portfolio?

Post by international001 » Wed Dec 12, 2018 4:54 pm

ThrustVectoring wrote:
Tue Dec 11, 2018 1:55 pm
international001 wrote:
Tue Dec 11, 2018 11:40 am
For those who say to just hold more stocks, it's not good enough. Just looking at PV's EF, last 40 years using LT in a 60/40 porfolio improves IT for 1% (same risk)
What about holding the same amount of stocks and more bonds of a shorter duration? Leverage is extraordinarily cheap in the highly liquid treasury futures market, so it's pretty straightforward to grab the DV01 of the cash bond you're considering buying and the DV01 of a shorter-term futures contract and buy those up until you've gotten your desired interest rate sensitivity. It's no more risky (by the measure of dollar change per basis point), and you get to use whatever part of the yield curve is most attractive.
You may as well invest in a combination of very risky stocks and plain cash

The point of investing in bonds and stocks is that they are uncorrelated. Ideally, you want two assets of a similar return/risk that are uncorrelated

This is just the statistics of it. If you have a hunch about LT bonds being bad in the next years, then don't invest on them.

Northern Flicker
Posts: 5061
Joined: Fri Apr 10, 2015 12:29 am

Re: Do long-term bonds belong in one's portfolio?

Post by Northern Flicker » Wed Dec 12, 2018 5:33 pm

vineviz wrote:
Sat Dec 08, 2018 6:50 pm
longinvest wrote:
Sat Dec 08, 2018 6:36 pm

EDV isn't a Bogleheads investment. It's pure speculation (a hope of making a profit by selling it to a greater fool).
You’re entitled to your opinion, obviously.

But you’re wrong.
In fact a regular treasury bond is equivalent to a portfolio of zero-coupon bonds each with a maturity date and value corresponding to a coupon payment or return of principal at maturity of the treasury bond.

T-bills are also zero-coupon instruments.
Index fund investor since 1987.

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Wed Dec 12, 2018 5:55 pm

jalbert wrote:
Wed Dec 12, 2018 5:33 pm
vineviz wrote:
Sat Dec 08, 2018 6:50 pm
longinvest wrote:
Sat Dec 08, 2018 6:36 pm

EDV isn't a Bogleheads investment. It's pure speculation (a hope of making a profit by selling it to a greater fool).
You’re entitled to your opinion, obviously.

But you’re wrong.
In fact a regular treasury bond is equivalent to a portfolio of zero-coupon bonds each with a maturity date and value corresponding to a coupon payment or return of principal at maturity of the treasury bond.

T-bills are also zero-coupon instruments.
Good point
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

JackoC
Posts: 1009
Joined: Sun Aug 12, 2018 11:14 am

Re: Do long-term bonds belong in one's portfolio?

Post by JackoC » Fri Dec 14, 2018 6:07 pm

vineviz wrote:
Wed Dec 12, 2018 2:38 pm
JackoC wrote:
Wed Dec 12, 2018 1:00 pm
1. I think you are now well into exaggeration here. There's no plausible basis to think the model results are random.
I never used the word random.

I'm suggesting that the model persistently overestimates the variance in the risk-free neutral rate, resulting in systematically inaccurate estimates of the term premium. During periods of volatile inflation expectations (e.g. 1960s through 1980s) the term premium is overestimated during periods of stable inflation expectations (e.g. 1990s through today) the term premium is underestimated.

I'm also suggesting that estimates for the term premium have confidence intervals that are so wide as to give the estimates limited utility in forming monetary or investment policy.

JackoC wrote:
Wed Dec 12, 2018 1:00 pm
The swap yield curve is *not* an "outside estimate" of the term premium.
2. Guggenheim were quite obviously using the term "outside estimate" to refer to estimates produced outside of Guggenheim. They mention that their own internal estimates for forward short-term rates are lower than the ACM estimate and so are "outside estimates" produced by economists and investors not employed by Guggenheim.
1. The argument you now present is new as far as I can see, and not the one in previous links you gave. Can you explain and support this further?

2. The very next paragraph in your original quote from G said
"Each of these independent forecasts suggests the ACM risk-neutral yield estimate is too high." They are claiming the OIS curve is an 'independent forecast' of future short rates as compared to the treasury curve. It's not.

But my simple point here, all along, is that this is basically wrong, not a matter of opinion, not a matter of who is smart or has a Phd. One market yield curve, the OIS swap curve, does not tell anything (by simple inspection) about the term premium in another market yield curve. It is IMO 100% clear this is what G is claiming, that forward rates on the OIS swap curve call into question the validity of the calculated term premium from the ACM model. And that's wrong and a significant mistake because a market indicator clearly contradicting model output would be a big deal for said model. It's not the case here.

Captain kangaroo
Posts: 94
Joined: Mon Nov 27, 2017 12:45 am

Re: Do long-term bonds belong in one's portfolio?

Post by Captain kangaroo » Fri Dec 14, 2018 7:17 pm

I am very happy with ICSH

Ishares Ultra Short Term Bond

Up 2.07% YTD in a rising interest market, with a 2.86% yield. Not too bad. Bonds are safety for me. Does't get much better then this.

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Fri Dec 14, 2018 8:11 pm

JackoC wrote:
Fri Dec 14, 2018 6:07 pm

1. The argument you now present is new as far as I can see, and not the one in previous links you gave.
I’m not sure how you’ve concluded this if you read what I wrote earlier, or the sources I provided.

[Can you explain and support this further?
JackoC wrote:
Fri Dec 14, 2018 6:07 pm
2. The very next paragraph in your original quote from G said
"Each of these independent forecasts suggests the ACM risk-neutral yield estimate is too high." They are claiming the OIS curve is an 'independent forecast' of future short rates as compared to the treasury curve. It's not.
I think I’ve explained twice already that you are misinterpreting this sentence.

JackoC wrote:
Fri Dec 14, 2018 6:07 pm
And that's wrong and a significant mistake because a market indicator clearly contradicting model output would be a big deal for said model. It's not the case here.
Guggenheim are not wrong, but you’re right to understand that the ACM model outputs ARE contradicted. How big a deal that is I couldn’t say.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

User avatar
siamond
Posts: 5131
Joined: Mon May 28, 2012 5:50 am

Re: Do long-term bonds belong in one's portfolio?

Post by siamond » Sun Dec 23, 2018 8:06 pm

stlutz wrote:
Sat Dec 08, 2018 6:51 pm
The answer to this question really depends on the answer to the question of why you even want bonds in your portfolio.

Two "risk parity" constructions are often discussed here:

a) The "bonds are for safety" approach. For these investors, they are seeking returns from domestic and international stocks and using bonds to lever down their exposure. With this approach, short term bonds are most appropriate.

b) The "non-correlated investments" approach. For these investors, they are looking for multiple sources of risk/volatility in their portfolio. Stocks and long-term bonds are both volatile. Sometimes they correlate positively; sometimes they don't.
This is a nice summary. It recently (finally!) dawned on me that a) and b) are not mutually exclusive. I use a modicum of IT bonds as a glorified emergency fund ('safety'). I am vaguely considering to replace some of my equities by LT bonds ('non-correlated'). Both approaches could complement each other. In the past 20 years, I ran the numbers, this would have worked very neatly (and even better with STRIPs). In previous crises, notably inflationary (40s and 70s), the case seems much less convincing. I'd be curious to see a good model of LT STRIPs for such challenging times though.

