High yield bonds and GNMA's in a portfolio, since 1988

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Beliavsky
Posts: 688
Joined: Sun Jun 29, 2014 10:21 am

High yield bonds and GNMA's in a portfolio, since 1988

Post by Beliavsky » Sun Sep 25, 2016 6:59 am

Many Bogleheads have recommended a 0% allocation to high yield bonds, saying bonds are for safety. According to Portfolio Visualizer, however, since 1988 (based on data availability), the highest Sharpe ratio portfolio choosing among

Portfolio Visualizer results

Code: Select all

VFINX   Vanguard 500 Index Fund
VUSTX   Vanguard Long-Term Treasury Fund
VFIIX   Vanguard GNMA Fund
VWESX   Vanguard Long-Term Investment Grade Fund Investor Shrs
VWEHX   Vanguard High Yield Corporate Fund


has been

Code: Select all

Optimized Portfolio
Allocation Ticker   Name
02.22%         VFINX   Vanguard 500 Index Fund
86.07%         VFIIX   Vanguard GNMA Fund
11.71%         VWEHX   Vanguard High Yield Corporate Fund


with Sharpe of 1.09. Admittedly, if I exclude VWEHX, the portfolio of

Code: Select all

Optimized Portfolio
Allocation   Ticker Name
06.06% VFINX   Vanguard 500 Index Fund
93.94% VFIIX   Vanguard GNMA Fund


has only a slightly lower Sharpe of 1.08. A much larger drop in Sharpe ratio, to 0.85, occurs when VFIIX is excluded:

Code: Select all

Optimized Portfolio
Allocation Ticker   Name
10.42% VFINX   Vanguard 500 Index Fund
40.15% VUSTX   Vanguard Long-Term Treasury Fund
49.43% VWEHX   Vanguard High Yield Corporate Fund


This occurs because VFIIX had the highest Sharpe ratio, 1.01. I chose candidate funds with long histories. Over the last few decades, the problem with high yield bonds, at least as represented by VWEHX, has been tax inefficency.

Should fixed income portfolios be tilted towards GNMA bonds? In optimizations they push out Treasuries and investment grade corporate bonds.

(Sorry for the misaligned tables. Is it possible to create nicely aligned ones on this site?)
Last edited by Beliavsky on Sun Sep 25, 2016 9:53 am, edited 1 time in total.

FillorKill
Posts: 1007
Joined: Sat Aug 06, 2011 7:01 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by FillorKill » Sun Sep 25, 2016 7:31 am

Beliavsky wrote:Should fixed income portfolios be tilted towards GNMA bonds? In optimizations they push out Treasuries and investment grade corporate bonds.

In future periods will the option-adjusted spread adequately or inadequately compensate you for negative convexity? That is the question....

Beliavsky
Posts: 688
Joined: Sun Jun 29, 2014 10:21 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by Beliavsky » Sun Sep 25, 2016 10:30 am

FillorKill wrote:
Beliavsky wrote:Should fixed income portfolios be tilted towards GNMA bonds? In optimizations they push out Treasuries and investment grade corporate bonds.

In future periods will the option-adjusted spread adequately or inadequately compensate you for negative convexity? That is the question....

I see a recent discussion of GNMAs you participated in, Critique Fixed Income Strategy - GNMA Dominance, where the negative convexity was mentioned.

Global high-yield bonds and GNMAs: A potentially attractive combination for volatile times by Deutsche Asset Management finds, as I did, that high yield bonds and GNMAs have been a good combination. They look at the last 10 years through June 2016.

FillorKill
Posts: 1007
Joined: Sat Aug 06, 2011 7:01 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by FillorKill » Mon Sep 26, 2016 4:37 am

Beliavsky wrote:According to Portfolio Visualizer, however, since 1988 (based on data availability), the highest Sharpe ratio portfolio choosing among....

Code: Select all

VFINX   Vanguard 500 Index Fund
VUSTX   Vanguard Long-Term Treasury Fund
VFIIX   Vanguard GNMA Fund
VWESX   Vanguard Long-Term Investment Grade Fund Investor Shrs
VWEHX   Vanguard High Yield Corporate Fund


has been

Code: Select all

Optimized Portfolio
Allocation Ticker   Name
02.22%         VFINX   Vanguard 500 Index Fund
86.07%         VFIIX   Vanguard GNMA Fund
11.71%         VWEHX   Vanguard High Yield Corporate Fund


with Sharpe of 1.09. Admittedly, if I exclude VWEHX, the portfolio of

Code: Select all

Optimized Portfolio
Allocation   Ticker Name
06.06% VFINX   Vanguard 500 Index Fund
93.94% VFIIX   Vanguard GNMA Fund


This occurs because VFIIX had the highest Sharpe ratio, 1.01. I chose candidate funds with long histories. Over the last few decades, the problem with high yield bonds, at least as represented by VWEHX, has been tax inefficency.

