+ All Categories
Home > Documents > Georgi Stoilov Econometrics

Georgi Stoilov Econometrics

Date post: 06-Apr-2018
Category:
Upload: georgi-stoilov
View: 233 times
Download: 0 times
Share this document with a friend

of 17

Transcript
  • 8/3/2019 Georgi Stoilov Econometrics

    1/17

    Introduction to Econometrics

    BS2220

    Dr. Kai Sun

    Report

    Analysis of the impact of academic and non-

    academic activities on first year students

    performance

    Candidate Number: 150851

  • 8/3/2019 Georgi Stoilov Econometrics

    2/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    2

    Contents:

    1. Introduction:

    1.1. Research Question and Importance of the Study

    1.2. Existing Studies

    1.3. Report Structure

    2. Conceptual Framework:

    2.1. Theories -

    2.2. Literature review

    2.3. Relationship prediction -

    3. Data and Methodology:

    3.1. Source and Explanation of Data

    3.2. Data Analysis

    3.3. Correlations

    3.4. Regression Model

    3.5. Estimation Strategy

    4. Econometrics Results:

    4.1. Estimated Coefficients

    4.2. Goodness of Fit and Diagnostic Tests

    5. Conclusion:

    5.1. Implications

    5.2. Suggestions

    6. References

  • 8/3/2019 Georgi Stoilov Econometrics

    3/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    3

    Introduction

    Research Question and Importance of the Study

    The main aim of this study is to examine and analyze the main factors that reflect on

    first year students academic performance. The UK is renowned for its excellent

    higher education and attracts thousands of new students each year. Many of them

    originate from different cultures and backgrounds and are used to lead different

    lifestyles. This may include after-school engagements with academic and non-

    academic clubs and societies, part-time employment and socializing with friends. The

    core purpose of this study is to investigate how many this differences in students

    lifestyle and background influence on their academic performance. It is also crucial to

    identify which determinants are statistically significant and have negative or positive

    impact on students performance, so that both students and the University can put

    their efforts on them to increase students efficiency and overall performance. In

    order to achieve the goal of our study we have gathered data from a survey,

    conducted online by hundred and seventy students doing the BS2220 course at

    Aston University, comprising of seventy-nine questions in four areas. For this

    research I have primarily focused on three main categories as I consider them most

    highly correlated with the academic performance of first year students: (i) Studying

    and proximity to the universitys educational facilities (ii) Time spent on leisure

    activities and part time jobs; (iii) Using the universitys educational facilities.

    Existing Studies

    This topic is not new and prior studies have been conducted to identify the variables

    related to academic success of students in a variety of countries. With the increasing

    diversity of students attending university, there is a growing interest in the factors

    predicting academic performance. In a study conducted in Australian university,

    McKenzie & Schweitzer (2001) identified previous academic performance as the

    most significant predictor of academic performance. Other predictors of university

    grades, found by the study, were integration into university, self efficacy andemployment responsibility. In other more recent research Lebcir, Wells and Bond

    (2008) suggest that the factors: level of details given in lectures, speed of lectures,

    academic internet sources, English Language Skills, group or individual assessment,

    the qualitative/quantitative content of assessment are important drivers of the

    academic performance of international students.

    Report Structure

    The first section of this analytical paper introduces a conceptual framework and

    literature review of previous research in the area and finishes with a prediction for theoutcomes of the examined question. The second section presents the means of data

  • 8/3/2019 Georgi Stoilov Econometrics

    4/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    4

    gathering and its explanation followed by presentation of the research methodology.

    The Econometrics results section provides evaluation of the performance of the

    regression model and examines it for violations. At the end a conclusion summarises

    the main findings and gives suggestions for better future performance, based onthese findings.

    Conceptual Framework

    Theories

    In order to identify the relationship between students lifestyle and their academic

    performance it is necessary first to give definition to the concept of academic

    performance. Academic performance can be defined as the general level of grades

    a student acquires in exams and course-works throughout the academic year.