This is the difficult thing with bonds, they tend to go through those multi-decades periods of time where they display returns completely different from the previous multi-decades periods of time. Stocks are much better behaved, if you ask me! :wink:

User avatar
willthrill81
Posts: 15146
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: Do long-term bonds belong in one's portfolio?

Post by willthrill81 » Sun Dec 23, 2018 10:57 pm

stlutz wrote:
Sat Dec 08, 2018 6:51 pm
The answer to this question really depends on the answer to the question of why you even want bonds in your portfolio.

Two "risk parity" constructions are often discussed here:

a) The "bonds are for safety" approach. For these investors, they are seeking returns from domestic and international stocks and using bonds to lever down their exposure. With this approach, short term bonds are most appropriate.

b) The "non-correlated investments" approach. For these investors, they are looking for multiple sources of risk/volatility in their portfolio. Stocks and long-term bonds are both volatile. Sometimes they correlate positively; sometimes they don't. If you can't handle the drawdown that could result from the two having positively correlation, then the portfolio should be levered down using short-term bonds.
Your view is the same as mine. In fact, I had a very similar post just this morning in another thread. Great minds think alike! :wink:
“It's a dangerous business, Frodo, going out your door. You step onto the road, and if you don't keep your feet, there's no knowing where you might be swept off to.” J.R.R. Tolkien,The Lord of the Rings

User avatar
siamond
Posts: 5131
Joined: Mon May 28, 2012 5:50 am

Re: Do long-term bonds belong in one's portfolio?

Post by siamond » Mon Dec 24, 2018 11:48 am

I am both curious and wary about LT STRIPs. Never paid attention to this until very recently. I was checking Vanguard EDV (Extended Duration Treasuries) and the fund size is a paltry $680M. For a fund that was created more than 10 years ago (2017), this seems... rather small?

One could argue that LT bonds are not for everybody, that is for sure, but Vanguard VBLTX (Total Bond Market Index) is close to $10B. It was created in 2001, by the way. Out of curiosity (this is apples and oranges, I know), Vanguard Total-Bonds is at $200B.

User avatar
JoMoney
Posts: 8166
Joined: Tue Jul 23, 2013 5:31 am

Re: Do long-term bonds belong in one's portfolio?

Post by JoMoney » Mon Dec 24, 2018 11:50 am

If I liked current interest rates, I would buy a SPIA rather than a 20 or 30 year bond.
"To achieve satisfactory investment results is easier than most people realize; to achieve superior results is harder than it looks." - Benjamin Graham

User avatar
willthrill81
Posts: 15146
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: Do long-term bonds belong in one's portfolio?

Post by willthrill81 » Mon Dec 24, 2018 11:52 am

JoMoney wrote:
Mon Dec 24, 2018 11:50 am
If I liked current interest rates, I would buy a SPIA rather than a 20 or 30 year bond.
You don't care for the "non-correlated asset" argument?
“It's a dangerous business, Frodo, going out your door. You step onto the road, and if you don't keep your feet, there's no knowing where you might be swept off to.” J.R.R. Tolkien,The Lord of the Rings

stlutz
Posts: 5506
Joined: Fri Jan 02, 2009 1:08 am

Re: Do long-term bonds belong in one's portfolio?

Post by stlutz » Mon Dec 24, 2018 11:59 am

willthrill81 wrote:
Sun Dec 23, 2018 10:57 pm
Your view is the same as mine. In fact, I had a very similar post just this morning in another thread. Great minds think alike! :wink:
Perhaps we should call this the "Final, Definitive" thread...? :sharebeer

User avatar
JoMoney
Posts: 8166
Joined: Tue Jul 23, 2013 5:31 am

Re: Do long-term bonds belong in one's portfolio?

Post by JoMoney » Mon Dec 24, 2018 12:05 pm

willthrill81 wrote:
Mon Dec 24, 2018 11:52 am
JoMoney wrote:
Mon Dec 24, 2018 11:50 am
If I liked current interest rates, I would buy a SPIA rather than a 20 or 30 year bond.
You don't care for the "non-correlated asset" argument?
Nope. I don't measure the 'standard deviation' or correlations of my portfolio.
A SPIA offers guaranteed income. Coupons on a long bond do as well, but it will change over time... A STRIP doesn't even offer a coupon.
"To achieve satisfactory investment results is easier than most people realize; to achieve superior results is harder than it looks." - Benjamin Graham

stlutz
Posts: 5506
Joined: Fri Jan 02, 2009 1:08 am

Re: Do long-term bonds belong in one's portfolio?

Post by stlutz » Mon Dec 24, 2018 12:13 pm

JoMoney wrote:
Mon Dec 24, 2018 11:50 am
If I liked current interest rates, I would buy a SPIA rather than a 20 or 30 year bond.
I see the argument if one was, say, 65 years old. What if one was 40 years old?

User avatar
JoMoney
Posts: 8166
Joined: Tue Jul 23, 2013 5:31 am

Re: Do long-term bonds belong in one's portfolio?

Post by JoMoney » Mon Dec 24, 2018 12:21 pm

stlutz wrote:
Mon Dec 24, 2018 12:13 pm
JoMoney wrote:
Mon Dec 24, 2018 11:50 am
If I liked current interest rates, I would buy a SPIA rather than a 20 or 30 year bond.
I see the argument if one was, say, 65 years old. What if one was 40 years old?
If I really liked the the interest rates, but didn't need the current income, maybe a delayed annuity with a start date down the road so the current interest was accumulated and paid out at an even larger rate later.
What I currently do, is have most of my investments in equities and a small amount of cash-short term bonds.
"To achieve satisfactory investment results is easier than most people realize; to achieve superior results is harder than it looks." - Benjamin Graham

longinvest
Posts: 4085
Joined: Sat Aug 11, 2012 8:44 am

Re: Do long-term bonds belong in one's portfolio?

Post by longinvest » Mon Dec 24, 2018 12:22 pm

siamond wrote:
Mon Dec 24, 2018 11:48 am
I am both curious and wary about LT STRIPs. Never paid attention to this until very recently. I was checking Vanguard EDV (Extended Duration Treasuries) and the fund size is a paltry $680M. For a fund that was created more than 10 years ago (2017), this seems... rather small?
A STRIPS (capital final S, as it stands for Separate Trading of Registered Interest and Principal of Securities) is a derivative product; while it is backed by the US government, it isn't directly sold by the Federal Reserve or a governmental agency. It has a "stripping" cost which hidden in its yield. Note that it would be a serious mistake to compare the yield to maturity (YTM) of a STRIPS with the YTM of a 30-year Treasury with coupons, due to their duration mismatch.

