Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Re: "Structural" break tests for non-time series data??


From   "Martin Weiss" <[email protected]>
To   <[email protected]>
Subject   st: Re: "Structural" break tests for non-time series data??
Date   Wed, 24 Jun 2009 23:27:16 +0200

<>

You might want to look into -suest- and http://www.stata.com/support/faqs/stat/chow.html



HTH
Martin
_______________________
----- Original Message ----- From: "kokootchke" <[email protected]>
To: "statalist" <[email protected]>
Sent: Wednesday, June 24, 2009 11:11 PM
Subject: st: "Structural" break tests for non-time series data??


Hello!!

I am running a regression of bond yield spreads* (non-finance people, see * below) on a bunch of variables. Let's group these variables in two sets, set A and set B. Other research has considered that these bond spreads depend on set A. My hypothesis is that the importance of set A in the determination of the spread should be lower once we take into account set B (omitted variables). Not only that, but that the variables in set B have become more and more important over time relative to those in set A.

If I run the regression:

spread = f(A,B,year dummies)

I observe that years 1998-99 are very important and I think that has to do with the Russian crisis which affected bond markets big time.

Now, when I run the same regression (without the year dummies) using two subsamples, one pre-1998 and one post-1998, I do confirm my hypothesis above, namely, that the magnitudes of the effects of A on spreads are lower in the post-1998 regression, while those of B become way higher.

I would like to know if there is a structural-break-type test that I can use to confirm that 1998 does indeed mark such a break in my dataset.

ONE IMPORTANT THING, THOUGH: my bond spreads data are NOT a time series -- that is, I don't observe a spread for bond i every day or every month or every year. I only observe it ONCE at the time the bond was launched, and then that's it (primary market spread).

So, basically, each row in my dataset corresponds to a given bond, and I have information on the spread (at launch, or at the time of issue), what company or what sovereign government issued it, what country they belong to, the total amount raised in the bond sale, the term or maturity of the bond... and then all of these other variables in sets A and B.

I hope that was clear enough!

Thank you very much in advance!

Best,
Adrian

* For the non-finance person, the yield spread of a (risky) bond is just the difference between its yield or return and the yield or return of another (safe) bond which is used as a benchmark. In general, greater returns are associated with greater risks.

_________________________________________________________________
Hotmail® has ever-growing storage! Don’t worry about storage limits.
http://windowslive.com/Tutorial/Hotmail/Storage?ocid=TXT_TAGLM_WL_HM_Tutorial_Storage_062009
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index