Econ 140 (Fall 2014) - Final Review

GSIs: Marion Aouad, Fenella Carpena,

∗
Alessandra Fenizia, Paolo Zacchia

Exercises
The probability of event A or B (i.e. P r(A or B)) equals:
a. P r(A) × P r(B) .
b. P r(A) + P r(B) if A and B are mutually exclusive.
c. P r(A)
P r(B)
d. P r(A) + P r(B) even if A and B are not mutually exclusive.
2. The cumulative probability distribution shows the probability:
a. That a random variable is less than or equal to a particular value.
b. Of two or more events occurring at once.
c. Of all possible events occurring.
d. That a random variable takes on a particular value given that an-
other event has happened.
3. The expected value of a discrete random variable is:
a. Is the outcome that is most likely to occur.
∗ We thank previous GSIs for the great section material they built over time. These section

notes are heavily based on their previous work.

1
b. Can be found by determining the 50% value in the c.d.f.
c. Equals the population median.
d. Is computed as a weighted average of the possible outcome of that
random variable, where the weights are the probabilities of that
outcome.
4. Two random variables X and Y are independently distributed if all of the
following conditions hold, with the exception of:
a. P r(Y = y|X = x) = P r(Y = y)
b. Knowing the value of one of the variables provides no information
about the other.
c. If the conditional distribution of Y given X equals the marginal
distribution of Y.
d. E(Y ) = E[E(Y |X)]

5. The correlation between X and Y (X & Y are random variables):

a. Cannot be negative since variances are always positive.
b. Is the covariance squared.
c. Can be calculated by dividing the covariance between X and Y by
the product of the two standard deviations.
d. Is given by ρ = corr(X, Y ) = var(X)var(Y
xy
cov(X, Y )
)

6. If Y is the sample mean drawn from an iid random sample of size N, the
variance of Y , σ , is given by the following formula:
2
Y

a. σY2

b. σ
√Y
N

c. σY2
N

d. σ2
√Y
N

2
7. The central limit theorem states that:
a. The sampling distribution of Y −σ µ is approximately normal.
Y
Y

b. Y →µ .
p
Y

c. The probability that Y is in the range µ ± c becomes arbitrarily

close to one as n increases for any constant c > 0 .
Y

8. An estimator is:
a. An estimate.
b. A formula that gives an ecient guess of the true population value.
c. A random variable.
d. A nonrandom number.
9. An estimate is:
a. Ecient if it has the smallest variance possible.
b. A nonrandom number.
c. Unbiased if its expected value equals the population value.
d. Another word for estimator.
10. An estimator µ̂ of the population parameter µ is unbiased if:
Y Y

a. µ̂Y = µY
b. Y has the smallest variance of all estimators.
c. Y →µ .
p
Y

d. E[µ̂Y ] = µY

11. An estimator µ̂ of the population value µ is consistent if:

Y Y

a. µ̂Y →µY
p
.
3
b. Its mean square error is the smallest possible.
c. Y is normally distributed.

d. Y → 0.
p

12. An estimator, µ̂, of the population parameter, µ , is more ecient when

compared to another estimator, µe , if:
Y

a. E[µ̂] > E[e

µ]
b. has a smaller variance.
µ̂
c. The cdf of µ̂ is atter than the cdf of the other estimator.
d. Both estimators are unbiased, and var(µ̂)< var(eµ).
13. The critical value of a two-sided t-test computed from a large sample:
a. Is 1.64 if the signicance level of the test is 5%.
b. Cannot be calculated unless you know the degrees of freedom.
c. Is 1.96 if the signicance level of the test is 5%.
d. Is the same as the p-value.
14. The standard error for the dierence in means of two independent random
variables M and W , when the two population variances are dierent and s is 2

the sample variance, is:

a.
s
s2M + s2W
nM + nW

b. s2M + s2W
nM + nW

c.
s
1 s2 + s2W
× M
2 nM + nW

d.
s
s2M s2
+ W
nM nW

15. The correlation coecient:

4
a. Lies between zero and one.
b. Is a measure of linear association.
c. Is close to one if X causes Y.
d. Takes on a high value if you have a strong nonlinear relationship.
16. When you are testing a hypothesis against a two-sided alternative, then the
alternative is written as:
a. E[Y ] > µY,0
b. E[Y ] = µY,0
c. Y 6= µY,0
d. E[Y ]6= µY,0

17. Suppose that a researcher, using data on class size (CS) and average test
scores from 100 third-grade classes, estimates the OLS regression.
520.4 5.82
T estScore
d = − × CS, R2 = 0.08, SER = 11.5.
(20.4) (2.21)

a) Construct a 95% condence interval for β , the regression slope co-

ecient.
1

b) Calculate the p − value for the two-sided test of the null hypothesis
H : β = 0. Do you reject the null hypothesis at the 5% level? At
the 1% level?
0 1

c) Calculate the p − value for the two-sided test of the null hypothesis
H : β = −5.6. Without doing any additional calculations, deter-
mine whether -5.6 is contained in the 95% condence interval for
0 1

β .
1

d) Construct a 99% condence interval for β .0

e) The regression R is 0.08. What are the units of measurement for

2

the R (score points? number of students? or is R unit-free?)?

