Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Mon Oct 23, 2017 8:01 pm

Frequently it's said that the bond market prices in information about future Fed funds rate increases and other action, potential Fed unwinding of the balance sheets, economic conditions, growth, etc. And that's true—the information is known to those transacting in the markets and is reflected in pricing. However, that doesn't tell us exactly why they're settling for the current prices. Are bond investors at any given moment more scared of reinvestment risk or inflation risk, or something else? What's the balance? How is the term risk being priced?

Overall this is related to the concept of the term premium and why on average we've seen higher returns from longer-term bonds than shorter-term bonds in the 20th century. It's a driving force behind why one would use bonds rather than cash to derisk portfolios.

As Ben Bernanke put it in a 2015 article here,
To explain the behavior of longer-term rates, it helps to decompose the yield on any particular bond, such as a Treasury bond issued by the US government, into three components: expected inflation, expectations about the future path of real short-term interest rates, and a term premium.
...
Briefly, a term premium is the extra return that lenders demand to hold a longer-term bond instead of investing in a series of short-term securities (a new one-year security each year, for example). Typically, long-term yields are higher than short-term yields, implying that term premiums are usually positive (investors require extra compensation to hold longer-term bonds instead of short-term securities).
The term premia of Treasuries across the yield curve cannot be directly observed, but can be estimated. The higher the value, the lower the price (higher yield) for longer-term bonds relative to what you'd get investing in short-term bonds and rolling those over.

One of the more-cited models for the yield curve term structure these days, building on decades of work since the expectations hypothesis and other earlier ideas that are widely regarded as empirically untrue now, is from Adrian, Crump, and Moench (ACM) of the New York Fed. They have a post here with some details, a link to updated data, a link to a paper explaining the estimation methodology, and a comparison to some other models and estimates.

For example, one alternative way to derive a term premium estimate would be to take expert surveys of expected short-term interest rate paths and compare those to current longer-term rates. Not all of the methods and models agree, and any one model may be substantially wrong.

All that said, I thought it would be interesting to go back and check what the ACM model is saying these days. I haven't checked it in a while. It does more than just estimate the term premia, but that's one of the things that falls out of it (as the difference between fitted Treasury yields and the risk neutral yield).

So I downloaded the data and graphed for the 2-year, 5-year, and 10-year Treasury bonds:

Image
(updated 2018-05-15, latest data point 2018-05-10)

As of the end of 2016, estimated term premia of -0.46% for the 2-year, -0.49% for the 5-year, -0.55% for the 10-year, with FFR in the 1.25-1.50% zone and between 2-3 increases expected in 2018. That's with 2-year, 5-year, and 10-year rates at 1.87%, 2.18%, and 2.43% (you'll see slightly different figures from other sources).

Turns out we're back to negative term premia according to this particular model, which effectively posits that currently the market is pricing Treasury bond yields so high that holding short-term T-bills would be expected to return more over time (averaged over potential futures). I can't tell you if this is actually correct, and I have no comments on the methodology for now, but I thought it was worth at least a look and possibly some discussion. No, I am not suggesting wholesale market timing.
Last edited by lack_ey on Tue May 15, 2018 10:37 am, edited 13 times in total.

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Sat Nov 18, 2017 1:47 pm

Updating. Also to bump for a little more exposure, especially given some other threads around.

User avatar
patrick013
Posts: 2172
Joined: Mon Jul 13, 2015 7:49 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by patrick013 » Sat Nov 18, 2017 2:23 pm

There's several articles every month but they all point to similar
observations depending of who has written them.

Federal Debt held by the public should rise to 90% of GDP.

Increased demand will keeps spreads down to 1% for TRSY 10
for the foreseeable future.

FFR will rise peaking in several years.

Projections of Interest Rates | Congressional Budget Office
The Budget and Economic Outlook: 2017 to 2027 | Congressional Budget Office

So as the FFR rises we will not see a 2% spread for awhile. Other economic factors
are just being trumped by more and more demand. If the FFR spikes or crashes that
would certainly be a new development. If the money going to TRSY's went to stocks
in confidence that would increase TRSY spreads as demand lessens. That's how it
makes sense to me.
age in bonds, buy-and-hold, 10 year business cycle

User avatar
SimpleGift
Posts: 2936
Joined: Tue Feb 08, 2011 3:45 pm
Location: Central Oregon

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by SimpleGift » Sat Nov 18, 2017 2:56 pm

lack_ey wrote:
Mon Oct 23, 2017 8:01 pm
Are bond investors at any given moment more scared of reinvestment risk or inflation risk, or something else? What's the balance? How is the term risk being priced?
Isn't this a case of the U.S. Fed raising short term rates, but international investors piling into longer term Treasuries (driving their prices up and their yields down) — since their home country government bond yields are so low? This table shows the recent yields of longer term bonds in Europe, Japan and the U.S.:
In a world where there's relatively free movement of capital between countries, global bond investors are not just looking at U.S. reinvestment risk or inflation risk (though a consideration, for sure), but rather how the U.S. debt market compares with other developed country debt markets. Right now, U.S. Treasury yields are quite compelling.