I am quite disturbed by the fact that the select forum members who started touting EDV, lately, have never discussed these things. They present EDV as if it was a normal "very long-term" Treasury index fund, which it isn't. For one thing, it isn't indexed to market caps. Instead, it aims to track an index of equal-par weighted 30 to 20 year STRIPS. For another, Vanguard explicitly warns potential investors as follows:

"The fund is primarily intended for institutional investors with extremely long-term liabilities—20 years or more. Prospective individual investors are urged to consult with their own advisors to determine if the fund is suitable for their overall investment programs and financial positions."
Source: EDV -- Portfolio & Management

People are free to put their money into whatever suits them. But, as this is the Bogleheads forum where we share investing advice inspired by Jack Bogle, I can only encourage forum members to first try find out about how high the hidden stripping fees are before putting money into a complex investment that they don't fully understand. Jack Bogle likes to remind us that cost matters!

Let me insist: EDV has a very low expense ratio. But, this doesn't include the stripping fees that were taken when creating STRIPS's from Treasuries.
siamond wrote:
Mon Dec 24, 2018 11:48 am
One could argue that LT bonds are not for everybody, that is for sure, but Vanguard VBLTX (Total Bond Market Index) is close to $10B. It was created in 2001, by the way. Out of curiosity (this is apples and oranges, I know), Vanguard Total-Bonds is at $200B.
The inception date of Vanguard Long-Term Bond Index Fund (VBLTX) was March 1st, 1994. Portfolio Visualizer's chart effectively goes back to March 1994: https://www.portfoliovisualizer.com/bac ... ion1_1=100

It is more interesting to look at the underlying market size, instead of the fund's size. By looking at Vanguard's Total Bond Market's breakdown by maturities, we discover that long-term bonds represent 17% of the total government and investment-grade nominal bond market (excluding municipal bonds).
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW

User avatar
siamond
Posts: 5131
Joined: Mon May 28, 2012 5:50 am

Re: Do long-term bonds belong in one's portfolio?

Post by siamond » Mon Dec 24, 2018 12:43 pm

longinvest wrote:
Mon Dec 24, 2018 12:22 pm
The inception date of Vanguard Long-Term Bond Index Fund (VBLTX) was March 1st, 1994.
Hmpf, you're right. Fat fingers when doing a quick search are NOT helpful! VBTLX is NOT VBLTX. :oops:
longinvest wrote:
Mon Dec 24, 2018 12:22 pm
It is more interesting to look at the underlying market size, instead of the fund's size. By looking at Vanguard's Total Bond Market's breakdown by maturities, we discover that long-term bonds represent 17% of the total government and investment-grade nominal bond market (excluding municipal bonds).
Yes, sure, I was just using an easy (and admittedly poor) proxy. Any clue about the market size for LT STRIPs?
longinvest wrote:
Mon Dec 24, 2018 12:22 pm
People are free to put their money into whatever suits them. But, as this is the Bogleheads forum where we share investing advice inspired by Jack Bogle, I can only encourage forum members to first try find out about how high the hidden stripping fees are before putting money into a complex investment that they don't fully understand. Jack Bogle likes to remind us that cost matters!

Let me insist: EDV has a very low expense ratio. But, this doesn't include the stripping fees that were taken when creating STRIPS's from Treasuries.
Fair point for sure. Thanks for being thorough, this is appreciated. Personally, I am very pragmatic and I simply looked at historical returns (as a separate asset class first, but much more importantly, as part of an aggregate portfolio). Now if you have a way to hone on those stripping fees, please share...
Last edited by siamond on Mon Dec 24, 2018 12:47 pm, edited 1 time in total.

HEDGEFUNDIE
Posts: 3973
Joined: Sun Oct 22, 2017 2:06 pm

Re: Do long-term bonds belong in one's portfolio?

Post by HEDGEFUNDIE » Mon Dec 24, 2018 12:47 pm

siamond wrote:
Mon Dec 24, 2018 11:48 am
I am both curious and wary about LT STRIPs. Never paid attention to this until very recently. I was checking Vanguard EDV (Extended Duration Treasuries) and the fund size is a paltry $680M. For a fund that was created more than 10 years ago (2017), this seems... rather small?

One could argue that LT bonds are not for everybody, that is for sure, but Vanguard VBLTX (Total Bond Market Index) is close to $10B. It was created in 2001, by the way. Out of curiosity (this is apples and oranges, I know), Vanguard Total-Bonds is at $200B.
EDV is only 680M but if you add all the share classes together (https://institutional.vanguard.com/VGAp ... undId=0930) it comes to $1.8B. Stepping back, why does the AUM matter? Either it works for your portfolio strategy or it doesn’t.
longinvest wrote:
Mon Dec 24, 2018 12:22 pm
I am quite disturbed by the fact that the select forum members who started touting EDV, lately, have never discussed these things. They present EDV as if it was a normal "very long-term" Treasury index fund, which it isn't. For one thing, it isn't indexed to market caps. Instead, it aims to track an index of equal-par weighted 30 to 20 year STRIPS.
To call EDV some exotic instrument that Bogleheads can’t understand underestimates the intelligence of Bogleheads. This is not some exotic product cooked up by hedge funds. It’s a portfolio of US Treasury debt with an average duration of 24 years. Vanguard Long Term Treasury has an average duration of 17 years. If you wanted longer duration than VUSTX then EDV is a good option. It’s as simple as that. And the longer duration makes EDV more volatile and more negatively correlated to stocks and financial shocks.

EDV has been around now for 11 years. We have history of its performance through the financial crisis and the mini crashes since then. We also have evidence of its performance through a rate rise cycle these past two years. What more are you looking for?
Last edited by HEDGEFUNDIE on Mon Dec 24, 2018 12:49 pm, edited 1 time in total.

longinvest
Posts: 4085
Joined: Sat Aug 11, 2012 8:44 am

Re: Do long-term bonds belong in one's portfolio?

Post by longinvest » Mon Dec 24, 2018 12:48 pm

siamond wrote:
Mon Dec 24, 2018 12:43 pm
Any clue about the market size for LT STRIPs?
I have no idea.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Mon Dec 24, 2018 12:48 pm

longinvest wrote:
Mon Dec 24, 2018 12:22 pm
A STRIPS (capital final S, as it stands for Separate Trading of Registered Interest and Principal of Securities) is a derivative product; while it is backed by the US government, it isn't directly sold by the Federal Reserve or a governmental agency. It has a "stripping" cost which hidden in its yield. Note that it would be a serious mistake to compare the yield to maturity (YTM) of a STRIPS with the YTM of a 30-year Treasury with coupons, due to their duration mismatch.
I'm curious about your persistent belief that zero coupon bonds have some sort of "hidden" cost. I'd love to see an explanation for this belief of yours.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

longinvest
Posts: 4085
Joined: Sat Aug 11, 2012 8:44 am

Re: Do long-term bonds belong in one's portfolio?

