The effect of investment on school building and student performance 1. Statement of the problem This paper will address the effect of investment on school facilities to student performance. We are interested especially in seeing whether additional money spend on school buildings lead to improvement of student performance, keeping every other factors constant. 2. Review of the literature There are a lot of literature which attempted to explain the effects of school resources on student performance, but the relationship between school resources and student performance has been controversial. First of all, Hanushek (1997) provided huge volume of literature review about the effects of school resources on student performance. After review of around 400 previous studies of student achievement, he found that there is not a strong or consistent relationship between student performance and school resources including the real resources of the classroom (teacher education, teacher experience, and teacher-pupil ratios), financial aggregates of resources (expenditure per student and teacher salary), and measures of other resources in school (specific teacher characteristics, administrative inputs, and facilities). On the other hand, Eide and Showalter (1998) studied the effect of school quality on student performance. In this paper, they found that the number of school enrollment and school year length have positive relation with student performance. They also observed that there may be differential school quality effects at different points in the test score gain conditional distribution. Wӧßmann (2003) investigated that the effects of family background, resources and institutions on students’ mathematics and science performance using an international database of more than 260,000 students from 39 countries. His result showed that international differences in student performance cannot be attributed to resource differences, but they are considerably related to institutional differences. He found that centralized examinations and control mechanisms, school autonomy in personnel and process decisions, individual teacher influence over teaching methods, limits to teacher unions’ influence on curriculum scope, scrutiny of students’ achievement and competition from private schools have positive effects on the student performance. 3. The data The data for this paper consists of 1,967 observations of Ohio Elementary School buildings for 2001-2002 year. For each observation we have 1) the number of teachers, 2) teacher attendance rate, 3) average years of teaching experience, 4) average teacher salary, 5) per pupil spending on instruction, 6) per pupil spending on building operations, 7) per pupil spending on administration, 8) per pupil spending on pupil support, 9) per pupil spending on staff support, 10) median of 4th grade prof scores(student performance), 11) building enrollment, 12) per capita income in the zip code area, 13) percent of p ...

1. The effect of investment on school building and student
performance
1. Statement of the problem
This paper will address the effect of investment on school
facilities to student performance. We are interested especially in
seeing whether additional money spend on school buildings lead
to improvement of student performance, keeping every other
factors constant.
2. Review of the literature
There are a lot of literature which attempted to explain the
effects of school resources on student performance, but the
relationship between school resources and student performance
has been controversial. First of all, Hanushek (1997) provided
huge volume of literature review about the effects of school
resources on student performance. After review of around 400
previous studies of student achievement, he found that there is
not a strong or consistent relationship between student
performance and school resources including the real resources
of the classroom (teacher education, teacher experience, and
teacher-pupil ratios), financial aggregates of resources
(expenditure per student and teacher salary), and measures of
other resources in school (specific teacher characteristics,
administrative inputs, and facilities).
On the other hand, Eide and Showalter (1998) studied the effect
of school quality on student performance. In this paper, they

2. found that the number of school enrollment and school year
length have positive relation with student performance. They
also observed that there may be differential school quality
effects at different points in the test score gain conditional
distribution.
Wӧßmann (2003) investigated that the effects of family
background, resources and institutions on students’ mathematics
and science performance using an international database of
more than 260,000 students from 39 countries. His result
showed that international differences in student performance
cannot be attributed to resource differences, but they are
considerably related to institutional differences. He found that
centralized examinations and control mechanisms, school
autonomy in personnel and process decisions, individual teacher
influence over
teaching methods, limits to teacher unions’ influence on
curriculum scope, scrutiny of students’ achievement and
competition from private schools have positive effects on the
student performance.
3. The data
The data for this paper consists of 1,967 observations of Ohio
Elementary School buildings for 2001-2002 year. For each
observation we have 1) the number of teachers, 2) teacher
attendance rate, 3) average years of teaching experience, 4)
average teacher salary, 5) per pupil spending on instruction, 6)
per pupil spending on building operations, 7) per pupil spending
on administration, 8) per pupil spending on pupil support, 9) per
pupil spending on staff support, 10) median of 4th grade prof
scores(student performance), 11) building enrollment, 12) per
capita income in the zip code area, 13) percent of population

