Econ 113 Problem Set No.

3: Solutions
Chris Limnios

Problem 3.3 from Chapter 3 Wooldridges Book

Biddle and Hamermesh (1990) study the tradeo between time spent sleeping and working and to look at other factors aecting sleep: sleep = 0 + 1 totwrk + 2 educ + 3 age + u where sleep and totwrk (total work) are measured in minutes per week and educ and age are measured in years. 1. If adults trade o sleep for work, what is the sign of 1 ? This tradeo implies that 1 wil be negative. 2. what signs do you think 2 and 3 will have? A higher level of education may imply a career involving less physical labor, which implies sleep may not be needed as much in comparison with a career involving a high level of physical labor, ceteris paribus. So we would expect a positive 2 . Some studies conclude that the older people get, the less sleep they need, while other studies reveal the opposite. So depending on which study we subscribe to, we would form our expectations accordingly. If we believe that older folks typically sleep more, we would expect a positive 3 . 3. using the data in SLEEP75, the estimated equation is: sleep = 3638.25 .148totwrk 11.13educ + 2.20age = u n = 706 R2 = .113 If someone works ve more hours per week, by how many minutes is sleep predicted to fall? Is this a large tradeo ? Five more hours of work per week corresponds to 300 more minutes per week, which implies that there will be a predicted fall of 300(0.148) = 44.4 minutes.

The tradeo posed is less than one-for-one and therefore represents a small relative tradeo. 4. Discuss the sign and magnitude of the estimated coecient on educ. While the sign is contrary to our hypothesis from earlier, the magnitude is quite small (an extra year of education decreases sleep by approximately 11 minutes per week). This signals that we may want to run a hypothesis test on this variable to see if it is statistically dierent than 0. 5. Would you say totwrk, educ and age explain much of the variation in sleep? What other factors might aect the time spent sleeping? Are these likely to be correlated with totwrk? Recalling from lecture that in social science applications, a value of R2 .10 is not uncommon, an R2 of 0.113 implies that the model specied is of average performance. Other factors which may eect time spent

sleeping may include the type of shift one works (graveyard vs. swing) or stress level (or possibly some other health-related factors). We would typically assume that these factors are probably correlated with totwrk (the more one works, the higher their stress level, typically and people who work graveyard may be putting in 12 hour shifts; most police and reghters who work graveyard typically put in 12 hour shifts).

Problem 3.7 from Chapter 3 Wooldridges Book

Which of the following cause OLS estimators to be biased? 1. heteroskedasticity Heteroskedasticity does not cause OLS estimators to be biased. 2. omitting an important variable Omitting an important variable would create biased OLS estimators (omitted variable bias). 3. a sample correlation coecient of .95 between independent variables both included in the model. Adding in a correlated variable would aect the standard errors in the OLS estimators but would not cause bias.

Problem 4.2 from Chapter 4 Wooldridges Book

Consider an equation to explain salaries of CEOs in terms of annual rm sales, return on equity (roe, in percentage form), and return on the rms stock (ros, in percentage form). log(salary) = 0 + 1 log(sales) + 2 roe + 3 ros + u 1. In terms of the model parameters, state the null hypothesis that, after controlling for sales and roe, ros has no eect on CEO salary. State the alternative that better stock market performance increases a CEOs salary. Our hypotheses would be structured as: H0 : 3 = 0 HA : 3 > 0. 2. Using the data in CEOSAL1, the following equation was obtained by OLS: log(salary) = 4.32 + 0.280 log(sales) + 0.0174 roe + 0.00024ros
(0.32) (0.035) (0.0041) (0.00054)

n = 209 R2 = 0.283 By what percent is salary predicted to increase if ros increases by 50 points? Does ros have a practically large eect on salary? If ros increases by 50 points, salary is predicted to increase by 0.00024 50 100 = 1.2% This depends on the actual unit of measurement. If by 50 points, we mean 0.5% (such as in basis-points), then it would have a large eect. If by 50 points, we mean 50%, then the eect is trivial.

