The following model can be used to study the salary of graduates from MBA programs: log( salary) = Bo + B1 testsc + 32 WAM + Barank + u (3.1) where salary is the annual salary of an MBA graduate five years after graduation, testsc is his/her test score on a professional exam, WAM is his/her weighted average (course) mark during the MBA degree and rank is world-wide rank of the MBA program they attended (where rank = 1 is the best). Based on a sample of 930 MBA graduates the following estimates were obtained: log(salary) = 11.000 + 0.050testsc + 0.048 WAM - 0.0009rank (0.143) (0.017) (0.023) (0.0001) n 930, R2 = 0.0927, SSR = 0.45511 (i) I want to test whether the two characteristics of the graduates - testsc and WAM - are jointly significant in explaining log(salary). When I re-estimate the model with the restrictions imposed I find R2 = 0.0645 and SSR = 0.46926. Carry out the appropriate F-test using a 1% significance level. (ii) I am concerned that the model above may be misspecified due to neglected nonlinearities. Outline the steps required to carry out the RESET test, and clearly state the null and alternative hypotheses of the test. (iii) A possible omitted variable in (3.1) is years of work experience. If we expect that their WAM during the MBA and their experience are positively correlated and assuming the usual positive relationship between annual salary and experience holds, what is the expected bias in 32 the coefficient on WAM? (iv) If I selected the sample for the estimation of this model based on salary, for example, choosing only those graduates with annual salaries in excess of $120,000, what is the consequence for my estimates? (iv) I have a proxy variable for the years of work experience for these MBA grads. It is their (age-6 - 20), since this should be the number of years they were not in some kind of schooling or tertiary education. Call this expec_exper. I think I may be able to use expec exper as a proxy variable. What are the requirements for this variable expec_exper to be a valid proxy