Introductory Econometrics - Exam

Professor: Michel SIMIONI

October . . . , 2018

1 Theoretical questions
1.1 Errors-in-variables

1. Consider the one-variable linear regression model Yi = β0 + β1 Xi + ui . What are the

three least squares assumptions used to derive the properties of ordinary least squares
estimator of β1 ?

2. Suppose that Yi is measured with error, so the data are Yei = Yi + wi , where wi is
the measurement error, which is i.i.d. and independent of Yi and Xi , and has a nite
fourth moment.
Consider the population regression Yei = β0 + β1 Xi + vi , where vi is the regression error,
using the mismeasured dependent variable, Y
ei . Show that vi = ui + wi .

3. Show that the regression Yei = β0 +β1 Xi +vi satises the three least squares assumptions.

4. Is the ordinary least squares estimator of β1 consistent?

5. Can condence intervals be constructed in the usual way?

6. Evaluate these statements: Measurement error in the X 's is a serious problem. Mea-
surement error in Y is not. Justify and argue your answer.

1.2 Heteroskedasticity

Consider the savings function:

√
Savings = β0 + β1 Income + u, with u= Income × e

where e is a random variable with E(e) = 0 and V ar(e) = σe2 . Assume that e is independent
of Income.

1. Show that E(u|income) = 0, so that the key zero conditional mean assumption
(LSA#1) is satised.

1
 2. Show that V ar(u|income) = σe2 × income so that the homoscedasticity assumption is
violated. In particular the variance of savings increases with income.

3. Provide a discussion that supports the assumption that the variance of savings increases
with family income.

4. Evaluate this statement: Having heteroskedastic errors is not really a serious problem.

2 Exercise
This exercise refers to Tables 1 and 2 (see below) where are presented the results of regression
estimations using data for 1998 from Current Population Survey. The data set consists of
information on 4000 full-time full-year worker. The highest educational achievement for each
worker was either a high school diploma or a bachelor's degree. The worker age ranges from
25 to 34 years. The data set also contains information on the region of the country where
the worker lived, marital status, and number of children. For the purposes of the exercise
let

ˆ AHE = Average Hour Earnings (in dollars 1998)

ˆ College = Binary variable (1 if college, 0 if high school)

ˆ F emale = Binary variable (1 if female, 0 if male)

ˆ Age = Age (in years)

ˆ N ortheast = Binary variable (1 if Region = Northeast, 0 otherwise)

ˆ M idwest = Binary variable (1 if Region = Midwest, 0 otherwise)

ˆ South = Binary variable (1 if Region = South, 0 otherwise)

ˆ W est = Binary variable (1 if Region = West, 0 otherwise)

Questions :

2
1. Compute R for each of the regressions and compare their values.

2. Using the regression results in column (1) of Table 1:

(a) Do workers with college degree earn more, on average, than workers with only
high school degrees? How much more?

(b) Do men earn more than women on average? How much more?

3. Using the regression results in column (2) of Table 1:

(a) Is age an important determinant of earnings? Explain.

2
 (b) Sally is a 29-year-old college graduate. Betsy is a 34-year-old college graduate.
Predict Sally's and Betsy's earnings.

4. Using the regression results in column (3) of Table 1:

(a) Do there appear to be important regional dierences

(b) Why is the regressor W est omitted from the regression? What should happen if
it is included?

∗∗∗ ∗∗ ∗
5. Add " " (1%), " " (5%), or " " (10%) to Table 2 to indicate the statistical sig-
nicance of the coecients. Explain why you can use the following quantiles of the
centered and standardized normal distribution: 1.64 (95%), 1.96 (97.5%), and 2.57
(99.5%).

6. Using the regression results in column (1) of Table 2:

(a) Is the college-high school earnings dierence estimated from this regression sta-
tistically signicant at the 5% level? Construct a 95% condence interval for the
dierence.

(b) Is the male-female earnings dierence estimated from this regression statistically
signicant at the 5% level? Construct a 95% condence interval for the dierence.
7. Using the regression results in column (2) of Table 2:

(a) Is age an important determinant of earnings? Use an appropriate statistical test

and/or condence interval to explain your answer.

(b) Construct a 95% condence interval for the expected dierence between the Sally's
and Betsy's earnings.

8. Using the regression results in column (3) of Table 2: Do there appear to be important
regional dierences? Use an appropriate test to explain your answer.

9. The regression, whose results are presented in column (2) of Table 2, was estimated
again, this time using data from 1992 (4000 observations selected at random from the
March 1993 CPS, converted in 1998 dollars using the consumer price index).The results
are

AHE
\ = 0.77+ 5.29 College− 2.59 F emale+ 0.40 Age,
(0.98) (0.20) (0.18) (0.03)
2
with SER = 5.85 and R = 0.21.
(a) Why and how have been 1992 earnings converted in 1998 dollars?

(b) Comparing this regression to the regression for 1998 shown in column (2) of Table
2 was there a statistically signicant change in the coecient on College?

3
10. Evaluate the following statement: "In all the regressions, the coecient on F emale is
negative, large, and statistically signicant. This provides strong statistical evidence
of gender discrimination in the U.S. labor market."

4
Table 1: Results of Regression of Average Hour Earnings on Gender and Education Binary
Variables and Other Characteristics using 1998 Data from the Current Population Survey

Dependent variable: Average Hour Earnings (AHE )

Regressor (1) (2) (3)
College 5.46 5.48 5.40
F emale -2.64 -2.62 -2.62
Age 0.29 0.29
N ortheast 0.69
M idwest 0.60
South -0.27
Intercept 12.69 4.40 3.75
Summary statistics
SER 6.27 6.22 6.21
R2 0.176 0.190 0.194
Number of observations 4000 4000 4000

Table 2: Results of Regression of Average Hour Earnings on Gender and Education Binary
Variables and Other Characteristics using 1998 Data from the Current Population Survey

Dependent variable: Average Hour Earnings (AHE)

Regressor (1) (2) (3)
College 5.46 5.48 5.40
(0.21) (0.21) (0.21)
F emale -2.64 -2.62 -2.62
(0.20) (0.20) (0.20)
Age 0.29 0.29
(0.04) (0.04)
N ortheast 0.69
(0.30)
M idwest 0.60
(0.28)
South -0.27
(0.26)
Intercept 12.69 4.40 3.75
(0.14) (1.05) (1.06)
Summary statistics
F for regional eects = 0 6.10
(2.61)
SER 6.27 6.22 6.21
R2 0.176 0.190 0.194
Number of observations 4000 4000 4000
Notes: Numbers given between parentheses are standard errors
except for F -test for regional eects = 0 where it is the value
of the 95% quantile of F (3, 3993).