If this is right, then as long as central banks in Europe and Japan are still "quantitatively easing" in their domestic economies — and the U.S. Fed is raising short term rates — the U.S. term premium may continue to be low or negative.
Cordially, Todd

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Sat Nov 18, 2017 3:18 pm

SimpleGift wrote:
Sat Nov 18, 2017 2:56 pm
Isn't this a case of the U.S. Fed raising short term rates, but international investors piling into longer term Treasuries (driving their prices up and their yields down) — since their home country government bond yields are so low? This table shows the recent yields of longer term bonds in Europe, Japan and the U.S.:
In a world where there's relatively free movement of capital between countries, global bond investors are not just looking at U.S. reinvestment risk or inflation risk (though a consideration, for sure), but rather how the U.S. debt market compares with other developed country debt markets. Right now, U.S. Treasury yields are quite compelling.

If this is right, then as long as central banks in Europe and Japan are still "quantitatively easing" in their domestic economies — and the U.S. Fed is raising short term rates — the U.S. term premium may continue to be low or negative.
When I checked a few of the countries, yields were similar after currency hedging, and generally it shouldn't be far off. So I don't know how much pricing pressure is actually from that. They're buying some bonds but what percentage of the market, anyway? I think term premia may be low across a number of countries, though if the US is likely to raise short-term rates more and faster, maybe lower in the US. Which if anything implies that the local bonds may be a better deal for the foreign investors, right, as least where the yield curves are not as flat.

Do you know of where to find stats on net change in ownership (who owns what) of Treasury bonds or other parts of the bond market?

User avatar
SimpleGift
Posts: 2936
Joined: Tue Feb 08, 2011 3:45 pm
Location: Central Oregon

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by SimpleGift » Sat Nov 18, 2017 3:50 pm

lack_ey wrote:
Sat Nov 18, 2017 3:18 pm
Do you know of where to find stats on net change in ownership (who owns what) of Treasury bonds or other parts of the bond market?
One can find monthly changes in foreign holdings of U.S. Treasuries here.

There looks to be both recent and historical data, broken down by individual foreign countries — but I haven't spent much time looking around at these data sets.
Cordially, Todd

lazyday
Posts: 3296
Joined: Wed Mar 14, 2007 10:27 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lazyday » Sat Nov 18, 2017 4:38 pm

lack_ey wrote:
Mon Oct 23, 2017 8:01 pm
One of the more-cited models for the yield curve term structure these days, building on decades of work since the expectations hypothesis and other earlier ideas that are widely regarded as empirically untrue now, is from Adrian, Crump, and Moench (ACM) of the New York Fed. They have a post here with some details, a link to updated data, a link to a paper explaining the estimation methodology, and a comparison to some other models and estimates.
Is ACMTP02 the modeled term premium for a 2 year Treasury note?

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Sat Nov 18, 2017 4:47 pm

lazyday wrote:
Sat Nov 18, 2017 4:38 pm
Is ACMTP02 the modeled term premium for a 2 year Treasury note?
Yeah, that's how it's structured. ACM model, TP term premium, 02 year Treasury. It's the top green line in the graph I posted.
SimpleGift wrote:
Sat Nov 18, 2017 3:50 pm
One can find monthly changes in foreign holdings of U.S. Treasuries here.

There looks to be both recent and historical data, broken down by individual foreign countries — but I haven't spent much time looking around at these data sets.
Great, thanks.

User avatar
patrick013
Posts: 2172
Joined: Mon Jul 13, 2015 7:49 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by patrick013 » Sat Nov 18, 2017 9:23 pm

Well it's interesting how they went from OLS to a 5 factor
system. How they got the fitted line so close is beyond
me.
age in bonds, buy-and-hold, 10 year business cycle

User avatar
Doc
Posts: 8134
Joined: Sat Feb 24, 2007 1:10 pm
Location: Two left turns from Larry

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by Doc » Sun Nov 19, 2017 9:32 am

patrick013 wrote:
Sat Nov 18, 2017 9:23 pm
Well it's interesting how they went from OLS to a 5 factor
system. How they got the fitted line so close is beyond
me.
There was a saying in school about regressions.

Something like with one factor you can draw a line, with two you can draw a curve ... with four you can draw an elephant and with five you can make that elephant wag it's tail.
A scientist looks for THE answer to a problem, an engineer looks for AN answer and lawyers ONLY have opinions. Investing is not a science.