2 2

What does R = 0.08 mean?

2

18. In a population µ = 100 and σ = 43. Use the Central Limit Theorem to
2

answer the following questions:

Y Y

a) In a random sample of size N = 100, nd P (Y < 101).

5
b) In a random sample of size N = 64, nd P (101 < Y < 103).
c) In a random sample of size N = 165, nd P (Y > 98).
19. Data on fth-grade test scores (reading and mathematics) for 420 school
districts in California yield Ȳ = 654.2 and sample standard deviation s = 19.5.
Y

a) Construct a 95% condence interval for the mean test score in the
population.
b) When the districts were divided into districts with small classes
(< 20 students per teacher) and large classes (≥ 20 students per
teacher), the following results were found:
Class Size Average Score Standard Deviation n
Small 657.4 19.4 238
Large 650.0 17.9 182
Is there statistically signicant evidence that the districts with smaller classes
have dierent average test scores than districts with larger classes? Explain.
20. Suppose the graph of wage and experience has an inverted-U shape in the
data, but you estimate the model W age = β + β Experience + ε , instead of
∗ ∗ ∗

the model that includes the quadratic term on experience. What can you say
i 0 1 i i

about the direction of the bias of the slope coecient? (Hint: use the omitted
variable bias formula).
21. In the case of perfect multicollinearity, OLS is unable to estimate the co-
ecients of the variables involved. Assume that you have included both X
and X as explanatory variables, and that X = X , so that there is an exact
1
2

relationship between the two explanatory variables. Does this pose a problem
2 2 1

for estimation?
22. You have obtained data on test scores and student-teacher ratios in region
A and region B of your state. Region B, on average, has lower student-teacher
ratios than region A. You decide to run the following regression:
Y = β0 + β1 X1 + β2 X2 + β3 X3 + ε

where X is the class size in region A, X is the dierence in class size between
region A and region B, and X is the class size in region B. Your regression
1 2

software shows a message indicating that it cannot estimate that model. What
3

is the problem and how can it be xed?

6
23. Suppose we have the following system of equations:
Yi = βXi + εi
Xi = γZi + ui

where Z is an instrument for X (i.e., Cov(X, Z) 6= 0 but Cov(Z, ε) = 0).

i i

a) Duo and Pande (2007) are interested in evaluating the eect of

1
dams on agricultural productivity. Let X be an indicator for whether
or not a district i in India has a dam, and let Y denote that district's
i

agricultural productivity. Why might Cov(X, ε) 6= 0?

i

b) Duo and Pande (2007) use the "slope of land" in district i as an

instrument, denoted by Z , for whether or not a district i has a dam.
The idea is that dams cannot be placed in areas that are too steep
i

or too at (otherwise they will not work). Do you think this is a
valid instrument? Why or why not?

24. Suppose you are interested in the eect of sales revenue on stock market
share price- you collect monthly data over 4 full years (t =1,2,...,48) on the
stock market share price (SP RICE ) of 4 companies (i=1,...,4) along with
their monthly revenues (REV EN U E ). You are concerned that other factors,
it

in particular industry (e.g., banking) and location (e.g., California) may aect
it

both the share price and revenues. Unfortunately, you do not have data on the
industry or the location of the companies.
a) Express a xed eects model for these data and explain how it can
solve the problem of the unobserved data.
b) b) Suppose that you also believe that the cost of capital facing all of
these companies varied over the four-year period but, again, you do
not have data on that variable. Explain how you could control for
this monthly variationin addition to industry and locationusing
a dummy variables regression.

1 Duo, E., and R. Pande. (2007). "Dams", Quarterly Journal of Economics (May), pp.
601-646.

7
Solutions
1. b
2. a
3. d
4. d
5. c
6. c
7. a
8. c
9. b
10. d
11. a
12. d
13. c
14. d
15. b
16. d
17.
a) The 95% condence interval for β is [−5.82±1.96×2.21]=[−10.152, −1.4884].
1

b) Calculate the t-statistic:

βˆ1 − 0 −5.82
tstat =  = = −2.6335.
SE β1 ˆ 2.21

The p − value for the test H0 : β1 = 0 vs.

H1 : β1 6= 0 is
 stat 

p − value = 2Φ − t  = 2Φ (−2.6335) = 2 × 0.0042 = 0.0084.

The p − value is less than 0.01, so we can reject the null hypothesis at the 5%
signicance level, and also at the 1% signicance level.
c) The t-statistic is:
βˆ1 − (−5.6) −0.22
tstat =   = = −0.10.
SE βˆ1 2.21

The p − value for the test H : β1 = −5.6

0 vs.
H1 : β1 6= −5.6 is
p − value = 2Φ − tstat  = 2Φ (−0.10) = 0.92.
 

