Econometrics PS # 2
Name:
Read carefully the questions and think before writing. When you do write, please do so in a clear way. Always, always, explain the reasoning behind your answer.
1. You want to investigate determinants of car fatalities. You have trac fatalities data on the lower 48 US states (i.e. excluding Alaska and Hawaii), for 1988. In particular, you have the following variables:
fatal vehicle fatalities per ten thousand people unemp unemployment rate income per capita personal income in 1987 (in thousands of dollars) beertax tax on case of beer in 1987 dollar miles average miles per driver jail =1 if state has mandatory jail for drunk driving service = 1 if state has mandatory community service youngdrivers percent of drivers aged 15-24.
f atal =0.5708 + 0.0490 unemp 0.0573 income + 0.1442 beertax + 0.0002 miles
(0.9333) (0.0363) (0.0248) (0.1317) (0.0001)
(b) What is the eect of an increase in income of $20, 000? Is this eect statistically signicant at 5%?
(c) Everything else being equal, how many more fatalities per 500 people would you expect if on average 10000 more miles were driven?
(d) There at least 2 errors in the following statement made by a politician who saw your regression analysis: Increasing beertax by .43 will increase fatalities by at least 0.144. This is strong evidence that there is no relationship between taxing beer and fatalities rate, thus, for the welfare of individuals, states should abolish taxes on alcoholic beverages. Can you spot the errors?
2. Earnings functions are one of the most investigated relationships in economics. These typically relate the logarithm of earnings to a series of explanatory variables such as education, work experience, gender, race, etc. Consider using data on a sample of Italian workers aged between 15 and 64. The sample is from the Labor Force Survey from ISTAT which is the most comprehensive survey on labor related issues. The sample we use here is a small subsample (3,000 observations) of the one provided by ISTAT We estimate rst a regression model linking lhourlywage, the logarithm of hourlywage, with age dummies and education dummies and a dummy for gender. Here a brief summary of the variables: hourlywage male educ | | | hourly wage =1 if individual is a male years of education
The summary is reported below. ## ## ## ## ## ## ## ## ## Min. Min. 1st Qu. 1.00 male :0.000 Min. 1st Qu.:0.000 Median :1.000 Mean Max. :0.541 :1.000 3rd Qu.:1.000 3.00 Median 5.00 educ : 0.0 1st Qu.: 8.0 Median :13.0 Mean Max. :11.5 :20.0 3rd Qu.:13.0 Mean 3rd Qu. 4.87 5.00 Min. : Max. 12.00 1.56 6.88 8.33 9.49
Consider the following regression: ## ## Coefficients: ## ## (Intercept) ## male ## educ ## --## Heteroskadasticity robust standard errors used ## ## Residual standard error: 6.26 on 2997 degrees of freedom ## Multiple R-squared: ## F-statistic: 0.0597,Adjusted R-squared: 0.0591 188 on 2 and Inf DF, p-value: <2e-16 Estimate Std. Error t value Pr(>|t|) 4.3612 0.4225 0.4254 0.3367 0.2248 0.0314 12.95 1.88 13.56 <2e-16 0.06 <2e-16
(b) What does this regression say about the returns on education in Italy? Has a big eect on wages?
(c) Based on the regression, can you say something denitive about female discrimination in Italy? (Explain)