Consider an empirical example in EXCEL. Download the Wooldridge data description ceosal1.des and the corresponding EXCEL file ceosal1.xls. It contains information on 209 Chief Executive Officers (CEOs) for the year 1990. Use the EXCEL Data Analysis Toolpack for the analysis. Keep the variables Y = lsalary and X = lsales. We will investigate the model

lsalaryj = 0 + 1lsalesj + j with j = 1, ..., 209. (2)

1. (5 pt) Determine summary statistics for Y and X. Are there any outliers?

2. (5 pt) Compute 95% confidence intervals for X and Y .

3. (5 pt) Compute the regression in (2). Provide the regression line and obtain esti- mates for 0 and 1. Interpret the estimates of the two parameters seperately.

4. (5 pt) Comment on the p-values for b0 and b1. Interpret the results.

5. (5 pt) Comment on the fit of the model. What measure(s) do you use?

6. (5 pt) Compute the Students t-values needed to test two different hypothesis tests: (i)H0 :1 =0vs. HA :1 =0and(ii)H0 :1 =0.5vs. HA :1 =0.5.

7. (5 pt) Comment on the exogeneity of the X variable.

Let us now extend the model in (2) by adding an additional regressor Return on Equity (ROE) (roe in the EXCEL file). We believe ROE to be an important measure of firm performance which might lead to a higher compensation for the CEO.

8. (5 pt) Estimate the extended model. Do estimates have expected sign and are they significant?

9. (5 pt) Compare model (2) and the extended model that also includes ROE. Can we compare the models using R2, R2 and an F-test?

You remember from the lectures that it might be a good idea to also add quadratic terms to the regression equation.

j. (5 pt) What could be the reason to add a new variable, say ROE2, to the regression? Is there empirical evidence to indeed include such a term?