Problem Set 2, Econ 120C
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Problem Set 2, Econ 120C
Question I Consider the causal model
Y = a + Xfβ + Xf(2)(y + Xo)
where Xf stands for the causal factor of interest, and Xo stands for other causal factors. (a) What is the ceteris paribus causal e§ect of Xf on Y? Please provide a mathematical
deÖnition.
(b) Does the ceteris paribus causal e§ect deÖned in (a) depend on the levels of Xf and Xo? Explain.
(c) Suppose (Xfi; Xoi) are iid draws from independent standard normal distributions, i.e., Xfi ~ iidN (0; 1); Xoi ~ iidN (0; 1), {Xfi} and {Xoi} are independent. We observe Xfi and
Yi = a + Xfiβ + Xf(2)i(y + Xoi)
for i = 1; :::; n: Suppose the sample size n is large, and Y is regressed on a constant and Xf: What do you expect the coe¢ cient estimator associated with Xf to be? Please present your answer in terms of the parameters a; β; and y:
Question II (a) Consider the causal model y = x3 + u: Given some iid draws (Xi; ui) where Xi and ui are independent standard normals, we obtain
Yi = Xi(3) + ui
Compute E (Yi|Xi = x) :
(b) Based on observations (Xi; Yi) from the above model, we run the following regression
Yi = a + Xib + errori
to obtain the OLS estimator OLS and OLS : What would you expect the value of OLS to be when n 二 &? What would you expect the value of OLS to be when n 二 &?(hint: EX = EX3 = 0; EX2 = 1; and EX4 = 3)
(c) Graph the functions y = x3 and y = a* + xb* in the same graph where a* and b* are the limit values of a* and b* that you Önd in (b). Do you expect the two functions to be close to each other? Explain.
Question III Consider the two-equation causal/structural model:
Y1 = 5 + Y2 + "2 (2)
where "1 and "2 are independent standard normal. To help you understand the equation system, you can think of the Örst equation as the demand curve where Y1 is the demand and Y2 is the price, and the second equation as the supply curve where Y1 is the supply and Y2 is the price.
1. Solve Y1 and Y2 in terms of ("1 ; "2 ).
2. Find cov (Y2; "1 ) and cov (Y2; "2 ) :
3. Given some iid draws ("1i; "2i) ; the two-equation system produces the observations (Y1i; Y2i) : Let OLS be the OLS estimator obtained by regressing Y1 on a constant and Y2: What do you expect OLS to be when n 二 &?
4. ìIn large samples, we expect OLS to fall in the interval [-2,1]î. Do you agree with this statement? Explain.
Question IV Consider the following causal model
y - bx + v;
x - cy + u;
where b 0; c 0 and bc 1: Suppose that the values of (u; v) are generated from
(U; V) ~ N ┌0; 、┐ :
We do not observe (U; V) but we could observe the equilibrium solution (X ,Y).
(a) Consider the special case with 7uu = 0 but 7vv 0 so that the model becomes
y - bx + v;
x - cy:
What is best linear prediction of Y given X (i.e., suppose I give you the value of X but withhold the value of Y; what would be your best guess of Y according to the MSE criterion?). What is best linear prediction of X given Y?
(b) Consider the special case with 7vv = 0 but 7uu 0 so that the model becomes
y - bx;
x - cy + u
What is best linear prediction of Y given X? What is best linear prediction of X given Y?
(c) Now suppose 7uu 0 and 7vv 0: Under what condition(s) is X useless as a predictor of Y? Under what condition(s) is Y useless as a predictor of X? Explain.
Question V (Optional. Your grade will not be a§ected whether you work on this one or not ) (a) Explain the terms ìactive predictionîand ìpassive predictionî
(b) We have been discussing two di§erent betaís: β and β * : What is the di§erence between these two betaís? Under what conditions, β = β * ? When should we use β? When should we use β * ?
Question VI (Optional. Your grade will not be a§ected whether you work on this one or not ) Read the Stata program below
clear
scalar alpha = sqrt(2)
scalar beta = 3 .14
scalar gamma = exp(1)
capture postclose tom
set seed 1
forvalues i = 1(1) 100 {
qui drop _all
qui set obs 100
gen z = rnormal()
gen x = z + rnormal()
gen u = rnormal()
gen y = alpha + x*beta + z*gamma+u
drop u
/*save xyz .dta, replace */
quietly reg y x z,r
scalar beta_hat = _b[x]
scalar gamma_hat = _b[z]
quietly reg y x,r
scalar beta_star_hat = _b[x]
quietly reg z x,r
scalar delta_hat = _b[x]
post fileid (beta_star_hat) (beta_hat) (gamma_hat) (delta_hat)
}
postclose fileid
use OVB .dta,clear
gen bias = beta_star_hat - beta_hat
gen proxy_effect = gamma_hat*delta_hat
sum
The sum command provides the summary statistics for each variable in the data set OVB.dta. What would you expect the mean of each variable to be? Explain. Is the mean of ëbiasíclose to that of ëproxy_e§ectí? Explain.
2022-08-18