Post by longinvest » Mon Dec 24, 2018 12:50 pm

Dear HEDGEFUNDIE,
HEDGEFUNDIE wrote:
Mon Dec 24, 2018 12:47 pm
longinvest wrote:
Mon Dec 24, 2018 12:22 pm
I am quite disturbed by the fact that the select forum members who started touting EDV, lately, have never discussed these things. They present EDV as if it was a normal "very long-term" Treasury index fund, which it isn't. For one thing, it isn't indexed to market caps. Instead, it aims to track an index of equal-par weighted 30 to 20 year STRIPS.
To call EDV some exotic instrument that Bogleheads can’t understand underestimates the intelligence of Bogleheads.
I have not written that Bogleheads can't understand EDV.

Best regards,

longinvest
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW

User avatar
siamond
Posts: 5131
Joined: Mon May 28, 2012 5:50 am

Re: Do long-term bonds belong in one's portfolio?

Post by siamond » Mon Dec 24, 2018 12:56 pm

HEDGEFUNDIE wrote:
Mon Dec 24, 2018 12:47 pm
EDV has been around now for 11 years. We have history of its performance through the financial crisis and the mini crashes since then. We also have evidence of its performance through a rate rise cycle these past two years. What more are you looking for?
We actually have 20 years of history, thanks to the corresponding index (Bloomberg Barclays 20-30Y Treasury Strips TR USD). Which is good. But nowhere near good enough to assess any investment vehicle, *notably* bonds with their decades-long changing patterns. Ideally, we should develop a better understanding of how such vehicle would have worked in inflationary time periods, e.g. post WW-II and mid 70s. We have the history for LT Treasuries (or at least a model that seems pretty reliable). We don't have the equivalent for STRIPs - so far. It would be cool to find a way to fill this gap - if feasible.

Don't misread me, I am NOT pushing back on your ideas. I AM intrigued. I just would like to see more facts on the table (hidden costs, historical data, etc), so that we can all make more informed judgments for our own personal goals. Let's keep an open discussion going, this is very informative.

HEDGEFUNDIE
Posts: 3973
Joined: Sun Oct 22, 2017 2:06 pm

Re: Do long-term bonds belong in one's portfolio?

Post by HEDGEFUNDIE » Mon Dec 24, 2018 1:02 pm

siamond wrote:
Mon Dec 24, 2018 12:56 pm
HEDGEFUNDIE wrote:
Mon Dec 24, 2018 12:47 pm
EDV has been around now for 11 years. We have history of its performance through the financial crisis and the mini crashes since then. We also have evidence of its performance through a rate rise cycle these past two years. What more are you looking for?
We actually have 20 years of history, thanks to the corresponding index (Bloomberg Barclays 20-30Y Treasury Strips TR USD). Which is good. But nowhere near good enough to assess any investment vehicle, *notably* bonds with their decades-long changing patterns. Ideally, we should develop a better understanding of how such vehicle would have worked in inflationary time periods, e.g. post WW-II and mid 70s. We have the history for LT Treasuries (or at least a model that seems pretty reliable). We don't have the equivalent for STRIPs - so far. It would be cool to find a way to fill this gap - if feasible.

Don't misread me, I am NOT pushing back on your ideas. I AM intrigued. I just would like to see more facts on the table (hidden costs, historical data, etc), so that we can all make more informed judgments for our own personal goals. Let's keep an open discussion going, this is very informative.
During an inflationary period you should expect a 24 year duration bond fund to suffer significant losses, worse than what “ordinary” long term treasuries would suffer. This happened in the late 70s early 80s. But keep in mind that stocks typically do well during inflationary periods. This is why I keep EDV to only 20% of my portfolio. If it crashes 50% I expect that loss to be made up by my stocks.

Here is an 11 year backtest of all four Vanguard Treasury funds. Each one behaves exactly as you would expect them to behave given their different durations.

https://www.portfoliovisualizer.com/bac ... ion3_3=100

longinvest
Posts: 4085
Joined: Sat Aug 11, 2012 8:44 am

Re: Do long-term bonds belong in one's portfolio?

Post by longinvest » Mon Dec 24, 2018 1:12 pm

Dear Vineviz,
vineviz wrote:
Mon Dec 24, 2018 12:48 pm
I'm curious about your persistent belief that zero coupon bonds have some sort of "hidden" cost. I'd love to see an explanation for this belief of yours.
I am quite surprised to see you writing this.

I really can't believe that you're not aware that the cost of a STRIPS is embedded in its yield!

This is quite similar to the cost of a Single-Premium Immediate Annuity (SPIA) which is embedded into its payout ratio. I don't think that anybody would question that an insurance company has to make some kind of a profit, when it sells a product.

Why should anyone believe that a third party (not the government nor the Federal Reserve) would buy a 30-year Treasury bond, strip it into its individual payments to create a collection 61 STRIPS, with all the administrative and legal overhead this involves, and not embed the cost and a profit margin in the yield of its newly created STRIPS, when selling them back individually to market participants? I know that it's Christmas time, but I've outgrown believing that Santa Claus comes through the Chimney, during the night, to put the gifts under the tree.

In an older post, I did an analysis to estimate the embedded fees of a SPIA, getting as a result that the embedded fees (including profits) were somewhere between 3% and 10%. Once we know this, we can ask ourselves, before buying a SPIA, is whether this seems like reasonable cost, or not, for the stability and longevity insurance that it provides.

I currently have no idea how high or low the embedded fees of STIPS are. Are you aware of any study about it?

Best regards,

longinvest
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Mon Dec 24, 2018 1:16 pm

siamond wrote:
Mon Dec 24, 2018 12:56 pm
HEDGEFUNDIE wrote:
Mon Dec 24, 2018 12:47 pm
EDV has been around now for 11 years. We have history of its performance through the financial crisis and the mini crashes since then. We also have evidence of its performance through a rate rise cycle these past two years. What more are you looking for?
We actually have 20 years of history, thanks to the corresponding index (Bloomberg Barclays 20-30Y Treasury Strips TR USD). Which is good. But nowhere near good enough to assess any investment vehicle, *notably* bonds with their decades-long changing patterns. Ideally, we should develop a better understanding of how such vehicle would have worked in inflationary time periods, e.g. post WW-II and mid 70s. We have the history for LT Treasuries (or at least a model that seems pretty reliable). We don't have the equivalent for STRIPs - so far. It would be cool to find a way to fill this gap - if feasible.

Don't misread me, I am NOT pushing back on your ideas. I AM intrigued. I just would like to see more facts on the table (hidden costs, historical data, etc), so that we can all make more informed judgments for our own personal goals. Let's keep an open discussion going, this is very informative.
STRIPS really aren't fundamentally different from Treasury bonds in any way.

Obviously, a STRIPS will have have more duration and higher convexity than a coupon-bearing Treasury of the same maturity but otherwise there's nothing about them that requires any distinct type of analysis.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

HEDGEFUNDIE
Posts: 3973
Joined: Sun Oct 22, 2017 2:06 pm

Re: Do long-term bonds belong in one's portfolio?

