- Question 2 18 Marks We Want To Estimate The Effect Of Non Labor Income Such As Carnings From Dividends Or Capital Inv 1 (61.6 KiB) Viewed 73 times
QUESTION 2 (18 marks) We want to estimate the effect of non-labor income (such as carnings from dividends or capital investment) on the number of worked hours. a) Suppose that you have a dataset with a random sample of 1 million individuals in Sweden. You run an OLS regression of worked hours y, on non-labor income xi. Explain why your OLS regression will not estimate the true effect of Xi on yi. In your data, you observe that 5,000 individuals have won a lottery prize, which implies that they experience an increase in non-labor income. b) You estimate a regression of worked hours y, on a dummy variable z that indicates if the individual won the lottery. Explain why this strategy will not estimate the true effect of winning the lottery on y (Hint: Not all the individuals in the sample bought tickets) c) Suppose that you run the same regression than in part b) but only for the subsample of individuals that bought tickets. Explain why this strategy is better if we want to identify the true effect of zon Yi. Explain also why we must be careful with the "external validity" of this strategy. d) We suspect that winning the lottery has a stronger effect on worked hours for females than for males. Write the equation of the regression that would allow us to capture this idea.