    Researchers have explored the study environment as a potential factor of academic

    performance outcome (Dorman, Fraser & McRobbie, 1997). It is also essential to

    understand the concept of lifestyle. According to Michman, Mazze, and Greco

    (2003) lifestyle emerges from various social influences, as well as an individuals

    personal value system and personality. Students spend their time influenced by their

    activities, interest and options available. After we defined these variables we can

    identify the key relationships between different student lifestyles and academic

    results.

    Literature review

    Throughout years a large amount of researches have been carried out in various

    universities worldwide in an attempt to most accurately identify the

    variables, determining students performance. Attendance in lectures and tutorials is

    considered to be essential variable in determining students academic results. This

    was proven in a research carried out by Walbeek, C. (2004) that students, attending

    all lectures and tutorials, are expected to achieve 7.3 percent more than their

    colleagues, who have not attended any lectures. Other supporting evidence is the

    study of E. Burd & B. Hodgson (2004) which identifies a strong correlation between

    lecture attendance and academic performance. I am also interested in therelationship between the usage of University facilities and services and academic

    results. In particular I am interested in the usage of mathematics tutorials, as Aston

    University provides valuable mathematics support, through its Learning Development

    Center, for students who are not very good with numbers and aims to improve their

    understanding and knowledge of the subject thus increase their academic

    performance. Studies among academics, including Mark Fenster (1998) and others

    have shown that there is a positive correlation between the amount of mathematics

    taken and exam result. Another component of students lifestyle is the time they

    spend watching television. Chernin & Limburger (2005) have found in their research

    a negative relationship between the amount of television watched and academic

  • 8/3/2019 Georgi Stoilov Econometrics

    5/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    5

    performance. Additionally, part-time work is also a segment of students lifestyle,

    especially for the students, who cover their living expenses on their own, without any

    financial support from their parents. The Cire Rochford, Michael Connolly &

    Jonathan Drennan (2009) research, conducted on students doing nursery, supportthe theory, that part-time employment has a negative correlation with academic

    overall result.

    Relationship Prediction

    Based on previous studies, the findings listed above and my personal preferences I

    will choose six variables to find out what determines academic performance based on

    our sample. In this paper I will study the relationship between academic performance

    and dwelling location, mathematics tutorials and study skills workshops taken, time

    spent watching TV, part-time employment during university and alcohol consumption.

    I predict that students, who live on campus, have higher attendance of study skills

    workshops and mathematics tutorials, spend less hours watching TV and doing part-

    time work and consume smaller units of alcohol are tend to have higher grades. In

    my research paper I am going to test these assumptions using correlations,

    regression analysis and other statistical methods.

    The Data and Methodology

    Source and Explanation of Data

    In order to complete this investigation, a survey has been conducted on hundred and

    seventy students from Aston University, Birmingham, taking the BS2220 module

    Introduction to Econometrics taught by Dr. Kai Sun. The questions in the survey

    were divided into four main groups; students personal information, lifestyle, family

    background and academic related information. The cross-sectional data collected

    from thirty eight questions in the survey represents students from diverse ethnicity

    and cultural backgrounds, which is something normal for Aston University and all

    other UK universities as whole. The sample size examined seems to be relatively

    small comparing to the scale of Aston University but I presume that the quality of the

    students in it is relatively high, as the Econometrics module is considered to be oneof the most complicated and demanding modules throughout the Business School. A

    lot of the variables in this dataset are highly correlated with students academic

    performance, for example, proximity of accommodation to university, prior academic

    achievements, after school engagements, such as part-time jobs, clubs, societies

    and sport activities, number of lectures and tutorials missed, time spend studying and

    revising prior and during exam period and time spend on leisure activities, such as

    TV watch, Facebook socializing and alcohol consumption. For the matter of my

    investigation I will analyse a set of six variables, which I think have a high impact on

    students performance and cover all aspects listed above.