User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by in_reality » Sun Nov 19, 2017 10:21 am

lack_ey wrote:
Sat Nov 18, 2017 3:18 pm
SimpleGift wrote:
Sat Nov 18, 2017 2:56 pm
Isn't this a case of the U.S. Fed raising short term rates, but international investors piling into longer term Treasuries (driving their prices up and their yields down) — since their home country government bond yields are so low? This table shows the recent yields of longer term bonds in Europe, Japan and the U.S.:
In a world where there's relatively free movement of capital between countries, global bond investors are not just looking at U.S. reinvestment risk or inflation risk (though a consideration, for sure), but rather how the U.S. debt market compares with other developed country debt markets. Right now, U.S. Treasury yields are quite compelling.

If this is right, then as long as central banks in Europe and Japan are still "quantitatively easing" in their domestic economies — and the U.S. Fed is raising short term rates — the U.S. term premium may continue to be low or negative.
When I checked a few of the countries, yields were similar after currency hedging, and generally it shouldn't be far off. So I don't know how much pricing pressure is actually from that. They're buying some bonds but what percentage of the market, anyway? I think term premia may be low across a number of countries, though if the US is likely to raise short-term rates more and faster, maybe lower in the US. Which if anything implies that the local bonds may be a better deal for the foreign investors, right, as least where the yield curves are not as flat.

Do you know of where to find stats on net change in ownership (who owns what) of Treasury bonds or other parts of the bond market?
Why would foreign investors need to hedge though? Doing so will reduce their yields. Currency carry trading would have them leveraged and unhededged. And heck 10-1 leverage on hedged holdings in many cases is signifantly better than local rates.

garlandwhizzer
Posts: 1825
Joined: Fri Aug 06, 2010 3:42 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by garlandwhizzer » Sun Nov 19, 2017 11:21 am

It seems to me that the bottom line for investors is that at present you aren't getting paid much to take on duration risk in the bond market. The spreads between short term and long term bonds have been narrowing, just like the spreads between highest and lowest quality bonds. Add to this the expected loss of bond principal value in a rising rate environment which increases directly with duration. Taking on duration risk in the bond market has been well rewarded for 30+ years as both inflation and interest rates have relentlessly declined from historical highs (1982) to historical lows (excepting the Great Depression for inflation). This has given a consistent boost to returns from principal appreciation in longer duration bonds.

No one knows the future, but we may be at or near an inflection point where both inflation and interest rates begin to increase steadily and slowly for a long time, possibly decades. If that is the case, duration risk in the bond market may not be as well rewarded in the future as it was in the past. Likewise lower quality higher yielding bonds have done well in recent years as investors stretched for yield. Spread between high and low quality bonds are low by historical standards which means you aren't getting paid much to take on default risk either. Backtesting results in the bond market over the last 3+ decades is not IMO to be trusted as defining the future. This may be an opportune time to review your bond allocation if yield seeking in recent years has stretched you too far into duration or default risk.

Garland Whizzer

User avatar
patrick013
Posts: 2172
Joined: Mon Jul 13, 2015 7:49 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by patrick013 » Sun Nov 19, 2017 2:37 pm

Doc wrote:
Sun Nov 19, 2017 9:32 am
patrick013 wrote:
Sat Nov 18, 2017 9:23 pm
Well it's interesting how they went from OLS to a 5 factor
system. How they got the fitted line so close is beyond
me.
There was a saying in school about regressions.

Something like with one factor you can draw a line, with two you can draw a curve ... with four you can draw an elephant and with five you can make that elephant wag it's tail.
The averages of the averages...but he's doing something else. Sticking small samples
together or using autocorrelation. Which does give me an idea about something else.
Still well beyond me.
age in bonds, buy-and-hold, 10 year business cycle

User avatar
Doc
Posts: 8134
Joined: Sat Feb 24, 2007 1:10 pm
Location: Two left turns from Larry

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by Doc » Sun Nov 19, 2017 3:51 pm

patrick013 wrote:
Sun Nov 19, 2017 2:37 pm
by patrick013 » Sun Nov 19, 2017 1:37 pm

Doc wrote: ↑Sun Nov 19, 2017 8:32 am
patrick013 wrote: ↑Sat Nov 18, 2017 8:23 pm
Well it's interesting how they went from OLS to a 5 factor
system. How they got the fitted line so close is beyond
me.
There was a saying in school about regressions.

Something like with one factor you can draw a line, with two you can draw a curve ... with four you can draw an elephant and with five you can make that elephant wag it's tail.
The averages of the averages...but he's doing something else. Sticking small samples
together or using autocorrelation. Which does give me an idea about something else.
Still well beyond me.
I found the five factor quote:
With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.
Attributed to von Neumann by Enrico Fermi, as quoted by Freeman Dyson in "A meeting with Enrico Fermi" in Nature 427 (22 January 2004) p. 297
https://en.wikiquote.org/wiki/John_von_Neumann

This has little to do with where the term premia is or where it might go. It is just an comment on regression analysis beyond the limits of credulity.
A scientist looks for THE answer to a problem, an engineer looks for AN answer and lawyers ONLY have opinions. Investing is not a science.

