23.07.2013 Views

samlet årgang - Økonomisk Institut - Københavns Universitet

samlet årgang - Økonomisk Institut - Københavns Universitet

samlet årgang - Økonomisk Institut - Københavns Universitet

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

366<br />

NATIONALØKONOMISK TIDSSKRIFT 2005. NR. 3<br />

is the students’ age when admitted. The ’error’ is a stochastic error term assumed to<br />

follow a normal distribution with zero mean and a constant variance.<br />

The interpretation of the model is as follows:<br />

– The expected grade of a student who has been admitted via quota-1 is:<br />

0 + 5 *AGEADMIT,<br />

i.e. the intercept of the model plus the coefficient of »ageadmit« times the age of the<br />

student at the time of the admission. If this age is 20.3 years, we have: 0 + 5 *20.3.<br />

– Similarly, the expected grade of a non-Scandinavian quota-2 student is:<br />

0 + 1 + 2 *GPA + 3 *MOTAVR + 5 *AGEADMIT<br />

For example, if the student has a GPA = 8.46 (the average), has MOTAVR = 12.80<br />

(the average) and is 23.78 years old when admitted, we have: 0 + 1 + 2 *8.46<br />

+ 3 *12.80 + 5 *23.78.<br />

– The expected grade of a Scandinavian quota-2 student is:<br />

0 + 1 + 2 *GPA + 3 *MOTAVR + 4 + 5 *AGEADMIT<br />

For example, if the student has GPA = 8.87 (the average), has MOTAVR = 12.80 and<br />

is 23.7 years old when admitted, we have: 0 + 1 + 2 *8.87 + 3 *12.8 + 4<br />

+ 5 *23.7<br />

4. The results<br />

The results are presented in the following two tables; table 1 for the first year courses<br />

and table 2 for the second year courses. 17 Note that the tables present a parsimonious<br />

model for each course. The »-« indicate the variables that are omitted compared to equation<br />

(1) due to acceptance of the null hypothesis that the coefficients of these variables<br />

were simultaneously zero (F-test results are reported in the final rows of the tables). 18<br />

Two misspecification tests are performed: the White’s general test for heteroscedasticity<br />

and the Jarque-Bera (JB) test for normality of the error term. In both cases<br />

a p-value smaller than 0.05 would indicate that the model was mis-specified. As can<br />

be seen, this is not the case.<br />

Our model has some explanatory value for all but two courses, viz. Intercultural<br />

Communication (123) and International Business Law (241). For these two courses<br />

none of the variables (that we examined) have an effect on grades. In these two courses,<br />

the expected average grade is 9.17 and 8.66, respectively. For all other courses the<br />

17. Instead of running regressions for each course (which reduces the number of observations available), we<br />

could have run one pooled regression that includes a dummy for each course. This would have increased the<br />

number of observations and presumably reduced the standard deviations and improved the fit. We decided,<br />

however, not to follow this procedure as that will remove the valuable information that one gets from running<br />

individual regressions – as will be discussed below, some variables work for some courses but not for<br />

others. Running one regression would blur that picture.<br />

18. For completeness we report the estimation results of the full models based on equation (1) in appendix 2.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!