3. that is non-white, 14) poverty percent of population in poverty,
15) % of population attending public schools.
After pre-screening of data set, 5 observations are excluded
from original dataset because they have abnormal values
(decimal points on the number of teachers, negative amount for
per pupil spending on pupil support and staff support). So our
final sample consists of 1,962 observations.
4. The empirical results and conclusions
At first, we include all variables that might have relationship
with the student performance. Here the first ordinary least
squares regression is:
STUDENT PERFORMANCE
=-108.613
-0.327 ∗
0.036 ∗
(23.485)
(0.100)
(0.030)
0.505 ∗
0.001
∗
(0.116)
(0.000001)
0.001 ∗
0.002 ∗
0.002 ∗
(0.0004)

4. (0.001)
(0.003)
0.004 ∗
0.004 ∗
0.015 ∗
(0.001)
(0.002)
(0.007)
13.241 ∗
19.015 ∗
30.749 ∗
38.009 ∗
(2.290)
(2.677)
(5.396)
(6.830)
, se in parentheses, R2=.357, n=1962.
This regression shows that there is relationship between per
pupil spending on building operation and median of 4th grade
proficiency score, but only at the relatively high p-value
(p<0.1). In addition, some variables are not statistically
significant even at the p-value of 0.1 (the rate of teacher
attendance, per pupil spending on administration and staff
support). Thus, we excluded those variables and get the new
regression model.

5. STUDENT PERFORMANCE
107.543
0.32 ∗
0.559 ∗
(23.379)
(0.100)
(0.112)
0.0004 ∗
0.001 ∗
0.002 ∗
(0.000001)
(0.0004)
(0.001)
0.004 ∗
0.013 ∗
13.666 ∗
(0.001)
(0.007)
(2.280)
19.574 ∗
30.747 ∗

6. 37.563 ∗
, se in parentheses, R2=.357, n=1962.
(2.649)
(5.389)
(6.826)
In this model, all variables are statistically significant (p<0.01),
except spending on building operation (p<0.06) and building
enrollment (p<0.05). In addition, although we excluded some
variables, the value of R2 doesn’t change. To check multi
collinearity problem, we checked the VIF (Variance Inflation
Factors) values of all variables. All VIF values of variables are
below 10, so we can assume that there is no multi collinearity
problem in the model. However, the White’s test shows that
there is heteroskedasticity problem in the model ( P[Chi-
square(77) > 558.405]=0.000 ). To solve this problem, we
regress the model again with robust standard errors.
STUDENT PERFORMANCE
107.543
0.32 ∗
0.559 ∗
(22.870)
(0.141)
(0.110)
0.0004 ∗

7. 0.001 ∗
0.002 ∗
(0.0001)
(0.001)
(0.001)
0.004 ∗
0.013 ∗
13.666 ∗
(0.002)
(0.009)
(2.297)
19.574 ∗
30.747 ∗
37.563 ∗
, se in parentheses, R2=.357, n=1962.
(3.183)
(5.502)
(7.526)
In this final model, we can see that there is positive relationship
between spending on building operation and student

8. performance at the 5% level. It means that we would predict for
the median proficiency score of students to increase by 0.002
for 1 unit increase in per pupil spending on building operation.
We also see that number of teachers, spending on pupil support
are significant at the 5% level, and average year of teaching,
average of teacher salary, per capita income, percent of non-
white, percent of poverty, and percent of attending public
school are significant at the 1% level. However, with robust
standard errors, spending on instruction and building enrollment
are no longer significant in this model.
5. Possible extensions and limitations of your study
One of limitation in this study is the interpretation of the
results. We cannot find plausible explanation for the unexpected
signs of variables. Unlike our expectation, the sign of number
of teachers was negative, and the sign of percent of attending
public school was positive. In addition, the coefficient for the
main independent variable is too small, so it is hard to conclude
that there is meaningful relationship between dependent
variable and independent variable, although existence of the
statistical significance in coefficient and overall model. It could
be the result of omitted variables, such as some important
factors which were not included in the data set. In the future
study, scrutiny with each variables will be needed.
Finally, it would be better if we can include the qualitative
variables in the model. As previous studies suggested,
qualitative factors might have important impacts on the
performance of students. In the future study, we can improve
quality of the study including those kinds of variables.
6. References
Hanushek, Eric A., “Assessing the Effects of School Resources
on Student Performance: An Update.” Educational Evaluation

