turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- Structural Break Test (CUSUM or Chow)

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

06-30-2011 02:09 PM

Hello,

First let me tell you that I'm not so experienced in SAS.

I was given a task to find a structural break in labor productivity using the following regression model:

Labor_Productivity = Unemployment Oil_Prices Real_Interest_Rates.

When I use growth rates, which I was advised to do, and run the regression, the R-square shows only 4% and t-statistics for the variables are not so good.

However, when I use logs of the variables, the R-square is around 82% and t-statistics are much better, but Durbin-Watson statistic shows a positive serial correlation.

Why is it such a difference? I think I need to use growth rates in my regression, but due to low statistics, the Chow test gives wrong output (I think) or maybe I'm doing it wrong.

Because I don't know when the break (breaks) occurred, I tried CUSUM test. Here is the code I used:

proc autoreg data=AJ.WD1 ;

model prod_change = unem_change oilprice_change realintrate_change;

*model ln_prod = ln_unem ln_oilprice realintrate;

output out=cusum cusum=cusum cusumub=ub cusumlb=lb;

run;

proc print data=cusum;

var lb cusum ub;

run;

However SAS gives CUSUM ERROR and dots for lb cusum ub. What am I doing wrong?

Is there any other way to find unknown structural breaks in panel data?

Please see the attached file with the data.

Please, any advice or suggestion, or help would be greatly appreciated.