Post by HEDGEFUNDIE » Mon Dec 24, 2018 1:19 pm

longinvest wrote:
Mon Dec 24, 2018 1:12 pm
Dear Vineviz,
vineviz wrote:
Mon Dec 24, 2018 12:48 pm
I'm curious about your persistent belief that zero coupon bonds have some sort of "hidden" cost. I'd love to see an explanation for this belief of yours.
I am quite surprised to see you writing this.

I really can't believe that you're not aware that the cost of a STRIPS is embedded in its yield!

This is quite similar to the cost of a Single-Premium Immediate Annuity (SPIA) which is embedded into its payout ratio. I don't think that anybody would question that an insurance company has to make some kind of a profit, when it sells a product.

Why should anyone believe that a third party (not the government nor the Federal Reserve) would buy a 30-year Treasury bond, strip it into its individual payments to create a collection 61 STRIPS, with all the administrative and legal overhead this involves, and not embed the cost and a profit margin in the yield of its newly created STRIPS, when selling them back individually to market participants? I know that it's Christmas time, but I've outgrown believing that Santa Claus comes through the Chimney, during the night, to put the gifts under the tree.

In an older post, I did an analysis to estimate the embedded fees of a SPIA, getting as a result that the embedded fees (including profits) were somewhere between 3% and 10%. Once we know this, we can ask ourselves, before buying a SPIA, is whether this seems like reasonable cost, or not, for the stability and longevity insurance that it provides.

I currently have no idea how high or low the embedded fees of STIPS are. Are you aware of any study about it?

Best regards,

longinvest
Here is the Treasury regulation governing STRIPS creation. Section 356.31

https://www.treasurydirect.gov/instit/s ... cfr356.pdf

Give it a read and let us know where all these exorbitant fees come in.

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Mon Dec 24, 2018 1:27 pm

longinvest wrote:
Mon Dec 24, 2018 1:12 pm
I currently have no idea how high or low the embedded fees of STIPS are. Are you aware of any study about it?
I doubt such study exists because there IS no "embedded fee" in STRIPS.

An investor who buys a zero-coupon bond in the secondary market is quoted a yield-to-maturity that is EXACTLY what they'll get if they they hold that bond to maturity. There is no "fee", and certainly no "hidden fee". You get the yield at which you purchase the bond. Period.

Is there some expense involved in splitting the coupon payments from the principal payments? Undoubtedly, although given the ease with which the NY Fed will do that for a bank I suspect the cost is pretty low. But I'm sure there is a cost, and that the banks that originate the creation of STRIPS are surely compensated for doing that origination. But any such compensation is accounted for the in the quoted YTM for the zero coupon bond at the time of purchase.

If a retail investor, for whatever reason, wanted to synthetically reconstitute the Treasury bond by purchasing the stripped principal and the stripped coupons separately then I have no doubt they'd pay a premium relative to the cost of buying the bond as issued. I can't imagine why any investor would want to do that, nor can I imagine that such an investor would be surprised to find they'd paid more to create the bundle than if they'd just bought the bundle in the first place. Either way, it makes no sense to refer to the extra cost of reconstituting the bond as "hidden fee".
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Mon Dec 24, 2018 1:30 pm

longinvest wrote:
Mon Dec 24, 2018 1:12 pm
Why should anyone believe that a third party (not the government nor the Federal Reserve) would buy a 30-year Treasury bond, strip it into its individual payments to create a collection 61 STRIPS, with all the administrative and legal overhead this involves, and not embed the cost and a profit margin in the yield of its newly created STRIPS, when selling them back individually to market participants?
You should keep in mind that STRIPS, although not directly issued by the Treasury, are in fact created by the Federal Reserve Bank of New York for the benefit of banks and/or brokerages.

https://www.newyorkfed.org/aboutthefed/ ... fed42.html
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

longinvest
Posts: 4085
Joined: Sat Aug 11, 2012 8:44 am

Re: Do long-term bonds belong in one's portfolio?

Post by longinvest » Mon Dec 24, 2018 2:29 pm

siamond wrote:
Mon Dec 24, 2018 12:43 pm
Personally, I am very pragmatic and I simply looked at historical returns (as a separate asset class first, but much more importantly, as part of an aggregate portfolio). Now if you have a way to hone on those stripping fees, please share...
A while ago when we were trying to model TIPS rates and returns pre-1997, I created a TIPS simulator spreadsheet which, among other things, estimated a zero-coupon yield curve out of the normal Treasury yield curve.

I think that I could use this estimated zero-coupon yield curve to create a synthetic EDV-like fund: a zero-coupon ladder-like fund 30-to-21 years, equal-par weighted. I'll try to do that in the near future and post the result on the Historical Bond Returns - Shiller: From Rates to Returns thread.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW

User avatar
siamond
Posts: 5131
Joined: Mon May 28, 2012 5:50 am

Re: Do long-term bonds belong in one's portfolio?

Post by siamond » Mon Dec 24, 2018 3:00 pm

longinvest wrote:
Mon Dec 24, 2018 2:29 pm
I think that I could use this estimated zero-coupon yield curve to create a synthetic EDV-like fund: a zero-coupon ladder-like fund 30-to-21 years, equal-par weighted. I'll try to do that in the near future and post the result on the Historical Bond Returns - Shiller: From Rates to Returns thread.
Yeah, my first Xmas gift! :wink:

More seriously, yes, this would be terrific, but no rush. :beer

JackoC
Posts: 1009
Joined: Sun Aug 12, 2018 11:14 am

Re: Do long-term bonds belong in one's portfolio?

Post by JackoC » Thu Dec 27, 2018 3:34 pm

vineviz wrote:
Fri Dec 14, 2018 8:11 pm
JackoC wrote:
Fri Dec 14, 2018 6:07 pm

1. The argument you now present is new as far as I can see, and not the one in previous links you gave.
I’m not sure how you’ve concluded this if you read what I wrote earlier, or the sources I provided.

[Can you explain and support this further?
JackoC wrote:
Fri Dec 14, 2018 6:07 pm
2. The very next paragraph in your original quote from G said
"Each of these independent forecasts suggests the ACM risk-neutral yield estimate is too high." They are claiming the OIS curve is an 'independent forecast' of future short rates as compared to the treasury curve. It's not.
I think I’ve explained twice already that you are misinterpreting this sentence.

JackoC wrote:
Fri Dec 14, 2018 6:07 pm
And that's wrong and a significant mistake because a market indicator clearly contradicting model output would be a big deal for said model. It's not the case here.
Guggenheim are not wrong, but you’re right to understand that the ACM model outputs ARE contradicted. How big a deal that is I couldn’t say.
Sorry for the delay in responding.

1. I didn't see that argument made in any of your sources. But in any case, please indulge me and explain it further in your own words.

2. You said twice that I was 'misinterpreting' that sentence but I was not. Their concept is clearly that looking at forward rates on the OIS swap curve can tell you something about the term premium. If that wasn't the claim, why refer to the OIS swap curve at all? But that's a basic misunderstanding by them. Simple observation of the forward rates on one market yield curve can't tell you anything about the term premium embedded in the forward rates on another market yield curve. The same risk preferences of the market are embedded in both curves.

