Universidad Carlos III de Madrid

Master in Economics & Master in Industrial Economics & Markets.

Econometrics I
1. Correlation, Causal e¤ects and Randomized Trials
Problem Set 1

1. In a population of couples (man-woman), the men income has an expectation of 50,000e per year with a standard
deviation of 15,000e. The women income has an expectation of 48,000e per year with a standard deviation
of 13,000e. The correlation between men and women income is 0.90. Let C be the total income of a couple
randomly chosen.

a. Compute the expectation of C:

b. Compute the covariance between the income of men and women.
c. Compute the standard deviation of C.
d. Transform the previous answers from euros (e) to dollars ($).

The objective of exercises 2 to 6 is to review the concepts of correlation and best linear predictor (linear
dependence) and conditional expectation (dependence on the mean). The concept of conditional expectation
is central to this course. The conditional expectation will be used to model the causal relationship between
variables.
2. Consider three random variables X; Y; Z. Suppose that Y takes only k values y1 ; :::; yk ; that X takes ` values
x1 ; :::; x` and that Z takes m values z1 ; :::; zm : The joint probability function of X; Y; Z is Pr (X = x; Y = y; Z = z)
and the conditional probability function is
Pr (Y = y; X = x; Z = z)
Pr (Y = y jX = x; Z = z ) = :
Pr (X = x; Z = z)

a. Explain how the conditional probability function of Y (Pr (Y = y)) can be calculated from the joint proba-
bility function.
b. Use the previous results to prove that E (Y ) = E [E (Y jX; Z )] : (Hint: check Stock & Watson Ch. 2).

3. This exercise provides an example where the conditional expectation is not constant (it varies with the explanatory
variable) but the explained and explanatory variable are not correlated. Consider to random variables X and
Z independently distributed as standard normal variates and let Y = X 2 + Z:

a. Compute E (Y jX ) : Is the function m (x) = E (Y jX = x ) a constant function?

b. Show that Y = E (Y ) = 1:
c. Show that E (Y X) = 0 and therefore X;Y = 0:

4. The following table give the joint probability distribution between employment status and college graduation
among those either employed or looking for work (unemployed) in the working age in U.S. on 2008.

Joint Distribution of Employment Status and College of population older than 25 years old in U.S., 2008
Unemployed Employed
Total
(Y = 0) (Y = 1)
Non college grads (X = 0) 0.037 0.622 0.659
College grads (X = 1) 0.009 0.332 0.341
Total 0.046 0.954 1.000

a. Compute E (Y ) ; E (X) ; V ar (Y ) ; V ar (X) ; E (Y X) and Corr (Y; X) :

b. The unemployment rate is the fraction of the labor force that is unemployed. Calculate the unemployment
rate.
c. Calculate E ( Y j X = 1) and E ( Y j X = 0) ; the best linear predictors BLP ( Y j X = 1) and BLP ( Y j X = 0) :
Compare the conditional expectations with the best linear predictors.

1
 d. Calculate the unemployment rate for college graduates and non college graduates.
e. A randomly selected member of this population reports being unemployed. What is the probability that
the worker is a college graduate? and a non-college graduate?
f. Are educational achievement and employment status independent? Justify your answer

5. The joint probability function of Y and X (Pr (Y = y; X = x)) is summarized in the following table,

Pr (Y = y; X = x)
Y nX X=1 X=2 X=3
Y =0 0:15 0:10 0:15
Y =1 0:15 0:30 0:15

a. Find the conditional expectation function E ( Y j X = x) :

b. Find the best linear predictor BLP ( Y j X = x) :
c. Compare the conditional expectation function and the best predictor for the values x = 1; 2; 3 using a plot.
d. Find the marginal distribution of the random variables E ( Y j X) and BLP ( Y j X) :

6. Consider the conditional probability function of Y given X (Pr ( Y = yj X = x)) summarized in the following table

Pr ( Y = yj X = x)
Y nX X = 0 X = 5 X = 10
Y =5 1=3 1=2 1=3
Y = 10 1=3 1=2 1=3
Y = 15 1=3 0 1=3

where the marginal probability function of X (Pr (X = x)) is the following

Pr (X = x)
X=0 X = 5 X = 10
3=10 4=10 3=10

a. Find the conditional expectation function E ( Y j X = x) :

b. Find the best linear predictor BLP ( Y j X = x) :
c. Compare the conditional expectation function and the best linear predictor for the values x = 0; 5; 10 using
a plot.
d. Find the marginal distribution of the random variables E ( Y j X) and BLP ( Y j X) :

7. Compute the following probabilities:

a. If Y is distributed N (50:25) ;…nd Pr (40 Y 52) :

d. If Y is distributed N (0; 1) ; …nd Pr ( 1:96 Y 1:96) :
2
f. If Y is distributed 4 ; …nd Pr (Y 7:78)
2
h. If Y is distributed 10 ; …nd Pr (Y > 18:31)
n
8. Let fYi gi=1 independent and identically Bernoulli distributed random variables with mean 0:4:

a. Use the central limit theorem to compute approximations for

i. Pr Y > 0:43 with n = 100:
ii. Pr Y 0:37 with n = 400:
b. How large would the sample size need to be to ensure that Pr 0:39 Y 0:41 0:95? (Use the central
limit theorem to compute an approximate answer).