  • 8/3/2019 Georgi Stoilov Econometrics

    6/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    6

    Data Analysis

    The outcomes of some questions in the survey conducted were initially described

    with characters in the dataset, or non-numeric values, but they have been convertedinto numeric functions. For example, using the destring and encode commands in

    STATA we can create dummy variables, which convert the character variables into

    numeric functions, so that all answers yes become 1 and no become 0. The

    data coding was completed in advance and the data was ready to be analysed. The

    summary of the dependant variable and some of the independent variables

    conducted in STATA showed the following results:

    Variable Description Observations Mean Standard Deviation Min Max Measurment

    First Year Average Score 79 64.71139 12.60914 2.2 84 Percentage

    Uni Entry Score 79 81.60253 29.1013 3.6 300 Percentage

    Lecture Hours Missed 79 16.08861 24.62862 0 145 Hours

    Use of Study Skills Workshop 79 .2025316 .4044543 0 1 Yes=1 No=0

    Prior Exam Period Library Study 79 18.26582 39.88178 0 300 Hours

    Exam Period Library Study 79 20.18987 22.81816 0 105 Hours

    Living on Campus 79 .2151899 .4135799 0 1 Yes=1 No=0

    Part of Non-Academic Society 79 .7341772 .4445932 0 1 Yes=1 No=0

    Part-Time Job 79 9.563291 26.34709 0 160 Hours

    Money Spent on Phone 79 25.15823 12.92943 5 65 Pounds

    Alcohol consumption 79 5.93038 11.92061 0 75 Units

    Facebook Usage 79 6.772152 11.99834 0 75 Hours

    Hours Watching TV 79 5.898734 6.826785 0 26 Hours

    As we can see from the table above, there are some extreme and unrealistic results

    observed from the output. The maximum Average University Entry Score result was

    300, which is inconsistent because the variable is measured in percentage and the

    maximum possible value is 100. This student maybe has entered his UCAS score by

    mistake instead of his average mark. Another student has answered that he has

    studied for 300 hours prior exam period in the library, which again is impossible

    because of the time constraint of 168 hours per week. In addition, a student has

    inputted a figure of 160 hours as a part-time job which again is almost impossible

  • 8/3/2019 Georgi Stoilov Econometrics

    7/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    7

    considering the fore-mentioned time constrain. These outliers could have resulted

    due to students misinterpretation on the survey questions.

    Correlations

    In the table below we can examine several correlations between the explained

    variable Average Score and the explanatory variables. The STATA summary shows

    a positive, although very weak relationship of 7% between students pre-university

    academic scores and their current performance at university, which indicates that

    students who have put continuous efforts in High School tend to do the same in

    university and perform better than those with low pre-university grades. Recent UK-

    based study by Duff (2004) confirms our theory that good prior academic

    achievement, as measured by results in high school, leads to better student

    performance at university. According to the findings below, living in campus ispositively associated to students academic performance with a positive correlation of

    7%. This can be explained with the fact that students who live closer to university

    have easier access to university educational facilities at any time, especially during

    examination period, when the library facilities are open twenty four hours. On the

    other hand the amount of time students spend on non-academic activities, such as

    undertaking part-time work or speaking on the phone to their family and friends show

    a negative correlation to their academic performance. These are expressed by

    negative coefficients of -0.0721 and -0.1062 consecutively, which means that, the

    more time students spend working or talking on the phone, the lower their academic

    performance will be. Unexpectedly a positive correlation of around 3% exists

    between hours of lectures missed and students grades. One possible reason for this

    may be the fact that students utilize the hours missed in lectures on individual study

    or maybe due to the small size of the sample the results may not be totally accurate.

    Furthermore, there is a positive relationship of 0.0624 between Study Skills

    Workshops and first year average results. This indicates that the workshops, offered

    by the university, have provided solid background knowledge to students for their

    exams.