8
The p − value is larger than 0.10, so we cannot reject the null hypothesis at the
10%, 5% or 1% signicance level. Because β = −5.6 is not rejected at the 5%
level, this value is contained in the 95% condence interval for β .
1
1

d) The 99% condence interval for β is [520.4 ± 2.58 × 20.4]= [467.7, 573.0].
0

e) R is unit-free. R = 0.08 means that 8% of variation in T estScore is

2 2

explained by (or predicted by) class size (CS). Alternatively, 8% of the variation
in our data is explained by our model.
18. CLT: Y ∼ N µ, = N 100, 

2
 
σY 43
Y N N

a) P (Y < 101) = P √ < √ = P (Z < 1.52) = 0.9357.

Ȳ −100
43
100
101−100
43
100

b) P (101 < Y < 103) = P

 
101−100 −100
√ 43
< Ȳ√ 43
< 103−100
√ 43
= P (1.22 < Z < 3.66)
64 64 64

= P (Z < 3.66) − P (Z < 1.22) = 1 − 0.8888 = 0.1112 .

c) P (Y > 98) = P
 
Ȳ√−100
> 98−100
√ 43 = P (Z > −3.91) = 1 − P (Z < −3.91) =
1 − 0 = 1.
43
165 165

19.
a) The condence interval for the mean test score in the population is the
following:
19.5
654.2 ± 1.96 √ = 654.2 ± 1.8649 = [652.34, 656.07].
420
b) Assuming unequal population variances, the null hypothesis is that H :
vs. H : µ .
0
6= µ
Therefore, t = = 4.05, which is statistically signicant at conven-
µsmallclass =µ largeclass 1 smallclass largeclass
q657.4−650.0

tional levels whether you use a two-sided or one-sided alternative. Hence the
19.42 2
238 + 17.9
182

null hypothesis of equal average scores is rejected and so there is statistically

signicant evidence that the districts with smaller classes have dierent average
test scores than districts with larger classes.

9
20. True regression model: W age = β + β Experience + β Experience + ε
i 0 1 i 2
2
i i

Regression model used: W age = β + β Experience + ε

i
∗
0
∗
1 i
∗
i

We know that: βc → β + β ∗
1 1 2 as N goes to innity.
Cov(exper,exper 2 )
V ar(exper)

⇒the bias = β 2 < 0, because:

Cov(exper,exper 2 )
V ar(exper)

• the correlation between exper and exper is positive, but, 2

• the exper enters the error term ε with negative sign, since the wage
2 ∗

function is concave in exper, as mentioned in the question (the graph of

i

wage and experience has an inverted-U shape in the data). So the sign on
β is negative.
2

The negative bias means that βc is smaller than it would be if exper was
∗ 2

included in the regression. 1

21. There is no problem for estimation, since the second explanatory variable
is not linearly related to the rst. This is an example of a quadratic regression
model, which is frequently estimated in econometrics.
22. There is perfect multicollinearity present since one of the three explanatory
variables can be expressed linearly in terms of the other two. For example:
X = X − X . Hence there are not really three pieces of dierent information
contained in the three explanatory variables. Dropping one of the three variables
2 1 3

will solve the problem.

23.
a) Cov(X, ε) 6= 0 because regions where relatively more dams are built are also
likely to dier along other dimensions, such as potential agricultural productiv-
ity.
b) The "slope of land" does not seem to be a valid instrument, because the
elevation and gradient of a district are likely to inuence the agricultural pro-
ductivity directly:
Cov(Z, ε) 6= 0 since "slope of land" is omitted in Y = βX + ε i i i

⇒ Cov(Z, Y ) = Cov(Z, βX + ε) = βCov(Z, X) + Cov(Z, ε) 6= βCov(Z, X)

⇒ Z inuences Y above and beyond Cov(Z, X).

Thus, Z is not a valid instrument for X .

10
24.
a) The entity xed eects model SP RICE = β + β REV EN U E + f +
u can be used to take care of this problem.
it 0 1 it i
it

We need to include dummy variables for each rm and (if we include an in-
tercept, we drop one dummy), in order to control for characteristics that are
constant across entities, i.e. industry and location do not vary by rm. Includ-
ing dummy variables for each rm (i.e. Let D = 1 if rm i= 2, 0 otherwise)
controls for these types of characteristics.
2i

b) We need to create dummy indicators for the companies and the months:
SP RICE = β + β REV EN U E + γ D + γ D + γ D + δ BT + ... +
δ BT + u .
it 0 1 it 2 2i 3 3i 4 4i 2 2t
48 48t it

Notice that we excluded a dummy for one of the companies and also for one of
the rms so as to avoid perfect multicollinearity. D has the same denition
as in part a, but BT = 1 if t = j, 0 otherwise. Note, t is measured in months
ri

so t = 24 is the 2nd year of data.You would then run the regression using OLS
jt

including these dummy variables.

11