Question on How to Perform the Chow Break test

[Stefan Peterfi]

Hello everyone,

I am writing my Bachelor Thesis on seasonal anomalies on the Euronext Amsterdam (day-of-the-week effect on AEX, AMX, and ASCX indexes). In order to test whether the anomalies have dissipated in time, I need to split the time-series data in sub-samples, but this has to be done according to some benchmark, or even better, a statistical test. Therefore, a Chow Break test for an unknown break date would be suitable in my case. However, I have encountered some difficulties in running the test due to my regression equation. See picture below!

Now, I need to identify whether a structural break occurs at an unknown date in LogReturns, which is the dependent variable. For a Chow Break test however (Stata command: estat sbsingle, breakvars (*variable of interest*)), the variable of interest has to be an independent variable. By construction, I cannot use the Chow Break test to verify for a structural break in returns. Is there a simple way to go around this, or should I just resume to visually inspecting the returns and identifying some break dates? Any kind of help would be highly appreciated! Thank you in advance!

Regards,
Stefan

[Phil Bromiley]

You might be able to trick it by running the regression with just the intercept (reg LogReturns ) and then look for the break using the constant as the x variable. If it doesn't work with this form of ols, generate a column of 1's and run reg LogReturns ones, noconstant.