  • 8/3/2019 Georgi Stoilov Econometrics

    8/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    8

    Variable

    Description

    Average

    Score

    Uni Entry

    Score

    Skipped

    Lectures

    Study Skills

    Workshop

    Library

    Study

    Live on

    Campus

    Part-Time

    Job

    Phone

    Bill

    Facebook

    Usage

    Average score 1.0000

    Uni Entry Score 0.0679 1.0000

    Skipped Lectures 0.0304 -0.0434 1.0000

    Study Workshop 0.0624 -0.0806 -0.1280 1.0000

    Library Study -0.0774 -0.1276 -0.0413 0.2069 1.0000

    Live on Campus 0.0366 -0.1743 -0.1190 0.0427 0.1668 1.0000

    Part-Time Job -0.0721 -0.1035 -0.0252 -0.1239 -0.1619 -0.0524 1.0000

    Phone Bill -0.1062 -0.0414 0.0290 -0.0111 0.1618 0.0007 -0.1111 1.0000

    Facebook Usage0.0040 0.0978 0.0622 0.0189 0.3028 0.0007 -0.1136 0.0867 1.0000

    Regression Model

    This section describes an empirical regression model for the determinants of

    students performance using the cross-sectional data from the BS220 students

    survey. The analysis uses a multiple regression model to study the determinants of

    the dependant or Explained variable students academic performance which is

    measured by the students average score of first year modules at the university. To

    determine the significance of the independent variables that could explain the

    dependent variable I have executed several regression analyses between the

    originally selected ten explanatory variables and our dependant variable.

    Nevertheless, due to the small size and accuracy of our sample data most of these

    variables were not enough statistically significant at the 95% confidence interval to

    explain the dependant variable. Their P values were higher than 5%. However, Iindentified six independent variables to determine first year students performance.

  • 8/3/2019 Georgi Stoilov Econometrics

    9/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    9

    In my empirical research, I will perform the regression analysis on the subsequent

    explanatory variables: live on campus, study skills workshop, use of mathematics

    centre, TV watched, part-time employment and alcohol consumption.

    Estimation Strategy

    Having considered the above explanatory variables, the specification model was

    derived as follows:

    Acad.Perf. = 1 Live Camp. + 2 Study Work. +3 Math. Cent. + 4 TV Watch + 5

    Part-t. Job + 6 Alcohol + i

    The explanation of the variables in the model is the following:

    1 Live Camp. denotes whether the student live on campus or not

    2 Study Work. denotes have the students used the Study skills workshop

    service provided by the university

    3 Math. Cent. denotes have the students used the Math center service

    provided by the university

    4 TV Watch signifies the number of hours students spend watching TV

    weekly

    5 Part-t. Job signifies the number of hours students spend doing part-time

    job weekly

    6 Alcohol denotes how many units of alcohol the students consume weekly

    i is the error term.

    This model is used to measure how much does a unit change in one of explanatory

    variables leads to a particular unit change in the explained variable. The inclusion of

    the i (the residual term) means that we have some unobserved variables, random

  • 8/3/2019 Georgi Stoilov Econometrics

    10/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    10

    disturbance or the accuracy of our data is not reliable. Seeing as the sample data

    observed is quite small compared to the whole population, several problems are

    likely to arise such as the problems of multicollinearity. This mainly occurs when the

    regressors are more highly correlated with one another than with the dependent

    variable. To test for the presence of such abnormalities I will estimate the variance-

    inflation factor (VIF). Another possible statistical problem that primarily occurs in

    cross-section data is the heteroskedasticity. This exists when the magnitudes (of

    error term) continuously increase and decreasing with the number of variables

    observed. In order to indentify the presence of this problem I will perform a Whites

    General Test using STATA.

    Econometrics Results

    The estimation results for BS220 survey are reported in the table below. The total

    number of observations is 79. By looking at the T values of the variables we can

    determine which one of them are significant in explaining the students performance.

