Finance event study on abnormal returns for one security - non parametric test needed

yvha

New Member
#1
Hi everyone,

We are conducting a finance event study on abnormal returns and unexpected trading volume.

Our data consists of abnormal returns/volume (not normally distributed, leptokurtic, log transformation not enough) for an industry index security for 250 consecutive trading days for estimation period, and the following 250 consecutive trading days out-of-sample, in which we have three sets of 5 days of event periods.
We would like to test each of these days in each event period separately as AR, and as a combined period as CAR.

Previous similar studies apply Corrado rank tests, but we can only find examples with a portfolio of securities. Simply using 1 as N (securities in cross-section of portfolio) results in never significant, due to a high standard deviation compared to the standardized rank statistic.

We have considered a simple sign-test, but again with only one security in the portfolio, this is not sophisticated enough.
We currently use Excel/STATA but are open to using any kind of free software (or SAS Studio) that can help us.

Does anyone have any suggestions how to properly perform the Corrado test in a single security portfolio, or perhaps a completely different test? All help would be very much appreciated!!

Thank you,
a very desperate thesis-writer