In 1992, economists Card and Krueger collected data on fast food restaurants in New Jersey and Pennsylvania. In particular, they contacted all of the Burger King, Kentucky Fried Chicken, Wendy's, and Roy Rogers restaurants listed in the white pages of the telephone books in New Jersey and eastern Pennsylvania. Among the restaurants contacted, 473 had valid telephone numbers and 410 of these responded to the survey. Another economist supplemented the data set with demographic and economic variables that could potentially influence the prices of both food and beverages sold at fast food restaurants. Specifically, she obtained the zip code of each restaurant and added demographic and economic characteristics within that zip code. BK - 1 if Burger King, = 0 otherwise KFC = 1 if Kentucky Fried Chicken. - O otherwise RR = 1 if Roy Rogers, - 0 otherwise NJ = 1 if New Jersey, -0 otherwise (eastern Pennsylvania) PFRIES - price of a small order of French fries PSODA - price of a medium soda INCOME - median family income in the zip code (dollars) PRPBLCK - proportion of Black households in the zip code MOOLA BV 80 F MODEL A Dependent Variable: PSODA Method: Least Squares Sample: 1410 Included observations: 401 Variable Coefficient Std. Error 1-Statistic Prob C PRPBLCK BK KFC RR 0.961884 0.082852 0.085047 0.031048 0.135303 0.009993 0.020813 0.011359 0.012940 0.012425 96.25604 3.980755 7.487232 2.399336 10.88964 0.0000 0.0001 0.0000 0.0169 0.0000 1.044863 R-squared Adjusted R-squared SE of regression Sum squared resid Log likelihood Durbin-Watsog stat 0.287242 Mean dependent var 0.280042 SD dependent var 0.075345 Akalke info criterion 2.248052 Schwarz criterion 470.3771 F-statistic 1.462692 Prob(F-statistic) 0.088798 -2.321083 -2.271293 39.89703 0.000000 MODEL B Dependent Variable: PSODA Method: Least Squares Sample 1 410 Included observations: 401 Variable Coefficient Sid. Error Statistic Prob O B85365 0.010267 52 76948 0.0000 stv AR MODEL B Dependent Variable: PSODA Method: Least Squares Sample: 1410 Included observations: 401 Variable Coefficient Std. Error t-Statistic Prob. PRPBLCK INCOME BK 0.016967 0.022000 3.05E-07 0.895365 0.057638 2.49E-07 0.079611 0.025188 0.127442 0.077750 52.76948 2.619923 0.813772 7633000 2.124973 11.05053 8.360006 0.010430 0.011853 0.0000 0.0091 0.4163 0.0000 0.0342 0.0000 0.0000 KFC RR NJ 0.011533 0.009300 R-squared Adjusted R-squared SE of regression Sum squared resid Log likelihood Durbin Watson stat 0.406824 Mean dependent var 0.397791 S.D. dependent var 0.068909 Akaike info criterion 1.870889 Schwarz criterion 507.1990 F-statistic 1.746790 Prob(F-statistic) 1.044863 0.088798 -2.494758 2.425038 45.03678 0.000000 MODEL C Dependent Variable PSODA Method:Least Squares Sample 1 410 Included observations 391 JP stv MODEL C Dependent Variable: PSODA Method: Least Squares Sample: 1410 Included observations: 391 Variable Coefficient Std. Error 1-Statistic Prob 0.031288 PRPBLCK INCOME BK KFC 0,587703 0.036566 -1.09E-07 0.066577 0.038109 0.083831 0.042735 0.400242 0.019134 2.68E-07 0.009123 0.010344 0.010898 0.008673 0.035910 18.78366 1.911017 -0.405990 7297549 3.684257 7.692683 4.927477 11.14573 0.0000 0.0567 0.6850 0.0000 0.0003 0.0000 0.0000 0.0000 RR NJ PFRIES R-squared Adjusted R-squared SE of regression Sum squared resid Log likelihood Durbin-Watson stat 0.558109 0.550032 0.059311 1.347315 553.7961 1 858383 Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob{F-statistic) 1.044962 0.088419 -2.791796 -2.710595 69.10415 0.000000 OV X POW Question 2 1 pts In particular, does one of these models violate an important assumption? If so, which assumption? Model A violates the zero conditional mean assumption, causing omitted variable bias. Model B includes irrelevant variables. Model A is too simple. Model B has multicollinearity issues Question 3 1 pts Explain why you believe the model in question violates this particular assumption TV R. 75 S Question 3 Explain why you believe the model in question violates this particular assumption. If INCOME and NJ are excluded from the model. Elu (PRPBLCK) + O for all values of PRPBLCK, because INCOME and NJ affect the price of SODA. NJ should not be included in the model because it is irrelevant to the price of soda. IF INCOME and NJ are excluded from the model, Elu |PRPBLCK) * O for all values of PRPBLCK, because INCOME and States are correlated. If INCOME and NJ are excluded from the model, Eu |PRPBLCK) * O for all values of PRPBLCK, becamse u and PRPBLCK are correlated with INCOME and vary by state. Question 4 1 pts What are the implications of violating this assumption? How would you explain the issue to a fast-food executive? TV A OME and N PRPBLCK, because u and PRPBLCK are correlated with INCOME and vary by state. Question 4 1 pts What are the implications of violating this assumption? How would you explain the issue to a fast-food executive? We need to collect more data so our estimation can be more precise. We need to control for other relevant variables that are correlated with the racial composition of the zip code. Otherwise the coefficient on racial composition will partly reflect the roles of these other variables. We should avoid the multicollinearity issue, which makes the estimator biased. We should not include irrelevant variables, which make the estimator inefficient