Prof. Bryan Caplan

http://www.gmu.edu/departments/economics/bcaplan

Spring, 1999

HW#5 (due Friday, May 7 by noon)

Johnston/DiNardo:

Math problems: 13.4

Computing problems: 13.8 (but perform OLS on the censored data instead of a Tobit, since you don't have the software to do a Tobit).

1.

a. Use your unemployment/inflation data to estimate the following VAR:

u_t=b11+b12u_t-1+b13i_t-1+b14u_t-2+b15i_t-2

i_t=b21+b22u_t-1+b23i_t-1+b24u_t-2+b25i_t-2

b. Estimate the impact of a 1 SD innovation to one of the equations on the entire system using an impulse-response function. (You calculate the 1 SD innovation by taking the square root of e'e/(n-k) for that equation). Calculate the impulse-response function from t=0 out to t=10.

2. Go to my webpage to download the Excel file with data on 15 countries from 1881-1992. Pick out a subset of this data that your software can handle and import it into Eviews. Include at least 3 countries and 10 years of data. Manually set up the lags and country and year dummies (i.e., treat each lag as a unique variable such as nlag1 rather than just n(-1)). Then:

a. Estimate one pooled linear probability regression without country or year dummies of the form Y_t=b1+b2Y_t-1+b3X_t, where Y is a dummy variable and X is comprised of variables of your choice that might have an interesting connection with Y.

b. Re-estimate the above regression with country dummies but no year dummies.

c. Re-estimate the above regression with both country and year dummies.