9. A survey of 1055 registered voters is conducted, and the voters are asked to choose between candidate A and
candidate B: Let p and p^ denote the fraction of voters who prefer candidate A in the population and in the
sample, respectively.

2
 a. You are interested in the competing hypotheses H0 : p = 0:5 and H1 : p 6= 0:5: Suppose that you decide to
reject H0 if j^
p 0:5j > 0:02:
i. What is the size of this test?
ii. Compute the power of this test if p = 0:53:
b. In the survey p^ = 0:54.
i. Test H0 : p = 0:5 versus H1 : p 6= 0:5; using a 5% signi…cance level.
ii. Test H0 : p = 0:5 versus H1 : p > 0:5; using a 5% signi…cance level.
iii. Construct a 95% con…dence interval for p:
iv. Construct a 99% con…dence interval for p:
v. Construct a 50% con…dence interval for p:
c. Suppose that the survey is carried out 20 times using independently selected voters in each survey. For each
of these 20 surveys a 95% con…dence interval for p is constructed
i. What is the probability that the true value of p is contained in all 20 of these con…dence intervals?
ii. How many of these con…dence intervals do you expect to contain the true value of p?
d. In survey jargon, the "margin of error" is 1.96 Standard Error(^
p); that is, the half of the length of 95%
con…dence interval. Suppose you wanted to design a survey that had a margin of error of at most 1%.
That is, you wanted Pr (j^
p pj > 0:01) 0:05: How large should the sample size n be if the survey uses
simple random sampling?

10. To investigate possible gender discrimination in a …rm, a sample of 100 men and 64 women with similar job
descriptions were selected at random. A summary of the resulting monthly salaries follows:

Average Salary (Y ) Standard Deviation (sY ) n

Men 3,000$ 200$ 100
Women 2,900$ 320$ 64

a. What do these data suggest about wage di¤erences in the …rm? Do they represent statistically signi…cant
evidence that wages of men and women are di¤erent? (To answer this question, …rst state the null and
alternative hypothesis: second, compute the relevant t statistic; third, compute the p value associated
with the t statistic; and …nally use the p value to answer the question.)
b. Do these data suggest that the …rm is guilty of gender discrimination in its compensation policies? Explain.

11. Show the four properties of the conditional expectation in the slides of Chap. 1, page 15.
12. Derive the formulae of and ^ in the slides of Chap. 1 equations (2) and (6), respectively.

13. Show (3) in page 18 of the notes.

14. Obtain the consistency and asymptotic normality of ^ 1 under A1 and A2 in the notes. Check the expression
for the asymptotic variance of ^ 1 under homoskedasticity (A3).
ANSWERS: 1. a) 98,000e; b) 175:50 106 e2 ; c) 27; 295e; d) 1e=1.20$: a’) 117; 600$; b’) 25:272 106 $2 ; c’)
32; 754$:
3. a) No.
4. a) E[Y ] =0.954, E[X] =0.341, Var[Y ] = 4:388 4 10 2 ; Var[X] = 0:224 72; E[XY ] = 0:332; Corr(Y; X) =
6:732 7 10 2 ; b) 0.046; c) E ( Y j X = 1) = 0:974, E ( Y j X = 0) = 0:944, BLP ( Y j X = 1) = 0:974,
BLP ( Y j X = 0) = 0; 944; d) 0:026 and 0:056; e) 0:196 and 0:804; f) No, for ex. Pr(X = 0jY = 0) =
0:804 6= 0:659 = Pr (X = 0).
5. a) E ( Y j X = 1) = 0:5; E ( Y j X = 2) = 0:75; E ( Y j X = 3) = 0:5:
b) BLP ( Y j X = x) = 0:6 : BLP ( Y j X = 1) = 0:6; BLP ( Y j X = 2) = 0:6; BLP ( Y j X = 3) = 0:6:
d) Pr(E ( Y j X) = 0:5) = 0:6; Pr(E ( Y j X) = 0:75) = 0:4; Pr(BLP ( Y j X) = 0:6) = 1:

3
6. a) E ( Y j X = 0) = 10; E ( Y j X = 5) = 7:5; E ( Y j X = 10) = 10:
b) BLP ( Y j X = x) = 9 : BLP ( Y j X = 0) = 9; BLP ( Y j X = 5) = 9; BLP ( Y j X = 10) = 9:
d) Pr(E ( Y j X) = 10) = 0:6; Pr(E ( Y j X) = 7:5) = 0:4: Pr(BLP ( Y j X) = 9) = 1:
7. a) 0:6326; d) 0:950; f) 0:90; h) 0:05:
8. a) i) 0:27; ii) 0:11; b) n 9220:
9. a) i) 0.19; ii) 0.26; b) i) t = 2:61, with Pr (jN (0; 1)j > 2:61) = 0:0094 and reject to 5%; ii) Pr (N (0; 1) > 2:61) =
0:0047 and reject to 5%; iii) (0.51,0.57), iv) (0.5,0.58), v) (0.53,0.55); c) i) 0.36, ii) The 95% of the 20
con…dence intervals d) If n > 9604, the margin of error is less than 0.01 for any p:
10. a) t = 4:4721; p value 0: