You are on page 1of 13

Home assignment 1

Applied Econometrics
Name: Luu Nguyen Phu An
Student ID: 31151021903

C3:
(i)

N= 706
^
Sleep=3568.4+0.15 totwrk
R2= 0.103
The intercept implies that the estimate amount of time for sleeping for people, who do not work
is about 3586.4 minutes, or about 59.77 hours.
(ii) If

totwrk=2 hours ^
sleep=0.15 totwrk , then

^
sleep=0.15 120=18 mins
This is not a large effect because the total amount of time for sleeping in one week only
decreases 18 minutes, or just about 2.5 minutes per night.
STATA COMMAND:
. use "D:\Ses3\stat\DATA\SLEEP75.DTA", clear
. reg sleep totwrk

C4:
(i)

wage=957.95

.The average wage is 957.95 US dollars per month.

I Q = 101.23 .The average IQ is 101.28 in the sample. The sample standard deviation of IQ is
15.05, which is pretty close to the population standard deviation value of 15.
(ii)

wage=116.99+8.3 IQ ; n=935; R=0.096


The estimate simple regression model: ^
The predicted increase in wage for an increase in IQ of 15 points:
w
^
age=8.3 IQ=8.3 15=124.5 US dollars

No, IQ does not explain most of the variation of wage. It only explains only 9.6%.
(iii)

A
log-

wage=5.87+0.0088 IQ
level model: log ^

n=935, R2=0.099
If IQ=15,then log ^
wage=0.0088 15=0.132 , which is the approximate proportionate
change in predicted wage. The percentage increase is therefore approximately 13.2.
STATA COMMAND:
. summarize wage IQ
. reg wage IQ
. reg lwage IQ

C7: (i)

gift=7.44
^

P ( gift < 1 )=0.6000469 (The number of people who did not give any gift is
2561)

(ii)

^
mailsperyear = 2.049
(iii)

Min mailsperyear=0.25 Max mailsperyear:3.5

The estimate model:

^
gift=2.014 +2.649 mailsperyear

R2=0.014 n=4268
(iv)
The slope coefficient, which was calculated in (iii), means that with each change by decreasing
or increasing one mail per year, there would be a corresponding change in estimate gift by 2.649.
If each mailing costs one guilder, the charity is expected to make an estimate net gain on each
mailing by (2.649-1) guilder.
No, it does not mean the charity makes a net gain on every mailing. Because, the gain in only on
average. There could be some mailings, whose contribution is nothing or even less than mailing
cost. On the other hand, some other mailings could generate much more than the mailing cost.
(v)
The smallest mailsyear in the sample is 0.25, the smallest predicted value of gifts is
2.014+2.649=2.676. However, when we look at the overall population or reality, where some
people receive no mailings, the smallest predicted value is about two. Thus, with this estimated
equation, we never predict zero charitable gifts.

STATA COMMAND
-

Summarize gift
Count if gift<1
Proportion gift
Summarize mailsperyear
Reg gift mailsperyear

CHAPTER 3
C4:
(i)

Table:
atndrte
priGPA
ACT

Minimum
6.25
0.857
13

Maximum
100
3.93
32

Average
81.71
2.587
22.510

(ii)
^
atndrte=75.7+17.261 priGPA1.717 ACT
n= 680; R2=0.291
The intercept means that if a student, whose prior GPA score is zero and ACT score is zero, still
has predicted attendance rate is 75.7%. This intercept has no meaning because in fact, there is no
college student whose prior GPA score is zero and ACT score is zero.
(iii)
The coefficient on priGPA means that, if students have one more point higher in prior GPA score,
the predicted attendance rate of them is 17.261% higher. The ACT is fixed in this situation.
However, there is a bit surprise in the coefficient on ACT score, when it is negative. It means,
students with 5 points higher in ACT score, their predicted attendance rate is 8.59 lower at any
given level of priGPA.
As priGPA is a measures the performance of students in college (and at least partially, it could
reflects past attendance). On the other hand, ACT score measures the potential of students in
college, it appears that students who has innate ability, think they can get by with skipping
lectures.