Should fixed income portfolios be tilted towards GNMA bonds? In optimizations they push out Treasuries and investment grade corporate bonds.

Admittedly I'm a risk-on-the-equity-side investor. In theory GNMA is a lose/lose proposition for the mortgagor and a win/win for the mortgagee. Rates drop the mortgagee exercises the prepayment option. Rates rise the mortgagee extends you. However, in practice, there is no denying that GNMA has been a solid performer for a very long time. I will add, for your consideration, if you include VSGBX, VG short term federal fund to your optimization it completely and totally displaces GNMA.

Valuethinker
Posts: 33384
Joined: Fri May 11, 2007 11:07 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by Valuethinker » Mon Sep 26, 2016 6:50 am

The data shows bonds over a long bull market. Falling interest rates generally favours bonds with higher coupons (eg HY).

During that time, credit risk will cycle in and out. There is also structural change (the 1990 HY crash was about the arrest of Michael Milliken and the collapse of DrexelBL). Generally credit risk became extreme in 2008-9 and then recovered. That plus Quantitative Easing has driven US MBS to record lows (Fed can buy bonds which are US govt guaranteed like GNMAs).

Mortgage Backed Securities you would have had considerable prepayment risk and less extension risk.

Against the background of these supercycles, independent annual data doesn't tell you as much as you would like. We almost certainly won't repeat the bull market in bonds-- interest rates cannot keep falling like they have (nominal interest rates). Credit risk? Nobody really knows.

Beliavsky
Posts: 688
Joined: Sun Jun 29, 2014 10:21 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by Beliavsky » Mon Sep 26, 2016 10:28 am

FillorKill wrote:
Beliavsky wrote:According to Portfolio Visualizer, however, since 1988 (based on data availability), the highest Sharpe ratio portfolio choosing among....

Code: Select all

VFINX   Vanguard 500 Index Fund
VUSTX   Vanguard Long-Term Treasury Fund
VFIIX   Vanguard GNMA Fund
VWESX   Vanguard Long-Term Investment Grade Fund Investor Shrs
VWEHX   Vanguard High Yield Corporate Fund


has been

Code: Select all

Optimized Portfolio
Allocation Ticker   Name
02.22%         VFINX   Vanguard 500 Index Fund
86.07%         VFIIX   Vanguard GNMA Fund
11.71%         VWEHX   Vanguard High Yield Corporate Fund


with Sharpe of 1.09. Admittedly, if I exclude VWEHX, the portfolio of

Code: Select all

Optimized Portfolio
Allocation   Ticker Name
06.06% VFINX   Vanguard 500 Index Fund
93.94% VFIIX   Vanguard GNMA Fund


This occurs because VFIIX had the highest Sharpe ratio, 1.01. I chose candidate funds with long histories. Over the last few decades, the problem with high yield bonds, at least as represented by VWEHX, has been tax inefficency.

Should fixed income portfolios be tilted towards GNMA bonds? In optimizations they push out Treasuries and investment grade corporate bonds.

Admittedly I'm a risk-on-the-equity-side investor. In theory GNMA is a lose/lose proposition for the mortgagor and a win/win for the mortgagee. Rates drop the mortgagee exercises the prepayment option. Rates rise the mortgagee extends you. However, in practice, there is no denying that GNMA has been a solid performer for a very long time. I will add, for your consideration, if you include VSGBX, VG short term federal fund to your optimization it completely and totally displaces GNMA.

Only because of the discreteness of Portfolio Visualizer calculations. You can see that the Sharpe ratio when including VSGBX falls slightly from 1.09 to 1.07, which should not happen when adding a fund choice. I have emailed the creator of the site, and he admits it is a problem, which can be solved by generating more points on the efficient frontier. Portfolio Visualizer is a fine web site, but it is worth checking its results with other programs.

User avatar
grabiner
Advisory Board
Posts: 20973
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by grabiner » Mon Sep 26, 2016 11:32 pm

FillorKill wrote:Admittedly I'm a risk-on-the-equity-side investor. In theory GNMA is a lose/lose proposition for the mortgagor and a win/win for the mortgagee.


But the rates on GNMAs allow for that, as investors are aware of the risk. GNMAs are backed by the US Government, but they yield significantly more than Treasury bonds of the same duration; investors demand the higher yields because they will gain less if rates fall and lose more if rates rise. There is no free lunch, and in this case, the free lunch would apply if you could buy Treasuries and sell equal-yielding GNMAs short.

The same phenomenon applies to callable bonds. The effective duration of a callable bond increases if rates rise (and thus it loses more than a non-callable bond), because it becomes less likely that the bond will be called. Therefore, callable bonds have a higher yield than non-callable bonds.
David Grabiner

FillorKill
Posts: 1007
Joined: Sat Aug 06, 2011 7:01 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by FillorKill » Tue Sep 27, 2016 3:52 am

grabiner wrote:
FillorKill wrote:Admittedly I'm a risk-on-the-equity-side investor. In theory GNMA is a lose/lose proposition for the mortgagor and a win/win for the mortgagee.


