EEP 118 / IAS 118 Elisabeth Sadoulet and Daley Kutzman

University of California at Berkeley Fall 2013

Introductory Applied Econometrics

Midterm examination

Scores add up to 50 (5 points for each sub-question)

Your name:________________________________________ SID:__________________________

1. (5 points) Using data on birth weights, we estimated the following two models:
 = 119.8 − 0.52cigs
bwght
R2=0.023 n=1388
(0.6) (0.09)
 = 116.9 − 0.46cigs + 0.09 faminc
bwght
R2=0.03 n=1388
(1.0) (0.09) (0.03)

where bwght is birth weight (in ounces), cigs is the number of cigarettes smoked daily during pregnancy and faminc
is 1988 family income, in $1000.
How does the introduction of the variable faminc affect the estimated parameter on cigs? What can you infer about
the correlation between faminc and cigs? Justify your response.
2. (5 points) Below are summary statistics for the GPAs of a sample of 101 students from Michigan State
University. The dean of MSU, Lou Anna Simon, firmly believes that the true average GPA of her university is 3.1
and your sample below is an inaccurate representation. Should you be skeptical of Lou Anna’s claim? Support your
opinion with a hypothesis test at the 5% significance level.

Variable | Obs Mean Std. Dev. Min Max

-------------+--------------------------------------------------------
colGPA | 101 2.984 .3723103 2.2 4

3. (10 points) Using a small sample of households from Nicaragua, we estimate the relationship between the log of
energy expenditure (lenergyexp) and the log of household total expenditure per capita (ltotexppc), household
size (hhsize), and whether the household owns a stove (stove).
. regress lenergyexp ltotexppc lhhsize stove

Source | SS df MS Number of obs = 174

-------------+------------------------------ F( 3, 170) = 30.10
Model | 82.5864529 3 27.5288176 Prob > F = 0.0000
Residual | 155.493187 170 .914665804 R-squared =
-------------+------------------------------ Adj R-squared = 0.3354
Total | 238.07964 173 1.37618289 Root MSE = .95638

------------------------------------------------------------------------------
lenergyexp | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
ltotexppc | .7763713 .1014082 7.66 0.000 .5761897 .9765528
lhhsize | .4904821 .1654038 2.97 0.003 .1639723 .816992
stove | .358545 .2240303
_cons | -.7896446 1.091029 -0.72 0.470 -2.943354 1.364064
------------------------------------------------------------------------------

a. What is the p-value for the test that having a stove has no effect on the consumption of energy? Interpret your
results.

b. Calculate and interpret the R-squared for this estimation.

4. (15 points) You have n = 40 quarterly observations on the imports M of a country, an index of import prices PM ,
and real aggregate income GDP. Adding dummy variables Q2,Q3 , and Q4 for the 2nd, 3rd, and 4th quarters of the
year, you estimate the model:
log M = β0 + β1 log PM + β 2 logGDP + β3Q2 + β 4Q3 + β5Q4 + u
and find the following results:

log M = 4.30 − 0.58log PM + 1.45logGDP + .15Q2 + .10Q3+ .40Q4
R 2 = 0.253 , n = 40
(0.13) (0.21) (.10) (.05) (.12)
(a) Construct a 95% confidence interval for β1 . Interpret.

(b) Test the hypothesis β2 = 1 against β2 ≠ 1 at the 5% significance level. Interpret this result in economic terms.

(c) Why is a first quarter Q1 dummy variable not included in the model? Interpret the estimated parameters on
Q2,Q3 , and Q4 .
5. (15 points) Using data for the US gasoline market between 1960 and 1999, we estimated the following model:

. regress lng lninc lnpriceg lnprnewcar lnprusedcar

Source | SS df MS Number of obs = 40
-------------+------------------------------ F( 4, 35) = 440.79
Model | 2.14671037 4 .30667291 Prob > F = 0.0000
Residual | .019480675 35 .000695738 R-squared = 0.9910
-------------+------------------------------ Adj R-squared = 0.9888
Total | 2.16619104 39 .061891173 Root MSE = .02638

------------------------------------------------------------------------------
lng | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lninc | 1.692181 .2154954 7.85 0.000 1.250759 2.133604
lnpriceg | -.2325466 .0595617 -3.90 0.001 -.3545532 -.1105399
lnprnewcar | -.233414 .1795357 -1.30 0.204 -.6011761 .1343482
lnprusedcar | -.0597329 .0693655 -0.86 0.396 -.2018216 .0823558
_cons | -3.957868 2.581538 -1.53 0.136 -9.24591 1.330174
------------------------------------------------------------------------------
where: lng=log(total US gasoline consumption per capita), lninc = log(income per capita), lnpriceg =
log(gasoline price), lnprnewcar = log(price of new cars), and lnprusedcar = log(price of old cars)

a. Suppose the government imposes a tax on gasoline that induces a price increase of 15%. What would be the effect
on gasoline consumption?

b. Is the consumption of gasoline influenced by the price of new cars and the price of used cars, when you consider
them one at a time? Justify your response.

c. We now estimate the model without the prices of new and used cars (See table on the last page). Comparing the
two estimated models, would you say that the consumption of gasoline is influenced by the prices of cars, new or
used, considered together? (Do a joint test of significance on the two parameters).
 Formulae
Statistics
1
Covariance between two variables in a population: cov ( x, y ) = ∑ ( xi − x ) ( yi − y )
n i
cov ( a1 x + b1 , a2 y + b2 ) = a1a2 cov ( x, y )

( )
var ax + by = a 2 var x + b2 var y + 2abcov(x, y)

Variance for the difference in means of two independent samples:

( )
var x1 − x2 = var x1 + var x2 ( ) ( )
When y is a binary variable with probability prob(y = 1) = p, its variance is p (1– p)

OLS estimator

cov ( x, y )
β̂1 =
var x
with var β̂1 = ( )
σ2
SSTx

For multiple regression: var β̂ j = ( ) σ2

(
SST j 1 − R 2j )
n n n
SST = ∑ ( yi − y ) , SSE = ∑ ( ŷi − y ) , and SSR = ∑ ûi2
2 2

i=1 i=1 i=1

Test statistics:

F statistic for q restrictions in a regression done with n observations and k exogenous variables:
2
RUR (
− RR2 q )
~ F ( q, n − k −1)
(
2
1− RUR )
( n − k −1)

Table for question 5c

. regress lng lninc lnpriceg
Source | SS df MS Number of obs = 40
-------------+------------------------------ F( 2, 37) = 629.99
Model | 2.13986669 2 .534966673 Prob > F = 0.0000
Residual | .026324354 37 .000849173 R-squared = 0.9878
-------------+------------------------------ Adj R-squared = 0.9863
Total | 2.16619104 39 .061891173 Root MSE = .02914

------------------------------------------------------------------------------
lng | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lninc | 2.13019 .148784 14.32 0.000 1.826743 2.433637
lnpriceg | -.1528558 .0535811 -2.85 0.008 -.2621352 -.0435763
_cons | -5.605895 2.165544 -2.59 0.015 -10.02255 -1.18924
------------------------------------------------------------------------------