User avatar
Doc
Posts: 8134
Joined: Sat Feb 24, 2007 1:10 pm
Location: Two left turns from Larry

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by Doc » Sun Nov 19, 2017 4:08 pm

From The Only Guide to a Winning Bond Strategy You'll Ever Need, Larry E. Swedroe & Joseph H. Hempen p73
Fama and Bliss concluded that today’s yield curve contains information about future yield curves – current forward rates provide the best forecast of future spot interest rates. In other words, today’s yield curve is the best estimate we have of what future yield curves will be.
"But not a very good one." Swedroe again maybe from Bogleheads forum?

In any case I use today's yield curve for future projections without regard to risk, term or inflation premia as separate components.
A scientist looks for THE answer to a problem, an engineer looks for AN answer and lawyers ONLY have opinions. Investing is not a science.

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Sun Nov 19, 2017 4:14 pm

patrick013 wrote:
Sun Nov 19, 2017 2:37 pm
The averages of the averages...but he's doing something else. Sticking small samples
together or using autocorrelation. Which does give me an idea about something else.
Still well beyond me.
There's a decomposition involved into uncorrelated principal components, which further increases the ability to span the data.

Apparently they build on a prior paper with four factors (and other similar research) to explain term structure and claim that the fifth is necessary.
Doc wrote:
Sun Nov 19, 2017 4:08 pm
From The Only Guide to a Winning Bond Strategy You'll Ever Need, Larry E. Swedroe & Joseph H. Hempen p73
Fama and Bliss concluded that today’s yield curve contains information about future yield curves – current forward rates provide the best forecast of future spot interest rates. In other words, today’s yield curve is the best estimate we have of what future yield curves will be.
"But not a very good one." Swedroe again maybe from Bogleheads forum?

In any case I use today's yield curve for future projections without regard to risk, term or inflation premia as separate components.
These papers are all past Fama-Bliss. It's no longer really thought that "today's yield curve is the best estimate of what future yield curves will be" which is kind of the stay-the-same hypothesis, implying a certain random walk. It's just that earlier ideas such as expectations hypothesis are empirically even worse.

That said none of the forecasting models are particularly good.
in_reality wrote:
Sun Nov 19, 2017 10:21 am
Why would foreign investors need to hedge though? Doing so will reduce their yields. Currency carry trading would have them leveraged and unhededged. And heck 10-1 leverage on hedged holdings in many cases is signifantly better than local rates.
That's fair enough, was just speculating on covered interest rate parity and then assuming uncovered may not be way far off that in terms of which investments are obviously better.

User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by in_reality » Sun Nov 19, 2017 9:34 pm

lack_ey wrote:
Sun Nov 19, 2017 4:14 pm
in_reality wrote:
Sun Nov 19, 2017 10:21 am
Why would foreign investors need to hedge though? Doing so will reduce their yields. Currency carry trading would have them leveraged and unhededged. And heck 10-1 leverage on hedged holdings in many cases is signifantly better than local rates.
That's fair enough, was just speculating on covered interest rate parity and then assuming uncovered may not be way far off that in terms of which investments are obviously better.
Despite the theory, it seems that the expectation of UIP fails.

In a simple, risk neutral setting, uncovered interest parity (UIP) predicts that the currency of a
country with a high interest rate is expected to depreciate so that the profit is the same as from
investing in a low interest rate bond. Empirically, UIP fails and the expected profit from the carry
trade (investing in a high interest rate bond abroad, and then converting back to the home currency)
is positive.

http://www.lse.ac.uk/fmg/events/favilukisFGN.pdf


The carry trade, one of the oldest and most popular currency speculation strategies, is
motivated by the failure of uncovered interest parity (UIP) documented by Bilson (1981)
and Fama (1984).1 This strategy has received a great deal of attention in the academic
literature as researchers struggle to explain its apparent profitability

https://www.kellogg.northwestern.edu/fa ... /carry.pdf

Understanding the behaviour of currency markets has been an active area of research for the past few decades. Much of the literature has focused on the marginal behaviours of exchange rates and carry trade portfolio returns resulting from the established violations of uncovered interest rate parity. Such investment strategies are popular approaches which involve constructing portfolios by selling low interest rate currencies in order to buy high interest rate currencies, thus profiting from the interest rate differentials.

Our model is shown to provide superior risk-adjusted returns for a currency carry trade strategy over the period 1999 - 2014

http://ieeexplore.ieee.org/abstract/doc ... eload=true

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Sun Nov 19, 2017 9:43 pm

Yeah, I know there's a lot of evidence and live trading results on currency carry, with UIP failing. I think that was the wrong point to make earlier, and mostly was just meaning to ask about relative Treasury flows to different holders.

