CHAPTER 7. + ˆ δ. (1 nopc) + ˆ β1. =.157, so the new intercept is = The coefficient on nopc is.157.

Transcription

1 CHAPTER 7 SOLUTIONS TO PROBLEMS 7. (i) The coefficient on male is 87.75, so a man is estimated to sleep almost one and one-half hours more per week than a comparable woman. Further, t male = 87.75/ , which is close to the % critical value against a two-sided alternative (about 2.58). Thus, the evidence for a gender differential is fairly strong. (ii) The t statistic on totwrk is.63/.8 9.6, which is very statistically significant. The coefficient implies that one more hour of work (6 minutes) is associated with.63(6) 9.8 minutes less sleep. 2 (iii) To obtain R r, the R-squared from the restricted regression, we need to estimate the model without age and age 2. When age and age 2 are both in the model, age has no effect only if the parameters on both terms are zero. 7.3 (i) The t statistic on hsize 2 is over four in absolute value, so there is very strong evidence that it belongs in the equation. We obtain this by finding the turnaround point; this is the value of hsize that maximizes sat ˆ (other things fixed): 9.3/(2 2.9) 4.4. Because hsize is measured in hundreds, the optimal size of graduating class is about 44. (ii) This is given by the coefficient on female (since black = ): nonblack females have SAT scores about 45 points lower than nonblack males. The t statistic is about.5, so the difference is very statistically significant. (The very large sample size certainly contributes to the statistical significance.) (iii) Because female =, the coefficient on black implies that a black male has an estimated SAT score almost 7 points less than a comparable nonblack male. The t statistic is over 3 in absolute value, so we easily reject the hypothesis that there is no ceteris paribus difference. (iv) We plug in black =, female = for black females and black = and female = for nonblack females. The difference is therefore = 7.5. Because the estimate depends on two coefficients, we cannot construct a t statistic from the information given. The easiest approach is to define dummy variables for three of the four race/gender categories and choose nonblack females as the base group. We can then obtain the t statistic we want as the coefficient on the black female dummy variable. 7.5 (i) Following the hint, colgpa = ˆ β + ˆ δ ( nopc) + ˆ β hsgpa + ˆ β2 ACT = ( ˆ β + ˆ δ ) ˆ δ nopc + ˆ β hsgpa + ˆ β2 ACT. For the specific estimates in equation (7.6), ˆ β =.26 and ˆ δ =.57, so the new intercept is =.47. The coefficient on nopc is

2 (ii) Nothing happens to the R-squared. Using nopc in place of PC is simply a different way of including the same information on PC ownership. (iii) It makes no sense to include both dummy variables in the regression: we cannot hold nopc fixed while changing PC. We have only two groups based on PC ownership so, in addition to the overall intercept, we need only to include one dummy variable. If we try to include both along with an intercept we have perfect multicollinearity (the dummy variable trap). 7.7 (i) Write the population model underlying (7.29) as inlf = β + β nwifeinc + β2 educ + β3 exper + β 4 exper 2 + β5 age + β6 kidslt6 + β7 kidsage6 + u, plug in inlf = outlf, and rearrange: or outlf = β + β nwifeinc + β2 educ + β3 exper + β 4 exper 2 + β5 age + β6 kidslt6 + β7 kidsage6 + u, outlf = ( β ) β nwifeinc β2 educ β3 exper β4 exper 2 β5 age β6 kidslt6 β7 kidsage6 u, The new error term, u, has the same properties as u. From this we see that if we regress outlf on all of the independent variables in (7.29), the new intercept is.586 =.44 and each slope coefficient takes on the opposite sign from when inlf is the dependent variable. For example, the new coefficient on educ is.38 while the new coefficient on kidslt6 is.262. (ii) The standard errors will not change. In the case of the slopes, changing the signs of the estimators does not change their variances, and therefore the standard errors are unchanged (but the t statistics change sign). Also, Var( ˆ β ) = Var( ˆ β ), so the standard error of the intercept is the same as before. (iii) We know that changing the units of measurement of independent variables, or entering qualitative information using different sets of dummy variables, does not change the R-squared. But here we are changing the dependent variable. Nevertheless, the R-squareds from the regressions are still the same. To see this, part (i) suggests that the squared residuals will be identical in the two regressions. For each i the error in the equation for outlf i is just the negative of the error in the other equation for inlf i, and the same is true of the residuals. Therefore, the SSRs are the same. Further, in this case, the total sum of squares are the same. For outlf we have SST = n n 2 2 ( outlfi outlf ) = [( inlfi) ( inlf )] i= i= 2 2 = ( inlfi + inlf ) = ( inlfi inlf ), n i= i= n 35