But the rates on GNMAs allow for that, as investors are aware of the risk. GNMAs are backed by the US Government, but they yield significantly more than Treasury bonds of the same duration; investors demand the higher yields because they will gain less if rates fall and lose more if rates rise. There is no free lunch, and in this case, the free lunch would apply if you could buy Treasuries and sell equal-yielding GNMAs short.


Sure; they're a spread product and the option-adjusted spread may or may not adequately compensate that risk in future periods. If that spread didn't exist the demand for the product would be roughly zero. Not many are interested in return-free risk. :D

Valuethinker
Posts: 33384
Joined: Fri May 11, 2007 11:07 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by Valuethinker » Tue Sep 27, 2016 4:13 am

grabiner wrote:
FillorKill wrote:Admittedly I'm a risk-on-the-equity-side investor. In theory GNMA is a lose/lose proposition for the mortgagor and a win/win for the mortgagee.


But the rates on GNMAs allow for that, as investors are aware of the risk. GNMAs are backed by the US Government, but they yield significantly more than Treasury bonds of the same duration; investors demand the higher yields because they will gain less if rates fall and lose more if rates rise. There is no free lunch, and in this case, the free lunch would apply if you could buy Treasuries and sell equal-yielding GNMAs short.

The same phenomenon applies to callable bonds. The effective duration of a callable bond increases if rates rise (and thus it loses more than a non-callable bond), because it becomes less likely that the bond will be called. Therefore, callable bonds have a higher yield than non-callable bonds.


Remember that risk premium is not constant over time.

So there is probably a tendency for corporates to issue callable bonds when implied volatility is low (small OAS over an uncallable bond, same maturity and credit rating).

As an investor you live in the probability = 1 world, ie the world where the outcome happened (or didn't happen), the world of ex post. That's what makes callable corporate bonds diabolical.

With GNMAs I don't imagine there is any systematic timing re spread. Housing market too big and complex for that. Same problem though-- the risk could show up just when you don't want it.

FillorKill
Posts: 1007
Joined: Sat Aug 06, 2011 7:01 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by FillorKill » Tue Sep 27, 2016 4:31 am

Beliavsky wrote:Only because of the discreteness of Portfolio Visualizer calculations. You can see that the Sharpe ratio when including VSGBX falls slightly from 1.09 to 1.07, which should not happen when adding a fund choice. I have emailed the creator of the site, and he admits it is a problem, which can be solved by generating more points on the efficient frontier. Portfolio Visualizer is a fine web site, but it is worth checking its results with other programs.

OK you made me look. There's more to it than just that. The projections the program is using are exceedingly optimistic. Additional points on that frontier would be helpful but that's just one aspect of the issue. As an example, from your analysis you have optimization results that show a weighting of 86.07 GNMA/11.71 HY/2.22 S&P 500 and the expected Sharpe/Sortino of 1.09/1.84

If you plug-in that weighting to the historic period you see Sharpe/Sortino of .98/1.56. The idea that the future risk-adjusted returns of that portfolio will exceed those achieved of during a secular decline in rates (Jan 88 - Aug 16) is asking a little too much IMHO.

Valuethinker
Posts: 33384
Joined: Fri May 11, 2007 11:07 am

Re: High yield bonds and GNMA's in a portfolio, since 1988

Post by Valuethinker » Tue Sep 27, 2016 5:14 am

FillorKill wrote:
Beliavsky wrote:Only because of the discreteness of Portfolio Visualizer calculations. You can see that the Sharpe ratio when including VSGBX falls slightly from 1.09 to 1.07, which should not happen when adding a fund choice. I have emailed the creator of the site, and he admits it is a problem, which can be solved by generating more points on the efficient frontier. Portfolio Visualizer is a fine web site, but it is worth checking its results with other programs.

OK you made me look. There's more to it than just that. The projections the program is using are exceedingly optimistic. Additional points on that frontier would be helpful but that's just one aspect of the issue. As an example, from your analysis you have optimization results that show a weighting of 86.07 GNMA/11.71 HY/2.22 S&P 500 and the expected Sharpe/Sortino of 1.09/1.84

If you plug-in that weighting to the historic period you see Sharpe/Sortino of .98/1.56. The idea that the future risk-adjusted returns of that portfolio will exceed those achieved of during a secular decline in rates (Jan 88 - Aug 16) is asking a little too much IMHO.


Underlining the problem of using historic data for projection-- the Markowitz optimizers just tilt you to the previous best performing classes.

Just because a risk did not show up during a sample period, doesn't mean it is not a risk.

It's almost a given that 20% S&P500 improves the risk-return tradeoff of any portfolio -- I think that's as good and as stable a result as we can get from past data. And similarly that US Treasury bonds are very diversifying (to a portfolio with a lot of non-bond assets).

Post Reply