My question is how much the yield differential is really worth given the uncovered currency exposure with the FX risk (investing in USD when you want some other currency). That seems like a relatively low Sharpe ratio trade for foreign investors, unless they also want to bet that FX movements will go their way. US rates aren't that much higher, and there's a decent chance there will be greater inflation in the US than in some of the low-yielding countries. How much pricing pressure are they really exerting on US bonds across the yield curve? Would the 10-year be at 4% were it not for foreign investors? 3%? Not even that much? What's actually the impact?

User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by in_reality » Sun Nov 19, 2017 11:01 pm

lack_ey wrote:
Sun Nov 19, 2017 9:43 pm
How much pricing pressure are they really exerting on US bonds across the yield curve? Would the 10-year be at 4% were it not for foreign investors? 3%? Not even that much? What's actually the impact?
I'm not sure, nor am I sure that empirically it's known. This 7.2017 research suggests that for spot rates, it depends on the country and that for spot rates it's not relevant to the US.
These findings contradict the conventional view that sudden large movements in exchange rates are attributable to the carry trade. They suggest, instead, that the effects of the global carry trade are primarily concentrated in bond markets.
We believe that these interest-rate effects are due to the greater importance of carry trade
transactions in bond order flows. Future research using bond flows will determine whether
this conjecture is correct.
However, we do note that we could find no evidence of significant carry trade effects on the differential
between U.S. and E.U (Eurocurrency) interest rates. These are exactly the money markets where non-carry
factors driving order flows are likely to swamp the effects of the carry trade.
https://papers.ssrn.com/sol3/papers.cfm ... id=3057946

User avatar
patrick013
Posts: 2172
Joined: Mon Jul 13, 2015 7:49 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by patrick013 » Mon Nov 20, 2017 6:34 pm

lack_ey wrote:
Sun Nov 19, 2017 9:43 pm
How much pricing pressure are they really exerting on US bonds across the yield curve? Would the 10-year be at 4% were it not for foreign investors? 3%? Not even that much? What's actually the impact?
Foreign investors are buying about 22% of all TRSY sales. So it's in the trillions.
Certainly buying up the higher yielders across all maturities I suspect.
So it's interest rates or just plain imports using foreign currencies don't really
follow it, probably should.
age in bonds, buy-and-hold, 10 year business cycle

User avatar
patrick013
Posts: 2172
Joined: Mon Jul 13, 2015 7:49 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by patrick013 » Tue Nov 21, 2017 3:26 pm

https://myphotos.mypclinuxos.com/images ... sed660.png

My little summary of spread history shows we're definitely not in the
better frequency of longer term premium observations. :)
age in bonds, buy-and-hold, 10 year business cycle

User avatar
BlueEars
Posts: 3592
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by BlueEars » Tue Nov 21, 2017 9:51 pm

Speaking of the Treasury term premium, I recently took the Simba data and looked at how the short term Treasuries did versus the intermediate Treasuries in the 1950's through the early 1970's. Seems that the short term Treasuries returned more then the intermediates more often then not as rates rose erratically.

In this chart I took the 5 year rolling returns for each bond type. Then subtracted the returns to plot the 5 year rolling return differentials i.e. intermediate Treasury rolling return minus short term Treasury rolling return:

Image

and here is how the 5yr constant maturity Treasury behaved during those decades: https://fred.stlouisfed.org/series/GS5
I could not seem to post the image now, so this URL will have to do.
.

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Tue Jan 02, 2018 7:53 pm

Updated it today. It still shows term premia from 1-year Treasury to 10-year in the -0.36% to -0.55% range. The 1-year result implies the risk-neutral rate is 2.1%.

Is it just me or does it seem a bit high, given where the short rates are now? Even if there's a bump of about 0.75% through the year, it seems hard to get a cash/T-bills return up to 2.1% this year.

In any case, this is just one model of many that could be devised.

lazyday
Posts: 3296
Joined: Wed Mar 14, 2007 10:27 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lazyday » Wed Jan 03, 2018 12:49 am

Thanks for the update.

From another thread, David Swensen has cut duration:
as rate came down, we systematically reduced the duration of our bond portfolio. And today it’s probably nine months or a year. And so we don’t think there is any point in—hey, Marty. (Laughter.) My first boss from Salomon Brothers [gesturing to someone in audience] laughing because I told him I would never make any interest rate bets. And here I am, describing an interest rate bet to a room full of people. (Laughter.) We don’t see any point of owning 30-year Treasurys if the yield is 2 ½ percent.
37:20 in the video. viewtopic.php?f=10&t=232479

He mentions low rates, but not the term premium. I would think that if you're going to market time duration, an estimate of the term premium would be important. But I haven't seen much discussion on it.

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Wed Jan 03, 2018 1:31 am

lazyday wrote:
Wed Jan 03, 2018 12:49 am
Thanks for the update.

