Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

2022

BNAD 277 Final Project

COLLEGE AND UNIVERSITY ANALYSIS


GROUP #71: JOHNMICHAEL FILIPPONE, CHARLIE KRAFT,
ELSA LUNDBLAD, KARLY PETERSEN, & TAYLOR SCHULZE
Table of Contents
Table of Figures ......................................................................................................................... - 2 -
Introduction ................................................................................................................................ - 3 -
Data Exploration Using Tableau .............................................................................................. - 3 -
Statistical Analysis ..................................................................................................................... - 3 -
Regression Equation ............................................................................................................. - 3 -
Fit of the Model ..................................................................................................................... - 4 -
Test of Joint Significance...................................................................................................... - 4 -
Interpretation of the Coefficients ........................................................................................ - 4 -
Base Case Variables .............................................................................................................. - 4 -
Percent of First-Time Undergraduates In-State & Out-of-State ..................................... - 5 -
Percent of Total Enrollment That Are Caucasian ............................................................. - 5 -
Common Violations Check................................................................................................... - 5 -
Non-linear Patterns ........................................................................................................... - 5 -
Heteroskedasticity ............................................................................................................. - 5 -
Endogeneity ....................................................................................................................... - 5 -
Multicollinearity ................................................................................................................ - 5 -
Relevance of Findings ............................................................................................................... - 5 -
Recommendations ...................................................................................................................... - 6 -
Conclusion .................................................................................................................................. - 6 -
Appendix ..................................................................................................................................... - 7 -
Residual Plots ........................................................................................................................ - 8 -
Check For Multicollinearity and Endogeneity ................................................................. - 10 -

-1-
Table of Figures
Figure 1- Final Regression Output ................................................................................... - 7 -
Figure 2- % of First-Time Undergraduates In-State & Out-of-State Regression Output . - 7 -
Figure 3- % of Total Enrollment that are Caucasian Regression Output ......................... - 8 -
Figure 4- Average Tuition and Fees Residual Plot............................................................ - 8 -
Figure 5- Percent of Freshman Receiving any Financial Aid Residual Plot ...................... - 9 -
Figure 6- Full-Time Enrollment Residual Plot ................................................................. - 9 -
Figure 7- Part-Time Enrollment Residual Plot ................................................................. - 9 -
Figure 8- Percent of Total Enrollment that are Caucasian ............................................. - 10 -
Figure 9- Correlation Matrix.......................................................................................... - 10 -

-2-
Introduction
The purpose of this document is to identify factors that impact a university’s undergraduate
graduation rate for the United States Department of Education. Using data from over 1,500
Universities and Colleges from across the U.S., our team used the factors of Average Tuition and
Fees from 2010-2014, Public Institutions(d), Percent of Freshman Receiving any Financial Aid,
City(d), Town(d), Suburb(d), Average Tuition x Public Institutions(d), Full-Time Enrollment, Part-
Time Enrollment, and Percent of Total Enrollment that are Caucasian. We have analyzed the
data using a regression analysis at the significance level of 0.1.
Based on the fit of the regression model, we can conclude that we are 65.59% of the way toward
being able to perfectly predict Undergraduate Graduation Rate using the factors listed above. The
standard error of 11.7233 tells us that the Undergraduate Graduation Rate from our model is off
by an average of 11.7233 percentage points. From the model, we also found that Part-Time
Enrollment and Percent of Freshman Receiving any Financial Aid decreases undergraduate
graduation rate. This report also incorporates other attempted models using different data
variables and data exploration using Tableau to visualize the datasets. The visuals from Tableau
help show that type of institution (Public vs. Private) and urbanization location have the largest
impact on undergraduate graduation rate. Our team recommends that the United States
Department of Education use our final regression model when planning new institutions to
utilize a high undergraduate graduation rate.

Data Exploration Using Tableau


Using Tableau to analyze the 6-year graduation rate from zip codes and states across the country,
we found that the state with the highest 6-year undergraduate graduation rate is Rhode Island
with an average of 68.63%. The zip code with the highest 6-year undergraduate graduation rate
is 02912 in Providence, Rhode Island; at an average of 95%. Based on our visuals of graduation
rate compared to urbanization, small urban areas have the highest average number of students
receiving a bachelor’s degree. Moreover, midsize urban areas have the highest bachelor’s degree
graduation rate on average. Comparing graduation rate across institution type, private not-for-
profit colleges have a higher graduation rate for bachelor’s degrees than public colleges.

These Tableau visualizations can be shown here.

Statistical Analysis
Regression Equation

𝑈𝑛𝑑𝑒𝑟𝑔𝑟𝑎𝑑𝑢𝑎𝑡𝑒 ̂ 𝐺𝑟𝑎𝑑𝑢𝑎𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒 = 324.92 + 0.0014(𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑇𝑢𝑖𝑡𝑖𝑜𝑛)


