The RATS Software Forum

I am trying to implement the Gibbons, Ross and Shanken (1989) test "A Test of Efficiency of a Given Portfolio", Econometrica vol.57, pp.1121-1152 (the paper can be found here: http://home.business.utah.edu/finmll/fi ... en1989.pdf). This test is widely used for linear asset pricing models. Specifically, I would like to test that the alphas (intercepts) in a linear single or multiple factor asset pricing model (CAPM, Fama-French three factor) are jointly zero. The GRS test is essentially a joint exclusion test for the alphas. Its distrbution is chi-squared (asymptotically) or is F-distributed in small smaples.

My question is the following: If I estimate a Seemingly Unrelated Regressions system with the excess returns on 25 portfolios as dependent variables, is there any way to test a joint exclusion restrction for all the intercepts in the SUR system in one go (using, for example, TEST or RESTRICT)? That would essentially amount to the asymptotic version of the test. Or, would I have to follow consumer.RPF to do that (use NLSYSTEM)? and if I go with the second approach (NLSYSTEM), how do I obtain the test statistic?

Thanks.

The attached program will compute both a standard and heteroscedasticity-robust version of the GRS test. It's a "SUR" regression, but in a form where it can be done by equation by equation OLS. The most convenient way to handle that is by using the SYSTEM instructions and ESTIMATE.

I am expecting to use Gibbons, Ross and Shanken (1989) test in order to test the efficiency of the Fama and French Five Factor (FF5) model. Although I have downloaded the previous attachment, I could not get the expected output as it shows an error at the end of the programme, So will you be able to help me providing an appropriate code to run the GRS test for the five factor model?

Thank you.

You would have to attach a copy of the program and data that you tried to use.