From another thread, David Swensen has cut duration:
as rate came down, we systematically reduced the duration of our bond portfolio. And today it’s probably nine months or a year. And so we don’t think there is any point in—hey, Marty. (Laughter.) My first boss from Salomon Brothers [gesturing to someone in audience] laughing because I told him I would never make any interest rate bets. And here I am, describing an interest rate bet to a room full of people. (Laughter.) We don’t see any point of owning 30-year Treasurys if the yield is 2 ½ percent.
37:20 in the video. viewtopic.php?f=10&t=232479

He mentions low rates, but not the term premium. I would think that if you're going to market time duration, an estimate of the term premium would be important. But I haven't seen much discussion on it.
Market timing is always dubious or speculative at best, especially based on single measures.

But I agree that if your'e going to do any (hopefully not huge changes, regardless), it seems like you should have an idea about what the market is pricing in.

In terms of strategic asset allocation, we think of bonds as a diversifier for equities. The reason to use bonds instead of cash is to pick up additional return without increasing risk appreciably (of course you could just adjust stock/bond ratio to match the risk of a stock/cash allocation). You get an excess return over cash because term risk and credit risk are rewarded and furthermore distinct from equity risk—though there's of course more relationship between [credit and equity] than [term and equity].

But if the market is not pricing Treasury bonds for excess expected return above cash/T-bills (of course still with the possibility for higher return), then what's the rationale for this term risk diversification that might actually be deworsification? I guess if we're counting on recent trends of negative correlation with equities to continue, high quality bonds then have negative market beta and are an actual hedge, so you're paying extra to be able to hedge equity risk. If the economy tanks and short-term rates have to be dropped to zero again (or lower), okay, that would be probably bad for stocks and good for excess returns of bonds over cash. But this all seems a bit unsatisfying.

People sometimes think about this in terms of an inverted yield curve, but there can be a positive term premium priced in during an inverted yield curve, if short rates are expected to drop. Here even with an upwards sloping yield curve, there seems to be little term premium priced in, maybe even negative as the model says.

I don't mind the idea of taking risks—real risk, actually getting crushed sometimes, buying and holding on, with some chance of earning some extra return. Maybe we don't even get the extra return. But what if, instead of earning some term risk premium as we're told to do, we're all on the side buying insurance in the form of bonds and term risk?

Of course, it's not like a fractional percentage point of (expected) return in a fraction of one's asset allocation is a huge deal anyway. This is more for my own understanding than anything else.

Furthermore, there's always the issue that even if the market is pricing something in, that doesn't mean the market is even correct. It's frequently wrong about rates.

User avatar
BlueEars
Posts: 3592
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by BlueEars » Wed Jan 03, 2018 10:55 am

lazyday wrote:
Wed Jan 03, 2018 12:49 am
Thanks for the update.

From another thread, David Swensen has cut duration:
as rate came down, we systematically reduced the duration of our bond portfolio. And today it’s probably nine months or a year. And so we don’t think there is any point in—hey, Marty. (Laughter.) My first boss from Salomon Brothers [gesturing to someone in audience] laughing because I told him I would never make any interest rate bets. And here I am, describing an interest rate bet to a room full of people. (Laughter.) We don’t see any point of owning 30-year Treasurys if the yield is 2 ½ percent.
37:20 in the video. viewtopic.php?f=10&t=232479

He mentions low rates, but not the term premium. I would think that if you're going to market time duration, an estimate of the term premium would be important. But I haven't seen much discussion on it.
I haven't seen a discussion of Swensen's path and reasoning towards a duration as low as a year. Seems very radical.
Have any here recently reduced duration in your bond holdings?

I have reduced to average duration = 3.7%

lazyday
Posts: 3296
Joined: Wed Mar 14, 2007 10:27 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lazyday » Thu Jan 04, 2018 1:06 pm

lack_ey wrote:
Wed Jan 03, 2018 1:31 am
But if the market is not pricing Treasury bonds for excess expected return above cash/T-bills (of course still with the possibility for higher return), then what's the rationale for this term risk diversification that might actually be deworsification? I guess if we're counting on recent trends of negative correlation with equities to continue, high quality bonds then have negative market beta and are an actual hedge, so you're paying extra to be able to hedge equity risk. If the economy tanks and short-term rates have to be dropped to zero again (or lower), okay, that would be probably bad for stocks and good for excess returns of bonds over cash. But this all seems a bit unsatisfying.
My feelings are similar.

With the 30 at 2.8% nominal, I doubt long Treasuries provide much protection against a crash. But lots of inflation risk and real interest rate risk. Not sure it’s worth it if you’re not getting a term premium.

