Foreign Trade University
Faculty of Finance and Banking
*********
ECONOMETRICS ASSIGNMENT
Topic: “The impact of lover on study results of Foreign
Trade University students”
Contents
I. Introduction................................................................................3
II. Data description.........................................................................3
1.Scope.........................................................................................................3
2.Sources of data..........................................................................................3
3.Investigated factors and expectations.......................................................3
III. Empirical Results.......................................................................5
1.Building Regression Model.........................................................................5
a.Model 1...............................................................................................5
b.Model 2...............................................................................................6
2.Assumption Tests.......................................................................................8
a.Multicollinearity..................................................................................8
b.Heteroskedasticity..............................................................................9
c.Autocorrelation.................................................................................10
d.Normality..........................................................................................11
IV. Conclusion...............................................................................13
1.Interpretion..............................................................................................13
2.Suggestions.............................................................................................14
3.Limitations...............................................................................................14
4.Final words...............................................................................................15
2
Part I: Introduction
Love is inherently a part of human life, particularly with the young. Entering college
marks the stage of adult. From then on, you have the right to have a boy/girl friend
and you also have more opportunities to expand your relationship than high school
period. Percentage of college students who are in love is great. Besides issues such
as part time jobs, school work, or social activities, we cannot deny that love is an
important part of university student life. However, the last question is whether
university students should love or not. Love has positive or negative impacts on
academic performance. Indeed, it depends on the way you love and people you
choose. Therefore, our group decided to choose the topic “The impact of lover on
study results of Foreign Trade University students”. We hope we can bring a more
fully comprehensive view of students’ love, which suggests a reasonable and helpful
advice for you to balance between love and learning.
Part II: Data description
1. Scope:
Data collected from students in Foreign Trade University who already have a lover
by November 2012.
2. Sources of data:
We have conducted survey on totally 150 students, through both online-form and
offline-form, to run the model.
Of all the 150 answer sheets, we have had 111 acceptable results. The rest cannot
be used because the respondents omitted some questions, or had some unrealistic
answers
3. Investigated factors and expectations:
Dependent variable
Variable
Description
GPA
The GPA at the nearest semester of a student
3
Independent variable – Quantitative variable
Expec
Variable
Description
-
Note
tation
The gap between the
Age
Finance
The “Age” variable can have
age of the respondent
+/-
positive or negative impact
and her/his lover.
on the study results.
The finance condition of
The “Finance” variable can
the respondent’s
+/-
boy/girl friend
have positive or negative
impact on the study results.
The “Time” variable can
have negative impact on the
The average number of
Time
study results. If students
hours per week the
-
respondent spends with
spend more time on love,
they have to cut back the
his/her lover
time spending on studying
Qualitative variable
Variable
Description
Denote
0
Expec1
tation
Note
The “Gender”
variable can be
Gender
Gender of the
respondent
Femal
Male
e
+/-
positive/ or
negative impact on
the study results.
Appearance
The
Averag
above
appearance of
e
Above
+/-
The “Appearance”
variable can be
the
positive/ or
respondent’
negative impact on
boy/girl friend
the study results.
is above
4
average level
The
The “Appearance”
appearance of
Appearance
below
the
respondent’s
boy/girl friend
variable can be
Averag
Below
e
+/-
positive/ or
negative impact on
the study results.
is below
average level
The capacity
The higher
of study of the
Capacity
respondent’s
Averag
above
boy/girl friend
e
capacity of boy/girl
Above
+
level of motivation
is above the
for respondent
average level
The capacity
The lower capacity
of study of the
Capacity
respondent’s
Averag
below
boy/girl friend
e
of boy/girl friend,
Below
-
the lower level of
motivation for
is below the
respondent
average level
Distance
friend, the higher
The geography
The “Distance”
distance
variable can have
between the
respondent’s
Close
Far
+/-
positive or
negative impact on
the study results
place and
his/her lover’s
Extra activities
High
concentrate
Extra activities
Taking part in extra
implies part
activities may have
time jobs or
no
yes
+/-
positive or
other social
negative impact on
activities
the study result
The higher
Respondent
concentrate
no
yes
well on study.
+
concentration on
study, the higher
the study result
5
The worse
Respondents
Low
concentrate
concentrate
no
yes
-
concentration on
badly on study.
study, the lower
the study result
Part III: Empirical Results
1. Building regression model:
a. Model 1:
Y = 0 + 1*Gender + 2*Age + 3*Appearance_above +
4*Appearance_below + 5*Capacity_above + 6*Capacity_below +
7*Distance + 8*Finance + 9*Dedicated_time +
10*Low_concentrate + 11*High_concentrate + 12*Extra_activities
+ ui
By using Gretl following OLS method, we have the result below:
6
R-squared = 0,511498, which means all the independent variables explain
about 51,15% of the real outcome.
There are only 4 variables which have statistical significance (p-values ≤ 0.1).
These
are:
Age,
Dedicated_time,
Low_concentrate,
High_concentrate,
Extra_Activities. The signs of the coefficients of these variable are followed
our expectation. Other variables do not have statistical significance, so we
will omit them from the model and run another regression model.
b. Model 2:
After omitting insignificant variables, we run the model with the other 5 variables:
Y = 0 + 1*Age + 2*Dedicated_time + 3*Low_concentrate +
4*High_concentrate + 5*Extra_activities + ui
7
The R-squared now is 0,475943, which is smaller than the old R-squared. Thus, we
run the Ramsey RESET test to see if there is mispecification in our model or not.
As we can see, all the p-values are larger than
specification is adequate.
