Midterm II, Econ120C
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Midterm II, Econ120C
Spring 2019
1. (20 points, 4 pts each) Do you agree or disagree with the following statements? Circle your answer. You do not need to provide any explanation. Should you decide to change an answer, please write in words what your Önal answer is. For example, you may write ìMy Önal answer to this question is ëAgreeîí.
(a) In a simple IV regression model Yi = a + βXi + ui with a single regressor Xi and a single instrument Zi ; we can check whether Zi is uncorrelated with ui by calculating the sample correlation between Zi and the IV residual i .
Agree Disagree
(b) Consider a simple IV regression Y = a + Xβ + u with Z as the instrument. Suppose the sample covariance c一ov (Z; X) 0: If the Stata command ìreg Y Zîgives us a zero estimate for the coe¢ cient of Z, then the IV estimate of β must be zero too.
Agree Disagree
(c) In an IV regression with strong instruments, the two stage least squares estimator is con- sistent and unbiased.
Agree Disagree
(d) The distinction between endogenous and exogenous variables is whether the variables are correlated with the error term or not.
Agree Disagree
(e) Consider an IV regression with one endogenous regressor X and one exogenous regressor W and one instrument Z . If the coe¢ cient estimate for Z is zero in the Örst stage least square regression, then the second stage least square regression will su§er from perfect multicollinearity.
Agree Disagree
2. (23 points) Consider the demand and supply equations
Qi(s) = y + 6Pi + Vi ; (2)
for some β < 0 and 6 > 0; where Qi(d) and Qi(s) are the demand and supply, Pi is the sticker price, Ti is the general sales tax, and Ci is the product-speciÖc tax (i.e., the cigarettes exercise tax).
Note that i := Pi + Ti + Ci is the total price (TP) paid by consumers.
Solving the above two equations, we obtain the market price and sales:
a6 - βy β6 6Ui - βVi
6 - β 6 - β 6 - β
a - y β Ui - Vi
6 - β 6 - β 6 - β
Assume that
Ci = Ui + "i
and that "i; Ui; Vi and Ti are mutually independent of each other.
The observations are (Qi; Pi ; i; Ti; Ci) : We do not observe Ui and Vi:
(a) (7 pts) Let OLS be the OLS estimator of the slope coe¢ cient obtained by regressing Qi on Pi: Is OLS consistent for 6? If yes, explain why. If not, is OLS expected to be larger or smaller than 6?
(b) (8 pts) Suppose we estimate the supply curve using C and T as the instruments. That is, we run the IV regression ìivreg Q (P = C T)î to otbain the slope estimator IV : Is the IV estimator IV consistent for 6? Explain.
(c) (8 pts) Suppose we estimate the demand curve using C as the instrument. That is, we run the IV regression ìivreg Q (TP = C)îto obtain the slope estimator IV . Is the IV estimator IV consistent for β? Explain.
3. (20 points, 5pts each) This is a multiple-choice question. Write down your answers at the end of this question. You DO NOT need to explain your reasoning.
A researcher is interested in estimating the e§ects of crime rate on government spending on
public safety. Based on the data in 1996, the researcher speciÖes the following model crimespendpc = a0 + a1rate + a2 staterevpc + a3 experstud + "
where crimespendpc is the per capita state government expenditure on public safety; rate is the crime rate (number of crimes committed per 100,000 population); staterevpc is the state government revenue per capita; and experstud is the state government expenditure per student on K12 education in $. The researcher believes that rate is endogeneous and other included regressors are exogeneous.
(i) The researcher has information on rate90 (the crime rate in 1990) and democrat (the per- centage of votes for the democratic presidential candidate in 1996). If the researcher believes that rate90 and democrat are potentially valid instruments and proceeds to estimate the above model by 2SLS. In the Örst stage regression, he should
(a) Regress rate on a constant, rate90 and democrat
(b) Regress rate on a constant, rate90, democrat, staterevpc; and experstud (c) Regress crimespendpc on a constant, rate90 and democrat
(d) Regress crimespendpc on a constant, rate90, democrat, staterevpc; and experstud
(ii) With the Örst stage estimation, the researcher would like to test whether the instruments are jointly strong or not. He should
(a) test for the joint signiÖcance of rate90 and democrat
(b) test for the joint signiÖcance of all the exogenous variables
(c) check the reported overall F-test for signiÖcance
(d) all of the above
(iii) The researcher estimates the model using 2SLS and saves the residuals ( 2SLS). If he wants
(a) Regress 2SLS on a constant, staterevpc and experstud and test the joint signiÖcance of
(b) Regress 2SLS on a constant, rate90 and democrat and test the joint signiÖcance of the
(c) Regress 2SLS on a constant, rate90, democrat, staterevpc; and experstud and test the
(d) Regress 2SLS on a constant, rate90, democrat, staterevpc; and experstud and test the
(iv) Finally, the researcher performs the J-test correctly and Önds that the p-value is 0:001. This result would imply that the researcher should
(a) not reject the null that rate90 and democrat are jointly weak
(b) not reject the null that rate90 and democrat are jointly exogeneous
(c) reject the null that rate90 and democrat are jointly weak
(d) reject the null that rate90 and democrat are jointly exogeneous
Your choices for Question 3 are:
Question 3 (i):
Question 3 (ii):
Question 3 (iii):
Question 3 (iv):
4. (32 points) Read the following Stata program and output:
* beginning of the program clear
set memory 100m
postfile tempid beta1 beta2 beta3 reject2 reject3 using ///
mydata .dta,replace
forvalues i = 1(1)10000 {
drop _all
quietly set obs 1000
gen z1 = rnormal()
gen z2 = rnormal()
gen u = rnormal()
gen e1 = rnormal()
gen e2 = rnormal() /*e1,e2,z1, z2,and u are independent standard normals */
gen x1 = z1 + z2 + e1
gen x2 = u + e2
gen x = x1 + x2
gen y = 2*x + u
quietly reg x z1 z2, r
quietly predict x_hat, xb
reg y x_hat, r
sca beta1 = _b[x]
quietly reg y x, r
scalar beta2 = _b[x]
quietly test x = 2
scalar reject2 = (r(p)<0 .10)
quietly ivreg y (x=z1 z2), r
scalar beta3 = _b[x]
quietly test x = 2
scalar reject3 = (r(p)<0 .10)
post tempid (beta1) (beta2) (beta3) (reject2) (reject3)
}
postclose tempid
use mydata .dta, clear
sum
* end of the program
Part of the output is given in the table below but the numbers a; β , y and a and b are missing.
Variable |
Mean |
beta1 |
a |
beta2 |
β |
beta3 |
y |
reject2 |
a |
reject3 |
b |
(a) (12 points, 4 pts for each parameter) What would you expect the values of a; β and y to be? Present your argument in details. If you think any of the missing numbers cannot be determined from the information given here, explain why.
(b) (8 points, 4 pts for each parameter) What would you expect the values of a and b to be? Present your argument in details. If you think any of the missing numbers cannot be determined from the information given here, explain why.
(c) (6 pts) Do you expect b to be di§erent if the command ìgen x1 = z1 + z2 + e1îin the above problem is replaced by ìgen x1 = e1î? Explain.
(d) (6 pts) Do you expect b to be di§erent if the command ìquietly ivreg y (x = z1 z2), rîin the above problem is replaced by ìquietly ivreg y (x = z1 z2)î? Explain.
2022-08-18