I guess there’s an argument for matching duration to spending needs, but if you are looking at nominal bonds and real liabilities, I’m not sure that’s very useful.
Furthermore, there's always the issue that even if the market is pricing something in, that doesn't mean the market is even correct. It's frequently wrong about rates.
Maybe long bond prices are set partly by those who will own them no matter what the rate is, including those who must own them. Pension funds and insurance companies have to match liabilities, and I believe some are constrained by regulations. The fed holds bonds bought during quantitative easing. Some have claimed that China and other countries buy Treasuries without regard to rate.
Of course, it's not like a fractional percentage point of (expected) return in a fraction of one's asset allocation is a huge deal anyway. This is more for my own understanding than anything else.
I already favor CDs over Treasuries & TIPS, and a negative term premium reinforces that. As an individual investor we can sometimes get a significantly better rate than a Treasury of the same duration. Recently I’ve seen this in 2 to 5 year CDs. Someone with a liability matching portfolio might use CDs for the first 5 or 10 years of a ladder, and TIPS for the long end.

lazyday
Posts: 3296
Joined: Wed Mar 14, 2007 10:27 pm

Re: Treasury term premium estimate: negative this year (NY Fed Adrian, Crump, Moench model)

Post by lazyday » Thu Jan 04, 2018 1:17 pm

BlueEars wrote:
Wed Jan 03, 2018 10:55 am
I haven't seen a discussion of Swensen's path and reasoning towards a duration as low as a year. Seems very radical.
The reversal is interesting. I thought he argued strongly in Pioneering Portfolio Management against changing duration in response to the market, but couldn't find it in my copy.

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Mon Feb 05, 2018 3:58 pm

Updated again at the top. Here's a closer look at the last 5 years:

Image

As of last Friday, the model's estimated term premium now is around -0.265% at the 1-year to around -0.267% at the 10-year, all in a fairly narrow range between -0.22% and -0.30%.

Anyway, the estimate here is that most of the run-up of the 2-year since 2017Q3 (in mid-September, 1.26%, to today's 2.15%) is about expectations of future short rates changing, in addition to short rates actually changing and us getting several more months later. I suppose people were expecting a slower and more cautious tightening cycle then than they are now. The estimated term premium on the 2-year changed from -0.52% to -0.30% over that period.

Does that seem about right?

zeugmite
Posts: 1005
Joined: Tue Jul 22, 2008 11:48 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by zeugmite » Mon Feb 05, 2018 4:09 pm

So is term premium negative or positive now? I've seen estimates for both.

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Mon Feb 05, 2018 4:13 pm

zeugmite wrote:
Mon Feb 05, 2018 4:09 pm
So is term premium negative or positive now? I've seen estimates for both.
This model says negative across the yield curve (though only 1 yr, 2 yr, .., 10 yr is analyzed). See above. Others say different things of course.

In any case, I doubt it's very negative or very positive. People who see rates pushing 5%+ in this cycle probably would say very negative, but that seems not backed by much of anything. Of course anything's a possibility. The mean or expected outcome is rarely the actual one.

User avatar
BlueEars
Posts: 3592
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by BlueEars » Mon Feb 05, 2018 5:09 pm

One thing I track is the spread between the 10 year and 3 month Treasury. At the end of January 2018 it was 104 basis points. The median from 1987 to now is 183 bp.

So it's anemic. But I guess this is not necessarily a good indicator of forward yield differentials.

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Tue May 15, 2018 10:42 am

I saw the 10-year hit 3.06% today and remembered this, so I updated the graph at the top. Of course, their daily series is lagged a few days, so it can't capture the absolute latest changes.

Data through 2018-05-10 (last Thursday).

User avatar
patrick013
Posts: 2172
Joined: Mon Jul 13, 2015 7:49 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by patrick013 » Tue May 15, 2018 12:30 pm

Image
Most of the time the spreads are around their mean. As rates rise
the LT spread will likely be below it's mean. But it does help to
know what the mean is when looking at prices. Lately the credit
premium for LT corp was around 1% over the TRSY 20 spread.
age in bonds, buy-and-hold, 10 year business cycle

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Tue May 15, 2018 1:07 pm

Hm, the last couple posts have been about spreads.

They're good in that they're directly observable but don't distinguish between risk pricing and expected forward rates, which is kind of the point of the topic here.

Spreads on Treasury yields above FFR conditioned on the market expecting rate increases should be higher than the average spreads if other factors are held constant (they're not).

User avatar
patrick013
Posts: 2172
Joined: Mon Jul 13, 2015 7:49 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by patrick013 » Tue May 15, 2018 1:32 pm

lack_ey wrote:
Tue May 15, 2018 1:07 pm
Spreads on Treasury yields above FFR conditioned on the market expecting rate increases should be higher than the average spreads if other factors are held constant (they're not).
On average I think the term premium = mean spread. If it doesn't
something needs to be identified. Excessive demand or whatever.
The market appears to be pricing less than a year into the future
regarding expected rate increases if any. Being a more or less risk
free asset the only up to the price is the term premium. If one was
pricing AAA LT Corp bonds then a credit and term premium would need
to be evaluated to determine if the price was buyable based on knowledge
of the mean spreads for those.