(iv)

^
If priGPA=3.65 and ACT=20, then atndrte=104.38 . It is nonsense because the attendance
rate could not be greater than 100%.
In reality, we can predict the attendance rate of 100% for this student.
Yes, there is one student who has priGPA=3.65and ACT=20
(v)
The predicted difference in attendance rates between Student A and Student B is
17.261 ( 3.12.1 )1.717 ( 2126 ) =25.846
STATA COMMAND
. summarize atndrte
. summarize priGPA
. summarize ACT
. reg atndrte priGPA ACT
. sort priGPA
C7:
(i)

^
math 10=20.361+6.23 lexpend0.304 lnchprg
n= 408, R2=0.179
Yes, the signs of coefficients are what we expected. The percentage of students passing the math
exam increases with the increase on the expenditure per student. On the other hand, the
percentage of students passing the math exam decreases with the increase on the percentage of
students who receives subsidized lunch program.
(ii)
The intercept does not have much of interpretation. Due to the fact that it makes a little sense that
lnchprg is zero (maybe there is no student attending the school lunch program). However, it does
not make any sense if the lexpend is zero (it means that the expenditure per student is 1 dollar
because log(1)=0). In addition, the minimum of expend in data is 3,332 UD dollars.

(iii)

^
math 10=69.341+11.163 lexpend
The slope coefficient, which we have recently obtained, is greater than the estimate obtained in
part (i).
The estimated spending effect now is larger than in part (i).
^
(iv) lexpend=8.4280.002lnchprg
lexpend and lnchprg are negatively correlated. Actually, at the beginning, we expected that the
predicted expenditure per student would increase when the percentage of students who joined in
subsidized lunch program grows, all else are equal. However, in fact, we decrease the percentage
of students who participate in school lunch program, the expenditure per student lightly because
the school will spend its money on many other things.
(v)
Because log(expend) and lnchprg are negatively correlated, the effect of log(expend) on math10
will be big when we excluded lnchprg explanatory variable, On the other hand, if we include
lnchprg explanatory variable, the effect of lexpend would be smaller on math10. We can see that
the inclusion of lnchprg suppressed the coefficient on log(expend).
STATA COMMAND:
-

C8:

reg math10 lexpend lnchprg


reg math10 lexpend
reg lexpend lnchprg

(i)
prpblck
income

Average value
0.113
47053.78

Standard deviations
0.182
13179.29

Units of measurement
percent
dollars

(ii)

^
psoda=0.956+ 0.12 prpblck +0.000 income

n=401 ; R 2=0.0642

The coefficient on prpblck is 0.12.


The coefficient on prpblck means that when prpblck increases by 1, then the predicted price
of a soda will increase 12 cents.
I do not think this is economically large.
For instance, if prpblck increases by 10%, then psoda increases by 0.12 x0.1 = 0.01 or 1 cent.
This amount of money is very little. However, if the number of soda sold is enormous, then
this change could create a large amount of money.
In conclusion, on face, this change is not economically large.
(iii)

^
psoda=1.037+ 0.065 prpblck

n=401 ;R =0.018

The discrimination effect is estimated to be significantly smaller when we exclude income


from the regression.
(iv)

^
lpsoda=0.794+0.121 prpblck +0.077 lincome
n=401; R2=0.0681
prpblck is already a proportion, meaning that we can interpret the coefficient on prpblck in
terms of percentage changes. If prpblck increases by 0.2, then the price of soda is predicted
to increase by 0.12 x 0.12= 2.4 percent.
(v)

^
psoda=0.512+0.075 prpblck +0.142 lincome+ 0.396 prppov
n=401 ; R 2=0.085
The coefficient on prpblck now decreases from 0.12 to 0.075 when prppov (the proportion of
individuals living the poverty in a zip code) is included in the regression.
(vi)

Yes, it is roughly what I expected. I expected a negative correlation.


(vii)
I think this statement is nonsense because they can certainly be in the same regression.
Maybe, they will make it more difficult to identify the independent effects of each on the
price of soda. However, it does not mean that they have no business in the same regression.
STATA COMMAND
-

summarize prpblck
summarize income
reg prpblck income
reg psoda prpblck
reg lpsoda prpblck lincome

- reg lpsoda prpblck prppov


- reg lincome prppov

You might also like