    In our mode not all of them are significant enough and they explain only 8.60% of the

    variation in the dependent variable, which is relatively small. This means that a large

    part of around 91.4% of the variation is missing. This can be due to the accuracy of

    the survey data and the size of the sample which was quite small to accurately

    represent the whole population.

  • 8/3/2019 Georgi Stoilov Econometrics

    11/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    11

    Variables Regression

    Coefficient

    Standard

    Error

    T- values p>t [95% Conf Interval]

    Live On Campus .6524029 3.590289 0.18 0.856 -6.504706 7.809512

    Study Skills

    Workshops

    1.39903 3.56374 0.39 0.696 -5.705155 8.503215

    Use of Math Center -7.129188 3.15224 -2.26 0.027 -13.41306 -.8453122

    TV Watched -.2018369 .225476 -0.90 0.374 -.6513151 .2476412

    Part-Time Job -.0422431 .0548477 -0.77 0.444 -.1515801 .0670939

    Alcohol

    Consumption

    .0242731 .1251488 0.19 0.847 -.2252065 .2737527

    Cons 67.81386 2.592498 26.16 0.000 62.64581 72.98191

    R-squared 0.0859

    Prob > F 0.3552

    Estimated Coefficients

    The coefficients for Live on campus and Study skills workshops are positive which

    means that students who live close to university and attend workshops in theLearning Development center will increase their academic performance by 0.65 and

    1.4 marks respectively. On the other hand the negative coefficients of the variables

    TV watched and Part-time jobs indicate that a one unit increase in the time that

    students spend watching television or work tend to decrease their average marks by

    0.2 and 0.04 respectively. Unexpectedly the regression analysis outcome shows,

    that students who make use of the math center are going to have 7.1 marks lower

    than other students. This I consider as either a mistake in the dataset or maybe

    students who use the math center are really bad in mathematics and these personal

    tutorials will not help them to improve their mark. Finally, the positive coefficient of

    alcohol of 0.02 shows that every unit of alcohol consumed will lead to 0.02 increase

    in the student mark. This I can relate to the fact that when students go out with

    friends more often, they consume alcohol, but they also relax and release themselves

    from and stress and therefore are more productive in their exams. Unfortunately

    almost all of the variable, except the use of math center, have low T values and

    relatively high confidence levels. This means that the variables in the regression

    model do not affect the Explained variable significantly and other variables are

    responsible for this. I tried various different combinations with all the variables

    provided in the data set but the outcomes were the same or even with lower T values

    and confidence levels. To solve this problem we may need a larger sample from the

    population and conduct further tests in order to achieve satisfactory results.

  • 8/3/2019 Georgi Stoilov Econometrics

    12/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    12

    Goodness of fit and diagnostic tests

    As mention previously, the R-squared value of the regression explains only 8.6% ofour data variation, which is a very small proportion. This determines the need for an

    increase of our observations and further testing.

    I have also performed the VIF and White test to detect if our expression model

    contains any multicollinearity or heteroskedasticity. The output shown below indicates

    that the model is free of any multicollinearity as the VIF values for all the variables

    are less than 10. The Explanatory variables used in our expression model are not

    highly correlated to each other more than to the dependant variable.

    The White Test has also been performed in order to check whether there is problemof Heteroskedasticity. From the table below we can see that the value of probability >

    chi 2 is 0.5102, which is much higher than 0.1, therefore we accept the hypotheses of

    Homoskedasticity, and conclude that our expression model used in our data is free of

    any Heteroskedasticity; i.e. there is no variability in the magnitude of our error term.