8
α
= 0,05. So we conclude that
The signs of the variables follow our expectations: The variables of age, dedicated
time and low concentration have negative signs, means that they have negative
relationship with the increase of GPA. On the other hand, the variables of high
concentration and extra activities have positive signs, indicating that they have
positive relationship with the increase of GPA.
Our SRF now is:
Y = 7,76941 -0,0326185*Age – 0,0129144*Dedicated_time –
0,556837*Low_concentrate + 0,655956*High_concentrate +
0,3368*Extra_activities + ui
2. Assumption Tests
a. Multicollinearity:
At first, we use a correlation matrix to detect the presence of multicollinearity.
By using gretl, we have the following correlation matrix:
We can see from the matrix that all the correlation coefficients between the
variables have small absolute value. Thus, we can not conclude that variables are
9
strongly associated with each other. But we also can not conclude that
multicollinearity does not appear. That is why we have to run another test to see if
there is multicollinearity in the model.
According to this test, all the variables have Variance Inflation Factors (VIF) smaller
than 10. Therefore, we can conclude that the model does not have the problem of
multicollinearity.
b. Heteroskedasticity:
We try to see if there are any signs of heteroskedasticity by creating a scatter plot
of the model:
10
There is no sign of heteroskedasticity, so we move on to run the White Test. And
here is the result from gretl:
The p-value here is 0,505189 >
α = 0,05; so we can come to the conclusion that
there is no heteroskedasticity in this model.
11
c. Autocorrelation:
Because we use a cross-sectional data, we cannot just run the autocorrelation test.
In this case, we change our data to time-series and have the result of the BG test as
below:
In this case, all the p-values are larger than
α
= 0,05; so we can come to the
conclusion that the model does not face autocorrelation.
d. Normality
This part is to find out whether the error term ui in the model has normal
distribution. After running the Test statistic for Normality, here is our result:
12
The p-value in this case is 0,0153 <
α = 0,05; so we conclude that the error term is
not normally distributed.
To fix this, we add the variable of l_age to the current model:
13
We run the test again to see if the problem has been cured:
14
The p-value is 0,6013 >
α
= 0,05; so we can conclude that the error term is now
normally distributed.
Part III: Conclusion
1. Interpretation:
After running regression and testing all the assumptions for multiple regressions, we
have the final regression function with R-squared = 41, 3070%. This is not a high
value, but still can be accepted. It implies that our model could explain about 41,3%
of the outcome.
Our final regression model is:
15
Y = 7,85175 – 0,0154247*Dedicated_time – 0,481006*Low_concentrate +
0,517323*High_concentrate
+
0,367198*Extra_activities
–
0,162235*l_age + ui
From our final regression model, we can conclude that among 5 variables:
Dedicated–time, low–concentrate, high-concentrate, extra-activities and l_age, there
are 3 variables having negative impacts on GPA and 2 variables having positive
ones. Their influence mostly follow our first expectations.
1 = -0, 0154247 < 0: means that one hour increased in time for love leads
to 0,0154247 unit less in Y if other factors remain unchanged.
2 = -0, 481006 < 0: means that low concentration leads to 0,481006 unit
decreased in Y if other factors remain unchanged.
3= 0,517323 >0: means that high concentration leads to 0,517323 unit
more in Y if other factors remain unchanged.
β4 = 0,367198 > 0: means that taking part in extra-activities leads to
0,367198 unit increase in GPA.
β5 = –0,162235 < 0: means that a year increased in age gap leads to
-0,162235 unit decrease in GPA.
2. Suggestions:
Below are some suggestions for Foreign Trade University’s students that we
conclude from our analysis result:
Gender, the capacity/appearance/finance condition of the lover and distance
do not have affect on study result. Thus, we are free to love who we want
redardless of these factors.
As students, we should spend less time for love as studying is the most
important thing at this time. It seems like an opportunity cost if we dedicate
too much to dating. Love should be the motivation to archive higher marks, not
a reason for going backward.
16
Practicing
self-concentration
is
the
most
important
factor
on
study
performance. We should try to manage time effectively as well as identify our
goals clearly; by doing this, not only studying but other issues will get better.
Extra-activities are very essential, especially with students, it not only help us
develop social skills but also have positive effects on studying result. However,
we should balance between time for studying and time for these activities.
3. Limitations:
From our regression model, we can conclude that among 13 variables, there are
some variables that follow our expectation but some do not. This indicates the gap
between theory and reality, which can be unpredictable and impossible to fulfill
without the help of subjects like Econometrics.
Besides, during the process of preparing this report, we have to face some
problems. The most challenged problem arising from the subject econometric itself.
Econometrics is a difficult subject which requires good nationality, diligence and
time for research as well as analysis. However, because we had to complete
assignments
of
different
subjects
simultaneously
and
we
received
the
announcement in hurry, we had to worked in a rush and did not have enough time
to proofread this paper. Furthermore, our knowledge of the subject still has
limitations, which led us to choosing unsuitable variables. In other words, the model
which was run by us still had unavoidable mistakes. Besides, our topic is about
“Impacts of lovers on study results of FTU's students” – a sensitive one, so very few
people can provide exact data leading to limited number of observations: we
received only 150 surveys, and luckily 110 ones are accepted. Moreover, many
different groups are conducting surveys at this time, which makes students get
bored of filling surveys. Last but not least, it is also not easy to draw the right and
meaningful conclusion from the result of the research, which is an inevitably
important step in the process of research. However during the time of doing this
exercise we had chance to practicing team-working and understanding more about
the econometrics and its application in life.
4. Final words
17
Our group, thanks to the instructions of Dr Tu Thuy Anh and lecturer Thai Long,
has made great effort in collecting data and implementing the model. Though the
result did not turn out to be as well as we had expected, we have gained a lot
experiences in building the regression model.
18