9. and Policy Analysis, Vol. 19, No. 2, pp. 141-164, 1997.
Eide, Eric and Showalter, Mark H., “The effect of school
quality on student performance: A quantile regression
approach.” Economics Letters, Vol 58, pp. 345-350, 1998.
Wӧßmann, Ludger., “Schooling Resources, Educational
Institutions and Student Performance: the International
Evidence.” Oxford Bulletin of Economics and Statistics, Vol 65,
Issue 2, pp. 117-170, 2003.
Term paper information
Choose the project described below for your paper.
Try different specifications for your model, for example, you
may want to try a double logarithmic model as well as a model
in levels. Consider including squares of variables, including
interaction terms, or creating additional dummy variables if that
should be a good idea. Examine your dataset carefully; for
example, perhaps you will want to delete observations that you
think do not make sense. Try to be creative and critical. Try to
do things like add extra regressors if that should be possible.
Sometimes relationships between variables in your dataset are
very unlikely to be linear. Multiple regressions are very often
more informative than simple regressions for the reasons
discussed. Be aware that your instructor will only be able to
judge your work on what you hand in, and not on all the work
you may have done but cannot be found back in your final
product. Also, take great care not to misinterpret output or draw
conclusions that are verifiably incorrect.
For reporting your regression results, follow the following
format (example):.
WGE = 1234 + 56.3∗ EXP − 901∗ F + 341∗ W − 344∗ E −
861∗ S

10. (231) (26) (540) (562) (668) (659) se in parentheses; R^2=
0.63, n = 1031
Do not include Gretl output in your paper.
If you can find literature or references about the topic of your
choice, mention these in the appropriate section; it will improve
your paper’s quality. Try different model specifications, and try
to work towards a “final model” if possible. Take care to choose
a sensible model specification; for example, do not use a
regressor that is coded “1” for train, “2” for bus, “3” for
airplane.
Low t-values can still make an interesting project. The point
here is to do economic or social science style research; it is
your job to objectively report what is suggested by the data. If
you think there ought to be a relationship between y and x, but
you don’t find one, then that is a perfectly fine outcome of a
piece of research. Scientific research should ideally be free of
biases, and reporting unbiased conclusions in a verifiable and
correct way is what science should ideally be all about.
Your paper should not exceed four double-spaced typed pages.
Your paper should be typed. The four-page limit is not there to
make your life difficult, but instead, the upper limit serves to
limit the amount of work that you can usefully spend on this
project. The idea is for you to be precise and concise.
For your paper, use the following section format. You are
explicitly allowed and encouraged to make minor changes to
this format if you think this is appropriate; however, you need
to subdivide your paper in sections.
1. Statement of the problem
2. Review of the literature
3. The data
4. The empirical results and conclusions
5. Possible extensions and limitations of your study
6. Acknowledgements (not always necessary)
7. References
Your paper should be self-contained as much as is reasonably

11. possible. That is, the paper should be readable without prior
knowledge of the dataset, and you should introduce any
nonstandard symbols or notation. An Undergraduate
Econometrics student from say, University of Indiana should be
able to read your paper and make sense of it.
Project descriptions
Wages in Malaysia
Marcia Schafgans analyzed Malaysian wage data in an article
titled ”Ethnic Wage Differences in Malaysia: Parametric and
Semiparametric Estimation of the ChineseMalay Wage Gap”.
This article appeared in Journal of Applied Econometrics. At
the internet location
http://qed.econ.queensu.ca/jae/1998-v13.5/schafgans/
you will find a description a datafile and a zipped data file. This
data set contains wages and characteristics of 8748 people
living in Malaysia of different ethnicity (Malay, Indian, and
Chinese). We are interested in analyzing wage discrimination in
Malaysia, and casual conversations with Malaysian grad
students suggest the potential presence of such discrimination.
Examine the end of the datafile carefully - it looked to me as
though there was an obsolete number at the end of this datafile.
For this project, regress the “LWAGE” variable on regressors
and try to draw conclusions about ethnicity based wage
differences in Malaysia.
There is a substantial literature in Labor Economics about the
estimation of wage equations, and for this project it should not
be too difficult to write a decent overview of the literature.