3 which is the SST for inlf. Because R 2 = SSR/SST, the R-squared is the same in the two regressions. 7.9 (i) Plugging in u = and d = gives f ( z) = ( β + δ ) + ( β + δ ) z. * * * * * (ii) Setting f( z ) = f( z ) gives β + βz = ( β + δ) + ( β+ δ) z or = δ + δz. * * Therefore, provided δ, we have z = δ / δ. Clearly, z is positive if and only if δ / δ is negative, which means δ and δ must have opposite signs. (iii) Using part (ii) we have * totcoll =.357 /.3 =.9 years. (iv) The estimated years of college where women catch up to men is much too high to be practically relevant. While the estimated coefficient on female totcoll shows that the gap is reduced at higher levels of college, it is never closed not even close. In fact, at four years of college, the difference in predicted log wage is still (4) =.237, or about 2.% less for women. SOLUTIONS TO COMPUTER EXERCISES C7. (i) The estimated equation is colgpa = PC +.45 hsgpa +.77 ACT.38 mothcoll (.34) (.59) (.94) (.7) (.63) +.48 fathcoll (.63) n = 4, R 2 =.222. The estimated effect of PC is hardly changed from equation (7.6), and it is still very significant, with t pc (ii) The F test for joint significance of mothcoll and fathcoll, with 2 and 35 df, is about.24 with p-value.78; these variables are jointly very insignificant. It is not surprising the estimates on the other coefficients do not change much when mothcoll and fathcoll are added to the regression. (iii) When hsgpa 2 is added to the regression, its coefficient is about.337 and its t statistic is about.56. (The coefficient on hsgpa is about.83.) This is a borderline case. The quadratic in hsgpa has a U-shape, and it only turns up at about hsgpa * = 2.68, which is hard to interpret. The coefficient of main interest, on PC, falls to about.4 but is still significant. Adding hsgpa 2 is a simple robustness check of the main finding. 36

4 C7.3 (i) H : β 3 =. Using the data in MLB.RAW gives ˆ β 3.254, se( ˆ β 3 ).3. The t statistic is about.94, which gives a p-value against a two-sided alternative of just over.5. Therefore, we would reject H at just about the 5% significance level. Controlling for the performance and experience variables, the estimated salary differential between catchers and outfielders is huge, on the order of [exp(.254) ] 28.9% [using equation (7.)]. (ii) This is a joint null, H : β 9 =, β =,, β 3 =. The F statistic, with 5 and 339 df, is about.78, and its p-value is about.7. Thus, we cannot reject H at the % level. (iii) Parts (i) and (ii) are roughly consistent. The evidence against the joint null in part (ii) is weaker because we are testing, along with the marginally significant catcher, several other insignificant variables (especially thrdbase and shrtstop, which has absolute t statistics well below one). C7.5 The estimated equation is log( salary ) = log(sales) +.67 roe.226 rosneg (.29) (.34) (.4) (.9) n = 29, R 2 =.297, 2 R =.286. The coefficient on rosneg implies that if the CEO s firm had a negative return on its stock over the 988 to 99 period, the CEO salary was predicted to be about 22.6% lower, for given levels of sales and roe. The t statistic is about 2.7, which is significant at the 5% level against a twosided alternative. C7.7 (i) When educ = 2.5, the approximate proportionate difference in estimated wage between women and men is (2.5) =.297. When educ =, the difference is.227. So the differential at 2.5 years of education is about 7 percentage points greater. (ii) We can write the model underlying (7.8) as log(wage) = β + δ female + β educ + δ female educ + other factors = β + ( δ + 2.5δ ) female + β educ + δ female (educ 2.5) + other factors β + θ female + β educ + δ female (educ 2.5) + other factors, 37

5 where θ δ + 2.5δ is the gender differential at 2.5 years of education. When we run this regression we obtain about.294 as the coefficient on female (which differs from.297 due to rounding error). Its standard error is about.36. (iii) The t statistic on female from part (ii) is about 8.7, which is very significant. This is because we are estimating the gender differential at a reasonable number of years of education, 2.5, which is close to the average. In equation (7.8), the coefficient on female is the gender differential when educ =. There are no people of either gender with close to zero years of education, and so we cannot hope nor do we want to to estimate the gender differential at educ =. C7.9 (i) About.392, or 39.2%. (ii) The estimated equation is e 4k = inc.62 inc age.3 age 2.35 male (.8) (.6) (.5) (.39) (.5) (.2) n = 9,275, R 2 =.94. (iii) 4(k) eligibility clearly depends on income and age in part (ii). Each of the four terms involving inc and age have very significant t statistics. On the other hand, once income and age are controlled for, there seems to be no difference in eligibility by gender. The coefficient on male is very small at given income and age, males are estimated to have a.35 lower probability of being 4(k) eligible and it has a very small t statistic. (iv) Somewhat surprisingly, out of 9,275 fitted values, none is outside the interval [,]. The smallest fitted value is about.3 and the largest is about.697. This means one theoretical problem with the LPM the possibility of generating silly probability estimates does not materialize in this application. (v) Using the given rule, 2,46 families are predicted to be eligible for a 4(k) plan. (vi) Of the 5,638 families actually ineligible for a 4(k) plan, about 8.7 are correctly predicted not to be eligible. Of the 3,637 families actually eligible, only 39.3 percent are correctly predicted to be eligible. (vii) The overall percent correctly predicted is a weighted average of the two percentages obtained in part (vi). As we saw there, the model does a good job of predicting when a family is ineligible. Unfortunately, it does less well predicting correctly less than 4% of the time in predicting that a family is eligible for a 4(k). (viii) The estimated equation is e 4k = inc.6 inc age.3 age 2 38