    Variable VIF 1/VIF

    TV Watched 1.17 0.851935

    Alcohol Consumption 1.10 0.906962

    Live On Campus 1.09 0.915507

    Part-Time Job 1.03 0.966622

    Study Skills Workshops 1.03 0.971603

    Use of Math Center 1.03 0.971867

    Mean VIF 1.03

  • 8/3/2019 Georgi Stoilov Econometrics

    13/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    13

    White's test for Ho: Homoskedasticity

    Against Ha: unrestricted heteroskedasticity

    chi2(12) 23.16

    Prob > chi2 0.5102

    Conclusion

    The core purpose of this paper was to test the relationship that exists between First

    year students results and a set of different lifestyle variables. We have used a

    multiple regression linear model to estimate the coefficients of our independent

    variables and see how they influence the dependant variable. Using this information

    presented in our data set, we tried to get the most accurate results that are useful in

    understanding the factors that affect students performance.

    The results revealed that First year academic performance is dependent on students

    amount of TV watched and part-time job hours per week, because it represents theamount of time forgone watching TV or working, thus the opportunity cost of which is

    studying, if we assume that this is the best students opportunity forgone. The data

    has also revealed that attending study skills workshops and living on campus have

    increased individuals abilities to perform better on test. However, due to the

    inaccuracy in some parts of our data I am uncertain about the validity of this

    conclusion. Furthermore, the results also indicate that the utilization of the math

    centre service by students affect negatively their performance, which is probably dueto the failure of our data.

    Based on the above mentioned econometrics results the following implication are

    expected to Improve first year students performance:

    University governance should put high emphasize on providing easily

    accessible workshops, but also promote awareness about the services among

    the students

  • 8/3/2019 Georgi Stoilov Econometrics

    14/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    14

    University accommodation governance should charge lower accommodation

    fees for its students in order to attract them to live on campus and enable them

    to use university and library facilities more often

    It is strongly recommended for students to spend less time watching TV,

    because that would provide them will spare time for more beneficial activities

    such as studying or sports

    To reach a more accurate conclusion a larger data sample is required, as this will

    enable us to carry out more complex and reliable statistical analysis. This sample

    was taken only from students studying BS2220 module, and is therefore only

    representative of segmental students. I would recommend the survey to be stretched

    to all first year students in Aston University giving us a better indication of how the

    variables that have been chosen actually affect students educational success. Many

    of the results we have found may be coincidental and anomalous. However, using a

    greater sample size will eliminate this doubt and give us a more accurate result.

  • 8/3/2019 Georgi Stoilov Econometrics

    15/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    15

    References:

    1. McKenzie, K. & Schweitzer, R. (2001) Who Succeeds at University? Factors predicting

    academic performance in first year Australian university students. Higher Education

    Research & DevelopmentVol. 20(1): pp.21-33.

    2. Lebcir, R. M., Wells, H. & Bond, A., (2008) Factors affecting academic performance of

    international students in project management courses: A case study from a British Post 92

    University International Journal of Project Management pp.5-10.

    3. Dorman, J. P., Fraser, B. J., & McRobbie, C. J. (1997) Relationship between school-level and

    classroom level environments in secondary schools. Journal of Educational Administration

    Vol. 35(1): pp.74-91.

    4. Michman. R., Mazze, E., & Greco, A. (2003) Lifestyle Marketing: researching the new

    American consumer, 1 edition. London: Praeger Publishers.

    5. Burd, E. & Hodgson, B. (2004) Attendance, and Attainment: a five-year study, CETL- Active

    Learning in Computing, South Rd, Durham.

    6. Rochford, C. & Connolly, M. & Drennan, J. (2009) Paid part-time employment and academic

    performance of undergraduate nursing students Vol. 29(6): pp.601-60.

  • 8/3/2019 Georgi Stoilov Econometrics

    16/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No. 150851

    16

    Appendix

    Commands used:

    .sum

    .cor

    .reg

    .vif

    .esta imtest, white

    Log file

  • 8/3/2019 Georgi Stoilov Econometrics

    17/17

    Analysis of the impact of academic and non-academic activities on first year

    students performance2011

    BS2220: Introduction to Econometrics | Candidate No 150851

    17


Recommended