I think spreads are lower when the FFR is higher but right now that isn't
the case for future pricing. No magic formula but I think the market
guess on the sustainability of a higher FFR has a big effect on pricing.
A better stat program could do it but I think it would have to be weighted
better. I think if you included prior results directly into a future estimate
you'd be factoring in the expectations of several periods into the resulting
new equation of a stat model. Simple as that, unless you had 3 Phd's and
20 years of math research to use to do a better job. :)

Just have to wait and see what occurs then.
age in bonds, buy-and-hold, 10 year business cycle

lack_ey
Posts: 6450
Joined: Wed Nov 19, 2014 11:55 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by lack_ey » Tue May 15, 2018 2:18 pm

Okay, I'm not sure I'm understanding where you're going with a lot of that, so help me out here.
patrick013 wrote:
Tue May 15, 2018 1:32 pm
On average I think the term premium = mean spread. If it doesn't
something needs to be identified. Excessive demand or whatever.
By "mean spread" what is the averaging over? Time? Over the past, or generally? If you're averaging over time, doesn't it defeat the point of trying to understand what is happening now?
patrick013 wrote:
Tue May 15, 2018 1:32 pm
The market appears to be pricing less than a year into the future
regarding expected rate increases if any. Being a more or less risk
free asset the only up to the price is the term premium.
Why would the market only price one year into the future, how do you know this, and is this typical?
patrick013 wrote:
Tue May 15, 2018 1:32 pm
I think spreads are lower when the FFR is higher but right now that isn't
the case for future pricing. No magic formula but I think the market
guess on the sustainability of a higher FFR has a big effect on pricing.
A better stat program could do it but I think it would have to be weighted
better.
Spreads are lower when FFR is higher? It usually should be the case that [Y - X] is smaller when X is larger... no real surprise there unless Y basically scales more than 1x with X.

Or did you mean more along the lines of FFR getting higher (today FFR > last year's FFR) or similar? That would seem more meaningful.

What did you mean by "a better stat program" (program as in...?) and weighting of what?

User avatar
patrick013
Posts: 2172
Joined: Mon Jul 13, 2015 7:49 pm

Re: Treasury term premium estimate: negative in most of 2017, now 2018 too (NY Fed Adrian, Crump, Moench model)

Post by patrick013 » Tue May 15, 2018 3:15 pm

lack_ey wrote:
Tue May 15, 2018 2:18 pm
Okay, I'm not sure I'm understanding where you're going with a lot of that, so help me out here.
patrick013 wrote:
Tue May 15, 2018 1:32 pm
On average I think the term premium = mean spread. If it doesn't
something needs to be identified. Excessive demand or whatever.
By "mean spread" what is the averaging over? Time? Over the past, or generally? If you're averaging over time, doesn't it defeat the point of trying to understand what is happening now?

Just monthly yield spreads over the FFR over 55 years,
and the simple average of those. Point is if spreads are high it's a good
time to buy all things else being constant. With rising rates due to FFR
changes good time to wait as overall rates should get higher regardless
of spreads.

patrick013 wrote:
Tue May 15, 2018 1:32 pm
The market appears to be pricing less than a year into the future
regarding expected rate increases if any. Being a more or less risk
free asset the only up to the price is the term premium.
Why would the market only price one year into the future, how do you know this, and is this typical?

Well it's not market pricing based on a FFR of 3% but on the current
FFR of about 2% IMO and using the mean spreads.

patrick013 wrote:
Tue May 15, 2018 1:32 pm
I think spreads are lower when the FFR is higher but right now that isn't
the case for future pricing. No magic formula but I think the market
guess on the sustainability of a higher FFR has a big effect on pricing.
A better stat program could do it but I think it would have to be weighted
better.
Spreads are lower when FFR is higher? It usually should be the case that [Y - X] is smaller when X is larger... no real surprise there unless Y basically scales more than 1x with X.

Or did you mean more along the lines of FFR getting higher (today FFR > last year's FFR) or similar? That would seem more meaningful.

If you look at a chart where the FFR is over 4% you should see lower
spreads over the FFR for TRSY's of all maturities. Of course if the FFR
stays there for decades I would expect the spreads to be average or a
little higher than average due to mean reversion for the term premium.
Usually like that but could vary due to rate spiking or whatever.


What did you mean by "a better stat program" (program as in...?) and weighting of what?

I've seen this done in some least square models. Include the result of
the dependent variable as an independent variable in the next line of
data along with the other independent variable(s). This has the effect
of smoothing or weighting the model and prior expectations should be
enumerated and weighted into the equation by doing just that, resulting
in a better fit. Worth a try, hopefully, and much simpler than the method
by the NY Fed.

Hope the TRSY chart is clearer now. Just knowing the spreads is very helpful
to me.
age in bonds, buy-and-hold, 10 year business cycle

Post Reply