6 (.8) (.6) (.5) (.39) (.5).38 male +.98 pira (.2) (.22) n = 9,275, R 2 =.95. The coefficient on pira means that, other things equal, IRA ownership is associated with about a.2 higher probability of being eligible for a 4(k) plan. However, the t statistic is only about.62, which gives a two-sided p-value =.5. So pira is not significant at the % level against a two-sided alternative. C7. (i) The average is 9.72, the standard deviation is , the smallest value is 52.32, and the largest value is, Remember, these are in thousands of dollars. (ii) This can be easily done by regressing nettfa on e4k and doing a t test on ˆe4k β ; the estimate is the average difference in nettfa for those eligible for a 4(k) and those not eligible. Using the 9,275 observations gives ˆ β e4k = and te4k = 4.. Therefore, we strongly reject the null hypothesis that there is no difference in the averages. The coefficient implies that, on average, a family eligible for a 4(k) plan has $8,858 more on net total financial assets. (iii) The equation estimated by OLS is nettfa = e4k.278 inc +.3 inc age age 2 n = 9,275, R 2 =.22 (9.96) (.277) (.75) (.6) (.483) (.55) Now, holding income and age fixed, a 4(k)-eligible family is estimated to have $9,75 more in wealth than a non-eligible family. This is just more than half of what is obtained by simply comparing averages. (iv) Only the interaction e4k (age 4) is significant. Its coefficient is.654 (t = 4.98). It shows that the effect of 4(k) eligibility on financial wealth increases with age. Another way to think about it is that age has a stronger positive effect on nettfa for those with 4(k) eligibility. The coefficient on e4k (age 4) 2 is.38 (t statistic =.33), so we could drop this term. (v) The effect of e4k in part (iii) is the same for all ages, For the regression in part (iv), the coefficient on e4k from part (iv) is about 9.96, which is the effect at the average age, age = 4. Including the interactions increases the estimated effect of e4k, but only by $255. If we evaluate the effect in part (iv) at a wide range of ages, we would see more dramatic differences. (vi) I chose fsize as the base group. The estimated equation is 39

7 nettfa = e4k.24 inc +. inc age +.29 age 2 (.2) (.278) (.75) (.6) (.483) (.55).859 fsize fsize fsize fsize5 (.88) (.877) (.868) (2.) n = 9,275, R 2 =.24, SSR = 3,25,27.5 The F statistic for joint significance of the four family size dummies is about With 4 and 9,265 df, this gives p-value =.2. So the family size dummies are jointly significant. (vii) The SSR for the restricted model is from part (vi): SSR r = 3,25,27.5. The SSR for the unrestricted model is obtained by adding the SSRs for the five separate family size regressions. I get SSR ur = 29,985,4. The Chow statistic is F = [(3,25, ,985,4)/ 29,985,4]*(9245/2) With 2 and 9,245 df, the p-value is essentially zero. In this case, there is strong evidence that the slopes change across family size. Allowing for intercept changes alone is not sufficient. (If you look at the individual regressions, you will see that the signs on the income variables actually change across family size.) C7.3 (i) 42/ (ii) The OLS estimates of the LPM are ecobuy = ecoprc +.79 regprc +.55 faminc +.24 hhsize (.65) (.9) (.32) (.53) (.3) n = 66, R 2 = educ.5 age (.8) (.25) If ecoprc increases by, say, cents (.), then the probability of buying eco-labeled apples falls by about.8. If regprc increases by cents, the probability of buying eco-labeled apples increases by about.72. (Of course, we are assuming that the probabilities are not close to the boundaries of zero and one, respectively.) (iii) The F test, with 4 and 653 df, is 4.43, with p-value =.5. Thus, based on the usual F test, the four non-price variables are jointly very significant. Of the four variables, educ appears to have the most important effect. For example, a difference of four years of education implies an increase of.25(4) =. in the estimated probability of buying eco-labeled apples. This suggests that more highly educated people are more open to buying produce that is environmentally friendly, which is perhaps expected. Household size (hhsize) also has an effect. Comparing a couple with two children to one that has no children other factors equal the couple with two children has a.48 higher probability of buying eco-labeled apples. 4

8 (iv) The model with log(faminc) fits the data slightly better: the R-squared increases to about.2. (We would not expect a large increase in R-squared from a simple change in the functional form.) The coefficient on log(faminc) is about.45 (t =.55). If log(faminc) increases by., which means roughly a % increase in faminc, then P(ecobuy = ) is estimated to increase by about.45, a pretty small effect. (v) The fitted probabilities range from about.85 to.5, so none are negative. There are two fitted probabilities above, which is not a source of concern with 66 observations. (vi) Using the standard prediction rule predict one when.5 i ecobuy and zero otherwise gives the fraction correctly predicted for ecobuy = as 2/248.4, so about 4.%. For ecobuy =, the fraction correctly predicted is 34/42.825, or 82.5%. With the usual prediction rule, the model does a much better job predicting the decision to buy eco-labeled apples. (The overall percent correctly predicted is about 67%.) 4