And compare points in 1 and 2. 1 is pretty esoteric (to repeat your words: "the model persistently overestimates the variance in the risk-free neutral rate, resulting in systematically inaccurate estimates of the term premium. During periods of volatile inflation expectations (e.g. 1960s through 1980s) the term premium is overestimated during periods of stable inflation expectations [e.g. 1990s through today] the term premium is underestimated.") 2. is a very simple. They are saying that forward rates on the OIS swap curve show that the ACM term structure model is coming up with forward rates that are too high. But that's a simply wrong observation by them. Whatever the term premium is, it's just as much embedded in the OIS swap curve as it in the treasury curve.

So I'm left with a simple point that's clearly wrong (2), and an esoteric point (1) you don't seem willing to explain in your own words.

KJVanguard
Posts: 222
Joined: Fri Oct 05, 2018 3:39 am

Re: Do long-term bonds belong in one's portfolio?

Post by KJVanguard » Thu Dec 27, 2018 4:59 pm

I have NEVER owned long-term bonds. And I suspect that I never will. Well, if this were 1981 and long bonds yielded 18% then I might change my mind.

But in my 25 years of investing I have never wanted them. I don't even own intermediate bonds any longer (I used to when rates were higher). I am 100% in Short-Term Investment Grade Admiral as my one & only bond fund.

I figure that I take lots of risk in the stock portion of my portfolio, such that bonds are there for safety & security, so I want something that won't move around much.

longinvest
Posts: 4085
Joined: Sat Aug 11, 2012 8:44 am

Re: Do long-term bonds belong in one's portfolio?

Post by longinvest » Thu Dec 27, 2018 9:25 pm

siamond wrote:
Mon Dec 24, 2018 3:00 pm
longinvest wrote:
Mon Dec 24, 2018 2:29 pm
I think that I could use this estimated zero-coupon yield curve to create a synthetic EDV-like fund: a zero-coupon ladder-like fund 30-to-21 years, equal-par weighted. I'll try to do that in the near future and post the result on the Historical Bond Returns - Shiller: From Rates to Returns thread.
Yeah, my first Xmas gift! :wink:

More seriously, yes, this would be terrific, but no rush. :beer
OK, I'm a little late for Xmas, but it's done and uploaded. See this post for details.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW

User avatar
willthrill81
Posts: 15146
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: Do long-term bonds belong in one's portfolio?

Post by willthrill81 » Thu Dec 27, 2018 10:21 pm

KJVanguard wrote:
Thu Dec 27, 2018 4:59 pm
I have NEVER owned long-term bonds. And I suspect that I never will. Well, if this were 1981 and long bonds yielded 18% then I might change my mind.
The problem in 1981 was that long-term nominal Treasuries had just been destroyed in real terms over the last four years, suffering a 46.5% drawdown in inflation-adjusted dollars from 1977 to 1981. How many investors were signing up for more in 1981? I certainly wouldn't have been. I know of one investor who commented about how skittish people were to take out 5 year CDs in the mid 1980s for fear of inflation and interest rates.
Last edited by willthrill81 on Thu Dec 27, 2018 11:15 pm, edited 1 time in total.
“It's a dangerous business, Frodo, going out your door. You step onto the road, and if you don't keep your feet, there's no knowing where you might be swept off to.” J.R.R. Tolkien,The Lord of the Rings

Northern Flicker
Posts: 5061
Joined: Fri Apr 10, 2015 12:29 am

Re: Do long-term bonds belong in one's portfolio?

Post by Northern Flicker » Thu Dec 27, 2018 11:14 pm

An issuer of STRIPS can try to embed all the fees they want into the yield. Ultimately, they have to find an investor who is willing to buy the instrument, and that will lead to the market establishing the yield. How much of a premium are market participants willing to pay for a STRIP in comparison to a long-term treasury?
Index fund investor since 1987.

User avatar
siamond
Posts: 5131
Joined: Mon May 28, 2012 5:50 am

Re: Do long-term bonds belong in one's portfolio?

Post by siamond » Thu Dec 27, 2018 11:56 pm

longinvest wrote:
Thu Dec 27, 2018 9:25 pm
siamond wrote:
Mon Dec 24, 2018 3:00 pm
longinvest wrote:
Mon Dec 24, 2018 2:29 pm
I think that I could use this estimated zero-coupon yield curve to create a synthetic EDV-like fund: a zero-coupon ladder-like fund 30-to-21 years, equal-par weighted. I'll try to do that in the near future and post the result on the Historical Bond Returns - Shiller: From Rates to Returns thread.
Yeah, my first Xmas gift! :wink:

More seriously, yes, this would be terrific, but no rush. :beer
OK, I'm a little late for Xmas, but it's done and uploaded. See this post for details.
I ran a quick backtest of the zero-coupon 30-21 synthetic model against the index tracked by Vanguard EDV (i.e. Bloomberg Barclays 20-30Y Treasury Strips TR USD). Here is a growth chart expressed in real terms (in other words, a telltale chart against inflation) comparing the two trajectories since the index' inception. I also included the index tracked by Vanguard Long-Term Treasury Fund (VUSTX), and its corresponding model in longinvest's updated spreadsheet.

Image

The model tends to drift a tiny bit lower than reality (the index) in both cases, but this remains a really good match. Which means that what the returns computed by the model for the pre-1997 years is probably a pretty good approximation of what an LT STRIPs fund would have returned. Nice modeling job.

PS. I disabled the display of EDV itself on the chart, as it was overlapping very closely its index and was just cluttering the display. This indicates that whatever friction costs due to stripping should be very low. And that Vanguard does a really good job of tracking the index.

HEDGEFUNDIE
Posts: 3973
Joined: Sun Oct 22, 2017 2:06 pm

Re: Do long-term bonds belong in one's portfolio?

Post by HEDGEFUNDIE » Fri Dec 28, 2018 1:58 am

longinvest wrote:
Thu Dec 27, 2018 9:25 pm
siamond wrote:
Mon Dec 24, 2018 3:00 pm
longinvest wrote:
Mon Dec 24, 2018 2:29 pm
I think that I could use this estimated zero-coupon yield curve to create a synthetic EDV-like fund: a zero-coupon ladder-like fund 30-to-21 years, equal-par weighted. I'll try to do that in the near future and post the result on the Historical Bond Returns - Shiller: From Rates to Returns thread.
Yeah, my first Xmas gift! :wink:

More seriously, yes, this would be terrific, but no rush. :beer
OK, I'm a little late for Xmas, but it's done and uploaded. See this post for details.
Great work longinvest.

Taking a look at the data, 1978-1981 was the last prolonged bear market for our hypothetical long zero coupon fund. Cumulative drawdown of -44%, compared to a 20-year bond fund which only lost -7%.

However, the S&P 500 returned 54% over the same time period. And in 1982 the long zeros recovered the entire drawdown in just one year.

So the general observation about STRIPS still holds. Strong negative correlation to equities during times of crisis, highly volatile.

