Econometrics Exercise session for midterm preparation solutions Problem 1 Suppose that X is the number of free throws made by a basketball player out of two attempts and assume that the individual probabilities for each outcome of X are the following: pr(x=0)=0.2; pr(x=1)=0.44 and pr(x=2)=0.36 i) Define the random variable. X 0 1 2 P(X) 0.2 0.44 0.36 ii) Draw the probability distribution associated to the above random variable. iii) Calculate the expected value of the above random variable. E(x)=0*0.2+1*0.44+2*0.36=1.16 iv) Calculate the probability that the player makes at least one free throw P(X≥1)=1-P(X<1)=1-P(X=0)=1-0.2=0.8 Problem 2 We have a dataset containing data about births to women in the United States. Two variables of interest are the dependent variable, infant birth weight in ounces (bw), and an explanatory variable, average number of cigarettes the mother smoked per day during pregnancy (cigs). The following simple regression was estimated using data on 1,388 births: i) Think about possible factors contained in the error term 𝑢[𝑖] . Prenatal doctor visits, stressful environment of the mother, gender of a child, mother education, family income ii) Interpret the above regression results. If number of cigarettes smoked per day increases by one unit, then the birth weight of the child reduces by 0.5 ounces. Ceteris paribus, average birth weight of a child is 119.77 ounces iii) What is the predicted birth weight when cigs =10? What about when cigs = 20 (one pack per day)? Comment on the difference. 119.77-0.514*10=114.63 119.77-0.514*20=109.49 Problem 3 We have information about mortality rates (MORT=total mortality rate per 100,000 population) in a specific year for 51 States of the United States combined with information about potential determinants: INCC (per capita income by State in Dollars), POV (proportion of families living below the poverty line), EDU (proportion of population completing 4 years of high school), TOBC (per capita consumption of cigarettes by State) and AGED (proportion of population over the age of 65). Estimation results are presented in the following table: i) Interpret the slope coefficient in Model 1 and validate it at 1% significance level. The slope coefficient in Model 1 implies that increasing the proportion of population over the age 65 by 1 percentage points will increase the mortality rate per 100000 of population by 5 546. T statistics of the coefficient is t= , which is > 2.7, therefore, it is significant at the 1% significance level ii) Validate the joint significance of Model 2 in comparison to model 1 at 1% significance level? Joint significance of model 2 in comparison to model 1 implies that we test joint significance of the coefficients Incc and Edu, according to F test we have comparing this with the critical value 5.1 we reject the hypothesis that the coefficients Incc and Edu are jointly insignificant iii) Comment on the effect of INCC on MORT in the second model. Why do you think is a positive and significant effect? When per capita income by state is higher, the mortality rate is also higher. This at one glance does not make any sense because rich people should be able to afford better health care and therefore, extend longevity of their lives. However, we can make an argument that generally, older people are more likely to have higher income, therefore, those states with high income probably also have proportion of old people higher and hence, the mortality rate is higher. iv) In Model 3 we add two new explanatory variables: POV and TOBC. Test whether this inclusion helps to improve the quality of the model at 1% significance level. Is model 3 the best in terms of goodness-of-fit? This question indirectly asks to compare the model 3 to model 2, therefore, we need to test joint significance of the variables POV and TOBC Again we calculate an F testF= , which is still larger than the critical value 5.1, therefore, these two variables are jointly significant at the 1% significance level. The model is the best in terms of the goodness of fit, because R^adj is the highest v) Are the effects of these two new variables the expected ones? Are they individually significant at 1% significance level? These two new variables have “positive” impact on the mortality rate, which makes sense, more smokers- higher mortality, more poor people – higher mortality. T test for POV is 2.8 and for TOBC, 2.9, both are statistically significant at 1% significance level when comparing to critical value 2.7 vi) What about the individual significance of EDU in model 3 if compared with model 2? Why? Edu in model 2 is significant while it is not in model 3. The fact that POV and TOBC were omitted in the model 2 was causing a bias in the estimation of the coefficient Edu. The reason is that EDU is negatively correlated with both POV and TOBC – more educated people are less likely to smoke and less likely to be poor. Meanwhile, POV and TOBC are positively correlated with the explained variable MORT, meaning that the direction of bias is negative. Since coefficient on Edu is negative, this bias was making it more negative in absolute terms and this way it was making it significant. Problem 4 consider a simple model to compare the returns to education at junior colleges and four-year colleges; for simplicity, we refer to the latter as “universities.” The population includes working people with a high school degree, and the model is: (1) where jc is number of years attending a two-year college, univ is number of years at a four-year college. exper is months in the workforce. Note that any combination of junior college and four-year college is allowed, including jc =0 and univ = 0. Use the data twoyear.dta a) Test the hypothesis that . The hypothesis of interest is whether one year at a junior college is worth one year at a university. To test this hypothesis we instead want to test and plug it in the original regression: (2) Now run: genr unjc=univ+jc ols lwage const jc unjc exper so the return to a year at a junior college is about one percentage point less than a year at a university. Test statistic on jc t=0.0102/.0069 =-1.48. We need to compare this with one sided alternative critical value. At 10% one-sided significance level, critical value is -1.282. Therefore, there is some but not strong evidence against the null hypothesis. Check also command: ols lwage const jc univ exper. Make your own observation! (ii) The variable phsrank is the person’s high school percentile. (A higher number is better. For example, 90 means you are ranked better than 90 percent of your graduating class.) Find the smallest, largest, and average phsrank in the sample. summary phsrank (ii) Add phsrank to regression (2) and report the OLS estimates in the usual form. Is phsrank statistically significant? How much is 10 percentage points of high school rank worth in terms of wage? ols lwage const jc unjc exper phsrank phsrank has a t statistic equal to only 1.25; it is not statistically significant. If we increase phsrank by 10, log(wage) is predicted to increase by (.0003)10 = .003. This implies a .3% increase in wage, which seems a modest increase given a 10 percentage point increase in phsrank. (iii) Does adding phsrank to regression (2) substantively change the conclusions on the returns to two- and four-year colleges? Explain. Adding phsrank makes the t statistic on jc even smaller in absolute value, about 1.33, but the coefficient magnitude is similar to (2). Therefore, the base point remains unchanged: the return to a junior college is estimated to be somewhat smaller, but the difference is barely significant with one-sided test. Problem 5 A soda vendor at Louisiana State University football games observes that more sodas are sold the warmer the temperature at game time is. Based on 32 home games covering five years, the vendor estimates the relationship between soda sales and temperature to be where y is the number of sodas she sells and x is temperature in degrees Fahrenheit, (a) Interpret the estimated slope and intercept. Do the estimates make sense? Why, or why not? The intercept estimate b[1] =-240 is an estimate of the number of sodas sold when the temperature is 0 degrees Fahrenheit. A common problem when interpreting the estimated intercept is that we often do not have any data points near x =0 . If we have no observations in the region where temperature is 0, then the estimated relationship may not be a good approximation to reality in that region. Clearly, it is impossible to sell -240 sodas and so this estimate should not be accepted as a sensible one. The slope estimate b[2]=8 is an estimate of the increase in sodas sold when temperature increases by 1 Fahrenheit degree. This estimate does make sense. One would expect the number of sodas sold to increase as temperature increases. (b) On a day when the temperature at game time is forecast to be 80^0F, predict how many sodas the vendor will sell. If temperature is 80^0F, the predicted number of sodas sold is (c) Below what temperature are the predicted sales zero? If no sodas are sold, y =0, and (d) Sketch a graph of the estimated regression line.