3. Test the null hypothesis that ros has no eect on salary against the alternative that ros has a positive eect. Carry out the test at the 10% signicance level. We will run a one-sided test. The critical value for a one-sided test at the 10% signicant level for 209 - 3 1 degrees of freedom is 1.282. The test statistic is 0.00024 0 = 0.444 = 0 < t < tC 0.00054 so that we would accept the null hypothesis and conclude that the estimated coecient on ros is no dierent than 0 statistically. 4. Would you include ros in a nal model explaining CEO compensation in terms of rm performance? Explain. Given that we have concluded that the coecient on this variable is statistically 0, I would not include it in a nal model explaining compensation for this sample of CEOs (It may be a sample of non-prot organizations).

Problem 4.5 from Chapter 4 Wooldridges Book

Consider the estimated equation from example 4.3, which can be used to study the eects of skipping class on college GPA: colGP A = 1.39 + 0.412 hsGP A + 0.015 ACT + 0.083 skipped
(0.33) (0.094) (0.011) (0.026)

n = 141, R2 = 0.234. 1. Using the standard normal approximation, nd that 95% condence interval for hsGP A . The 95% condence interval for hsGP A is given by 0.412 1.980(0.094) = [0.22588, 0.59812], where 0.412 is for this variable, 1.980 is the 2-tailed, 5% signicance level at 137 degrees of freedom critical value and 0.94 is the standard error given for hsGP A . 2. Can you reject the hypothesis H0 : hsGP A = 0.4 against the two-sided alternative at the 5% level? The critical value tC is 1.980. The test statistic is t = 0.412 0.4 = 0.127659574 = 0 < t < tC , 0.094

which implies that we cannot reject the hypothesis. 3. Can you reject the hypothesis H0 : hsGP A = 1 against the two-sided alternative at the 5% level? The critical value tC is 1.980. The test statistic is t = 0.412 1 = 6.25531915 = t < tC < 0, 0.094

which implies that we can reject the hypothesis.

Problem C3.1 from Chapter 3 Wooldridges Book

bwght = 0 + 1 cigs + 2 f aminc + u (1)

where bwght is infant birth weight, cigs is smoking during pregnancy, and f aminc is family income. 1. what is the most likely sign for 2 ? Since a low level of family income may be correlated with a low level of nutrition in the diet, we would expect the coecient to be positive. 2. Do you think cigs and f aminc likely to be correlated? Explain why the correlation might be positive of negative? We would expect a negative level of correlation between these two variables. Typically, people with a high level of education have a better understanding of the health risks entailing smoking. Likewise, people with a high level of education typically have higher salaries. So we would expect to see families with a high level of income smoke less. 3. Now, estimate the equation with and without f aminc , using the date in BWGHT. Report the results in equation form, including the sample size and R-squared. Discuss your results, focusing on whether adding faminc substantially changes the estimated eect of the other variable. The code to perform what is asked of us is given below, along with the output. > > > > data = read.table("http://people.ucsc.edu/~archeng/teach/econ113/bwght.txt", header = T) with_family_income = lm(bwght ~ cigs + faminc, data) without_family_income = lm(bwght ~ cigs, data) summary(with_family_income)

Call: lm(formula = bwght ~ cigs + faminc, data = data) Residuals: Min 1Q -96.061 -11.543

Median 0.638

3Q Max 13.126 150.083

Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 116.97413 1.04898 111.512 < 2e-16 *** cigs -0.46341 0.09158 -5.060 4.75e-07 *** faminc 0.09276 0.02919 3.178 0.00151 ** --Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1 Residual standard error: 20.06 on 1385 degrees of freedom Multiple R-squared: 0.0298, Adjusted R-squared: 0.0284 F-statistic: 21.27 on 2 and 1385 DF, p-value: 7.942e-10 > summary(without_family_income) Call: lm(formula = bwght ~ cigs, data = data)