User avatar
vineviz
Posts: 5391
Joined: Tue May 15, 2018 1:55 pm

Re: Do long-term bonds belong in one's portfolio?

Post by vineviz » Fri Dec 28, 2018 7:53 am

siamond wrote:
Thu Dec 27, 2018 11:56 pm
longinvest wrote:
Thu Dec 27, 2018 9:25 pm
siamond wrote:
Mon Dec 24, 2018 3:00 pm
longinvest wrote:
Mon Dec 24, 2018 2:29 pm
I think that I could use this estimated zero-coupon yield curve to create a synthetic EDV-like fund: a zero-coupon ladder-like fund 30-to-21 years, equal-par weighted. I'll try to do that in the near future and post the result on the Historical Bond Returns - Shiller: From Rates to Returns thread.
Yeah, my first Xmas gift! :wink:

More seriously, yes, this would be terrific, but no rush. :beer
OK, I'm a little late for Xmas, but it's done and uploaded. See this post for details.
I ran a quick backtest of the zero-coupon 30-21 synthetic model against the index tracked by Vanguard EDV (i.e. Bloomberg Barclays 20-30Y Treasury Strips TR USD). Here is a growth chart expressed in real terms (in other words, a telltale chart against inflation) comparing the two trajectories since the index' inception. I also included the index tracked by Vanguard Long-Term Treasury Fund (VUSTX), and its corresponding model in longinvest's updated spreadsheet.

Image

The model tends to drift a tiny bit lower than reality (the index) in both cases, but this remains a really good match. Which means that what the returns computed by the model for the pre-1997 years is probably a pretty good approximation of what an LT STRIPs fund would have returned. Nice modeling job.

PS. I disabled the display of EDV itself on the chart, as it was overlapping very closely its index and was just cluttering the display. This indicates that whatever friction costs due to stripping should be very low. And that Vanguard does a really good job of tracking the index.
I haven’t looked at the spreadsheet yet, but I suspect the call feature embedded in Treasuries issued before 1985 might need to be accounted for. I don’t think any are currently outstanding, and I don’t know how often bonds were called, but that could account for some of the differential between the simulation and the index before 2015.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

linenfort
Posts: 2241
Joined: Sat Sep 22, 2007 9:22 am
Location: #96151D

Re: Do long-term bonds belong in one's portfolio?

Post by linenfort » Fri Dec 28, 2018 8:44 am

I read the first page of this thread and skimmed the rest, and I don’t see any comments addressing the quote below from the OP. (Is that you? CULater = CULlen Roche?)
CULater wrote:
Sat Dec 08, 2018 8:57 am
Here are some arguments for holding long term bonds
...
These are good lessons to remember. I like to think of the stock market as a 25 year high quality bond that will pay us about 7% per year if we wait it out. But you have to be willing to wait several decades to see that come to fruition with any certainty. Diversifying our portfolios across the bond market is how we reduce our average waiting time (and uncertainty) without reducing it to nothing (cash) and earning nothing. So, just as it never makes sense to own an all-or-nothing position in stocks it also makes no sense to own an all-or-nothing position in the bond market.
https://www.pragcap.com/3-reasons-hold- ... tes-rise/

...
Perhaps because that part is more about stocks. Is there a thread someone can point me to where the stocks-as-longbonds idea is debated? It sounds very Jeremy Siegel and I would like to believe it, but I no longer can. Shouldn’t that mid-paragraph sentence end “...and without any certainty”?

There are great arguments for owning stocks, and yet you can find the value declines by 60% in year 25, just as you retire, if you are unlucky.

Yes, inflation is a concern, but if you buy 30-year U.S. treasurys, you will get interest and in thirty years, you will get your principal back. Period. Not a 60% decline.

longinvest
Posts: 4085
Joined: Sat Aug 11, 2012 8:44 am

Re: Do long-term bonds belong in one's portfolio?

Post by longinvest » Fri Dec 28, 2018 9:10 am

Here's my attempt to uncover the implied cost (or yield differential) of STRIPS.

Here's what I did:
  • I downloaded EDV's lastest Annual Report 8/31/2018.
  • I chose a principal strip maturing on 8/15/48 (just a bit less than 30 years away) and noted its Market Value ($6,094,000) and Face Amount ($15,050,000).
  • I extracted the Constant Maturity Treasury Rates on 8/31/2018 from FRED for all available maturities: 1 year (2.46%), 2 years (2.62%), 3 years (2.70%), 5 years (2.74%), 7 years (2.81%), 10 years (2.86%), 20 years (2.95%), and 30 years (3.02%).
  • I used a linear approximation for all missing maturities between 1 and 30 years.
  • I derived zero-coupon yields for all 30 maturities.
  • I used a linear approximation to get the zero-coupon yield for a maturity of 30 years minus half a month.
  • I compared the resulting yield with the yield on the principal strip.
Here's the table of normal and zero-coupon yields that I got:

Code: Select all

Maturity Yield  Zero Yield
   1     2.46000  2.46000
   2     2.62000  2.62210
   3     2.70000  2.70363
   4     2.72000  2.72357
   5     2.74000  2.74401
   6     2.77500  2.78091
   7     2.81000  2.81824
   8     2.82667  2.83568
   9     2.84333  2.85343
   10    2.86000  2.87148
   11    2.86900  2.88089
   12    2.87800  2.89052
   13    2.88700  2.90034
   14    2.89600  2.91036
   15    2.90500  2.92055
   16    2.91400  2.93091
   17    2.92300  2.94144
   18    2.93200  2.95214
   19    2.94100  2.96302
   20    2.95000  2.97406
   21    2.95700  2.98253
   22    2.96400  2.99116
   23    2.97100  2.99995
   24    2.97800  3.00889
   25    2.98500  3.01799
   26    2.99200  3.02725
   27    2.99900  3.03668
   28    3.00600  3.04627
   29    3.01300  3.05604
   30    3.02000  3.06598
From this, I got the linear approximation: (3.05604% + (11.5/12) X (3.06598% - 3.05604%) = 3.06557% for a maturity of 29 years 11.5 months. The yield on the similar strip principal is 3.06377%. The difference in yields is (3.06377% - 3.06557%) = 0.0018%. That's 2 tenths of one basis point. It's smaller than the errors due to rounding and using annual coupons in the zero yield calculations. The cumulative drift of a 0.0018% lag would be (((1 - 0.000018) ^30) - 1) = -0.05% in 30 years.

So, at least for the specific strip principal maturing on 8/15/2048, I can say that I'm confident to say that its embedded stripping fees seem to effectively be insignificant.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW

User avatar
hdas
Posts: 1327
Joined: Thu Jun 11, 2015 8:24 am

Re: Do long-term bonds belong in one's portfolio?

Post by hdas » Fri Dec 28, 2018 9:43 am

longinvest wrote:
Fri Dec 28, 2018 9:10 am
Here's my attempt to uncover the implied cost (or yield differential) of STRIPS.

