Comments on: What the Treasury Yield Curve is Telling Us About Corporate Bond Yields http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/ Helping Institutions and Ordinary People Invest Better by Focusing on Risk Control Thu, 18 Mar 2010 02:23:05 -0400 http://wordpress.org/?v=2.8.4 hourly 1 By: David Merkel http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/comment-page-1/#comment-16352 David Merkel Tue, 01 Jan 2008 16:20:08 +0000 http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/#comment-16352 Bill, I can put all of that together, but I am adding it to my article list. I'll give you a quick answer on yield curve slope, though. I use the definition that is pretty standard among bond investors -- 10-year yield less 2-year yield. Difference, not ratio. Bond investors call that "the belly of the curve." (Sometimes that phrase is more closely applied to the 3-7 year portion of the curve.) It represents the bulk of the investable bond universe. I look at the whole curve, though. The front end tells us about the money markets, 0-2 years. The long end tells us about long term liability defeasance, such as with defined benefit pensions, or structured settlements (10+ years). So, my simplified yield curve has three joints to it: short (3-month), 2-year, 10-year, and the long end (usually 30-years). As for the correlation matrices, detailed data on corporate yields really did not exist until the early '90s, partially because corporates traded on a dollar basis, and spread data were not retained. There are a few yield series that are useful for relative purposes, like the Moody's series, but they are blends of many maturities. On Treasuries, we can go back a long way ('20s), but I think the period prior to floating exchange rates is less relevant to the way the yield curve currently behaves. Monetary policy is different, and so are yield curve dynamics. I used to do yield curve modeling for life insurance companies. My best model was a lognormal multivariate mean-reverting model, which behaved realistically. (The Society of Actuaries has a better one for the Treasury curve, but mine did credit spreads as well.) I don't have access to that model now (client property), but I could develop it again. It would take some time, though, and I am getting close to starting a new job, so time is limited. When I write the article, I can try to divulge it all. That's probably not one article, but a series. Bill, I can put all of that together, but I am adding it to my article list. I’ll give you a quick answer on yield curve slope, though. I use the definition that is pretty standard among bond investors — 10-year yield less 2-year yield. Difference, not ratio. Bond investors call that “the belly of the curve.” (Sometimes that phrase is more closely applied to the 3-7 year portion of the curve.) It represents the bulk of the investable bond universe.

I look at the whole curve, though. The front end tells us about the money markets, 0-2 years. The long end tells us about long term liability defeasance, such as with defined benefit pensions, or structured settlements (10+ years).

So, my simplified yield curve has three joints to it: short (3-month), 2-year, 10-year, and the long end (usually 30-years).

As for the correlation matrices, detailed data on corporate yields really did not exist until the early ’90s, partially because corporates traded on a dollar basis, and spread data were not retained. There are a few yield series that are useful for relative purposes, like the Moody’s series, but they are blends of many maturities.

On Treasuries, we can go back a long way (’20s), but I think the period prior to floating exchange rates is less relevant to the way the yield curve currently behaves. Monetary policy is different, and so are yield curve dynamics.

I used to do yield curve modeling for life insurance companies. My best model was a lognormal multivariate mean-reverting model, which behaved realistically. (The Society of Actuaries has a better one for the Treasury curve, but mine did credit spreads as well.) I don’t have access to that model now (client property), but I could develop it again. It would take some time, though, and I am getting close to starting a new job, so time is limited. When I write the article, I can try to divulge it all. That’s probably not one article, but a series.

]]> By: Bill aka NO DooDahs! http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/comment-page-1/#comment-16351 Bill aka NO DooDahs! Tue, 01 Jan 2008 13:16:51 +0000 http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/#comment-16351 Sounds good. What about your working definitions? They would be text data. Out of all the different timeframes available, how would you define the measurement of the T yield curve? Difference of two rates (which two?), ratio of two rates (which two?), or some mathematical function of more than two rates? Do you define the spread based corporate rates? If so, how? Difference of two rates (which two?), ratio of two rates (which two?), or some mathematical function of more than two rates? If they really are inversely related, a correlation coefficient and the definitions behind it (curve definitions, date ranges of data used, definition of data points [as in monthly evaluation points from 1975 through 2007]) all would be text data. Sounds good. What about your working definitions? They would be text data.

Out of all the different timeframes available, how would you define the measurement of the T yield curve? Difference of two rates (which two?), ratio of two rates (which two?), or some mathematical function of more than two rates?

Do you define the spread based corporate rates? If so, how? Difference of two rates (which two?), ratio of two rates (which two?), or some mathematical function of more than two rates?

If they really are inversely related, a correlation coefficient and the definitions behind it (curve definitions, date ranges of data used, definition of data points [as in monthly evaluation points from 1975 through 2007]) all would be text data.

]]> By: David Merkel http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/comment-page-1/#comment-16349 David Merkel Tue, 01 Jan 2008 03:12:14 +0000 http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/#comment-16349 Bill, I'm having some difficulty posting non-text data, but when I get it resolved, I'll post some graphs on this topic. Bill, I’m having some difficulty posting non-text data, but when I get it resolved, I’ll post some graphs on this topic.

]]> By: Bill aka NO DooDahs! http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/comment-page-1/#comment-16345 Bill aka NO DooDahs! Mon, 31 Dec 2007 20:57:17 +0000 http://alephblog.com/2007/12/31/what-the-treasury-yield-curve-is-telling-us-about-corporate-bond-yields/#comment-16345 What is the "spread curve of high yield bonds?" If I had access to the data directly, how would I calculate it? I assume it's a quality spread of corporates, but it's always nice to have a definition nailed down, so we're all talking about the same thing. While you're at it, a good definition of the T-yield curve would be nice, since I've seen it as the 10/2 difference, the 10/2 ratio, the 10/FFR difference, the 10/FFR ratio, and could imagine various other ways to calculate it. Definitional problems lead to other problems. It's easy to buy that the quality spread between junk corporate and Ts isn't the most relevant metric. It's NOT so easy to buy that the quality spread between, say for example, As and junk, tends to invert when the TIME spread on Ts, say for example, difference between 10/2, gets steep. I'd like to see data on that, if you have it. What is the “spread curve of high yield bonds?” If I had access to the data directly, how would I calculate it? I assume it’s a quality spread of corporates, but it’s always nice to have a definition nailed down, so we’re all talking about the same thing.

While you’re at it, a good definition of the T-yield curve would be nice, since I’ve seen it as the 10/2 difference, the 10/2 ratio, the 10/FFR difference, the 10/FFR ratio, and could imagine various other ways to calculate it. Definitional problems lead to other problems.

It’s easy to buy that the quality spread between junk corporate and Ts isn’t the most relevant metric.

It’s NOT so easy to buy that the quality spread between, say for example, As and junk, tends to invert when the TIME spread on Ts, say for example, difference between 10/2, gets steep.

I’d like to see data on that, if you have it.

]]>