Residuals: Min 1Q -96.772 -11.772

Median 0.297

3Q Max 13.228 151.228

Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 119.77190 0.57234 209.267 < 2e-16 *** cigs -0.51377 0.09049 -5.678 1.66e-08 *** --Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1 Residual standard error: 20.13 on 1386 degrees of freedom Multiple R-squared: 0.02273, Adjusted R-squared: 0.02202 F-statistic: 32.24 on 1 and 1386 DF, p-value: 1.662e-08 Re-expressing the above in equation form, bwghtw/f aminc = 116.97413 0.46341cigs + 0.09276f aminc
(1.04898) (0.09158) (0.02919)

n = 1388, R = 0.0298 bwghtw/of aminc = 119.77190 0.51377cigs

(0.57234) (0.09049)

n = 1388, R = 0.02273 As we can see, adding f aminc increases the standard error on cigs from 0.091 to 0.092, which is trivial. The R2 statistic increases from 0.02 to 0.03, which implies that the inclusion of f aminc is justied.

Problem C3.6 from Chapter 3 Wooldridges Book

Use the data set in WAGE2 for this problem. As usual, be sure all of the following regression contain an intercept. 1. Run a simple regression of IQ on educ to obtain the slope coecient, 1 . > data = read.table("http://people.ucsc.edu/~archeng/teach/econ113/WAGE2.txt", header = T) > no.1 = lm(IQ ~ educ, data) > summary(no.1) Call: lm(formula = IQ ~ educ, data = data) Residuals: Min 1Q -50.228 -7.262

Median 0.907

3Q 8.772

Max 37.373

Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 53.6872 2.6229 20.47 <2e-16 *** educ 3.5338 0.1922 18.39 <2e-16 *** ---

Signif. codes:

0 *** 0.001 ** 0.01 * 0.05 . 0.1

Residual standard error: 12.9 on 933 degrees of freedom Multiple R-squared: 0.2659, Adjusted R-squared: 0.2652 F-statistic: 338 on 1 and 933 DF, p-value: < 2.2e-16 The output implies that 1 = 3.5338. 2. Run the simple regression of log(wage) on educ, and obtain the slope 1 . > no.2 = lm(lwage ~ educ, data) > summary(no.2) Call: lm(formula = lwage ~ educ, data = data) Residuals: Min 1Q -1.94620 -0.24832

Median 0.03507

3Q 0.27440

Max 1.28106

Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 5.973062 0.081374 73.40 <2e-16 *** educ 0.059839 0.005963 10.04 <2e-16 *** --Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1 Residual standard error: 0.4003 on 933 degrees of freedom Multiple R-squared: 0.09742, Adjusted R-squared: 0.09645 F-statistic: 100.7 on 1 and 933 DF, p-value: < 2.2e-16 The output implies 1 = 0.059839. 3. Run the multiple regression of log(wage) on educ and IQ, and obtain the slopes 1 , 2 , respectively. > no.3 = lm(lwage ~ educ + IQ, data) > summary(no.3) Call: lm(formula = lwage ~ educ + IQ, data = data) Residuals: Min 1Q -2.01601 -0.24367

Median 0.03359

3Q 0.27960

Max 1.23783

Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 5.6582876 0.0962408 58.793 < 2e-16 *** educ 0.0391199 0.0068382 5.721 1.43e-08 *** IQ 0.0058631 0.0009979 5.875 5.87e-09 *** --Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1

Residual standard error: 0.3933 on 932 degrees of freedom Multiple R-squared: 0.1297, Adjusted R-squared: 0.1278 F-statistic: 69.42 on 2 and 932 DF, p-value: < 2.2e-16 The output implies 1 and 2 are equal to 0.0391199 and 0.0058631, respectively. 4. Verify that 1 = 1 + 2 1 . 0.0391199 + 0.0058631 3.5338 = 0.0598389228 0.059839, as desired.