Here's what I did:
  • I downloaded EDV's lastest Annual Report 8/31/2018.
  • I chose a principal strip maturing on 8/15/48 (just a bit less than 30 years away) and noted its Market Value ($6,094,000) and Face Amount ($15,050,000).
  • I extracted the Constant Maturity Treasury Rates on 8/31/2018 from FRED for all available maturities: 1 year (2.46%), 2 years (2.62%), 3 years (2.70%), 5 years (2.74%), 7 years (2.81%), 10 years (2.86%), 20 years (2.95%), and 30 years (3.02%).
  • I used a linear approximation for all missing maturities between 1 and 30 years.
  • I derived zero-coupon yields for all 30 maturities.
  • I used a linear approximation to get the zero-coupon yield for a maturity of 30 years minus half a month.
  • I compared the resulting yield with the yield on the principal strip.
Here's the table of normal and zero-coupon yields that I got:

Code: Select all

Maturity Yield  Zero Yield
   1     2.46000  2.46000
   2     2.62000  2.62210
   3     2.70000  2.70363
   4     2.72000  2.72357
   5     2.74000  2.74401
   6     2.77500  2.78091
   7     2.81000  2.81824
   8     2.82667  2.83568
   9     2.84333  2.85343
   10    2.86000  2.87148
   11    2.86900  2.88089
   12    2.87800  2.89052
   13    2.88700  2.90034
   14    2.89600  2.91036
   15    2.90500  2.92055
   16    2.91400  2.93091
   17    2.92300  2.94144
   18    2.93200  2.95214
   19    2.94100  2.96302
   20    2.95000  2.97406
   21    2.95700  2.98253
   22    2.96400  2.99116
   23    2.97100  2.99995
   24    2.97800  3.00889
   25    2.98500  3.01799
   26    2.99200  3.02725
   27    2.99900  3.03668
   28    3.00600  3.04627
   29    3.01300  3.05604
   30    3.02000  3.06598
From this, I got the linear approximation: (3.05604% + (11.5/12) X (3.06598% - 3.05604%) = 3.06557% for a maturity of 29 years 11.5 months. The yield on the similar strip principal is 3.06377%. The difference in yields is (3.06377% - 3.06557%) = 0.0018%. That's 2 tenths of one basis point. It's smaller than the errors due to rounding and using annual coupons in the zero yield calculations. The cumulative drift of a 0.0018% lag would be (((1 - 0.000018) ^30) - 1) = -0.05% in 30 years.

So, at least for the specific strip principal maturing on 8/15/2048, I can say that I'm confident to say that its embedded stripping fees seem to effectively be insignificant.
Kudos! for bringing objective analysis and numbers to the table to a worthwhile discussion while arriving at conclusions that contradicted your original bias. Very admirable. H
"whenever there is a randomized way of doing something, then there is a nonrandomized way that delivers better performance but requires more thought" ET Jaynes

longinvest
Posts: 4085
Joined: Sat Aug 11, 2012 8:44 am

Re: Do long-term bonds belong in one's portfolio?

Post by longinvest » Fri Dec 28, 2018 1:40 pm

On another thread, I replied to a post about the potential behavior of EDV in two specific hypothetical scenarios. This prompted me to compare its behavior to a consol (a perpetual bond). The result of the comparison is quite intriguing.

First, here's my post on that other thread:
longinvest wrote:
Fri Dec 28, 2018 12:58 pm
longinvest wrote:
Fri Dec 28, 2018 10:56 am
boglerdude wrote:
Fri Dec 28, 2018 2:53 am
Say you buy EDV. $10k 2% rate when you buy it. So you have a "bond" and the duration is always 25 years

What's your return if rates go to 10% and stay there for 20 years?

What if rates go to 1% for 20 years?

Is there a web calculator where you plug in different hold times and average rates over that timeframe. Total return including dividends
If you're up to it, just enter a full range of yields across the curve for every year of your scenario into our bond fund simulator and tell us about the outcome. Luckily, the newly released 1.18 version has added a simulation for an EDV-like fund.
OK. I did the exercise.

Scenario 1

Let's assume that yields, across the yield curve, from 1 year to 30 years, are initially 2% for all maturities. One year later, yields , across the yield curve, from 1 year to 30 years, have increased to 10% for all maturities. An EDV-like bond fund, making no distributions (as it's composed of zero-coupon strips) would lose -83.38% of its value during that year. It would immediately start returning +10%, annually, for as long as yields remain fixed at 10%.

So, that's a cumulative (((1 - 0.8338) X (1.10 ^ 20)) - 1) = +11.81% gain in 21 years (+0.53% annualized).

Scenario 2

Let's assume that yields, across the yield curve, from 1 year to 30 years, are initially 2% for all maturities. One year later, yields , across the yield curve, from 1 year to 30 years, have decreased to 1% for all maturities. An EDV-like bond fund, making no distributions (as it's composed of zero-coupon strips) would gain +29.69% in value during that year. It would immediately start returning +1%, annually, for as long as yields remain fixed at 1%.

So, that's a cumulative (((1 + 0.2969) X (1.01 ^ 20)) - 1) = +58.25% gain in 21 years (+2.21% annualized).
What's important to me, in this post, are the 1-year impact of the 2% to 10% increase in yields (-83.38%) and the 1-year impact of the 2% to 1% decrease in yields (+29.69%).

If, instead of holding an EDV-like fund, in such hypothetical scenarios, our investor held a consol with a $100 annual coupon. What would have happened?

Initially, yields are 2% across the yield curve. The consol pays $100 per year. The price of the consol is: ($100 / 0.02) = $5,000.

Scenario 1

Yields increase to 10% during the year: At the end of the year, the consol pays its $100 coupon. Its price has changed. It has become: ($100 / 0.10) = $1,000. The 1-year total return of the consol is thus: ((($1,000 + $100) / $5,000) - 1) = -78.00%.

Scenario 2

Yields decrease to 1% during the year: At the end of the year, the consol pays its $100 coupon. Its price has changed. It has become: ($100 / 0.01) = $10,000. The 1-year total return of the consol is thus: ((($10,000 + $100) / $5,000) - 1) = +102.00%.

Comparison

In scenario 1, the EDV-like fund lost a little more (-83.38%), on a total return basis, than the consol (-78.00%). Yet, in scenario 2, it gained significantly less (+29.69%) than the consol (+102.00%).

I found this intriguing. Given a choice of investing into either EDV or a consol, what would you choose? Why? (Please provide numerical justifications).
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW

HEDGEFUNDIE
Posts: 3973
Joined: Sun Oct 22, 2017 2:06 pm

Re: Do long-term bonds belong in one's portfolio?

Post by HEDGEFUNDIE » Fri Dec 28, 2018 2:43 pm

longinvest wrote:
Fri Dec 28, 2018 1:40 pm
I found this intriguing. Given a choice of investing into either EDV or a consol, what would you choose? Why? (Please provide numerical justifications).
Isn’t this a moot question? US Treasury consols don’t exist. I’ve heard speculation of a 50 year treasury but not consols.

Post Reply