+ 3.53(𝑃𝑢𝑏𝑙𝑖𝑐(𝑑) ) − 0.24(% 𝑜𝑓 𝐹𝑟𝑒𝑠ℎ𝑚𝑎𝑛 𝑅𝑒𝑐𝑒𝑖𝑣𝑖𝑛𝑔 𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝐴𝑖𝑑) + 5.65(𝐶𝑖𝑡𝑦(𝑑) )
+ 6.84(𝑇𝑜𝑤𝑛(𝑑) ) + 6.60(𝑆𝑢𝑏𝑢𝑟𝑏(𝑑) ) + 0.0009(𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑇𝑢𝑖𝑡𝑖𝑜𝑛 𝑥 𝑃𝑢𝑏𝑙𝑖𝑐(𝑑) )
+ 0.0012(𝐹𝑢𝑙𝑙-𝑇𝑖𝑚𝑒 𝐸𝑛𝑟𝑜𝑙𝑙𝑚𝑒𝑛𝑡) − 0.0019(𝑃𝑎𝑟𝑡-𝑇𝑖𝑚𝑒 𝐸𝑛𝑟𝑜𝑙𝑙𝑚𝑒𝑛𝑡)
+ 0.17(% 𝑜𝑓 𝑇𝑜𝑡𝑎𝑙 𝐸𝑛𝑟𝑜𝑙𝑙𝑚𝑒𝑛𝑡 𝑡ℎ𝑎𝑡 𝑎𝑟𝑒 𝐶𝑎𝑢𝑐𝑎𝑠𝑖𝑎𝑛)

-3-
Fit of the Model
R2 = 0.6559 which tells us we are 65.59% of the way toward perfectly predicting a university’s
undergraduate graduation rate.
Standard Error (SE) = 11.7233 which tells us our predictions of undergraduate graduation rate are
off by an average of 11.7233 percentage points.
Click here to see our final regression model.

Test of Joint Significance


H0: None of our independent variables impacts graduation rate
HA: At least one of our independent variables significantly impacts graduation rate.
Our significance F is 0 which less than our significance level of .1, this means that we reject our
null hypothesis and can say that at least one of our independent variables significantly effects
graduation rate.

Interpretation of the Coefficients


o Average Tuition and Fees – As average tuition and fees increase by $1000, undergraduate
graduation rate increases by 1.4 percentage points, on average and all else constant.
o Percent of Freshmen Receiving Any Financial Aid – As percent of freshmen receiving
any financial aid increases by 1 percentage point, undergraduate graduation rate
decreases by .2421 percentage points, on average and all else constant.
o Full-time Enrollment – As full-time enrollment increases by 1, undergraduate graduation
rate increases by .0012 percentage points, on average and all else constant.
o Part-time Enrollment – As part-time enrollment increases by 1, undergraduate graduation
rate decreases by .0019 percentage points, on average and all else constant.
o Percent of Total Enrollment that are Caucasian – As percent of total enrollment that are
Caucasian increases by 1 percentage point, undergraduate graduation rate increases by
.1564 percentage points, on average and all else constant.
o Public(d) – Public schools have an undergraduate graduation rate of 3.5338 percentage
points higher than private schools, on average and all else constant.
o City(d) – Schools in cities have an undergraduate graduation rate of 5.6482 percentage
points higher than schools in rural areas, on average and all else constant.
o Town(d) – Schools in towns have an undergraduate graduation rate of 6.8427 percentage
points higher than schools in rural areas, on average and all else constant.
o Suburb(d) – Schools in suburbs have an undergraduate graduation rate of 6.6047
percentage points higher than schools in rural areas, on average and all else constant.
o Average Tuition and Fees x Public(d) – Each additional $1000 of tuition and fees is worth
1.82 percentage points more for public schools than for private schools, on average and
all else constant.

Base Case Variables


o For the dummy variable Public(d), d = 1 when an institution is public and d=0 when an
institution is private.
o For the dummy variables City(d), Town(d), and Suburb(d), d =1 when an institution is in a
city, town, or suburb regarding urbanization and d = 0 when an institution is in a rural
area.

-4-
Percent of First-Time Undergraduates In-State & Out-of-State
Our team incorporated a new model which can be shown here to include the variables % of first-
time undergraduates in-state and % of first-time undergraduates out-of-state. These variables
reduced the standard error to 11.0004 and increased the adjusted R2 to 0.6656 compared to
0.6536. However, their p-values of 0.5114 and 0.8261 are not less than the significance level of
0.1 which indicates that these variables are not statistically significant in determining the
undergraduate graduation rate.

Percent of Total Enrollment That Are Caucasian


For this model, which can be found here, our team included the variable % of total enrollment
that are Caucasian. We found this variable to be statistically significant since its p-value of
2.9589E-23 is less than our significance level of 0.1. When adding this variable, we found that
the variables of total percent admitted, and total admissions became statistically insignificant
since their p-values were not less than our significance level of 0.1.

As Percent of Total Enrollment that are Caucasian increases by 1 percentage point,


Undergraduate Graduation Rate increases by 0.1513 percentage points, on average and all else
constant.

Common Violations Check


Non-linear Patterns
There is no evidence of non-linear patterns since the residual plots do not show any curved
patterns.
Heteroskedasticity
Heteroskedasticity is detected in the residual plots of Full-Time Enrollment, Part-Time
Enrollment and Percent of Freshmen Receiving any Financial Aid. This is shown by all three of
the residual plots have a funnel shape to the data. The presence of heteroskedasticity tells us that
the coefficient estimates are biased, and our hypothesis are no longer valid or trustworthy. To fix
this error, we would use White’s standard errors.
Endogeneity
Endogeneity is not present because the correlation between the residuals and the independent
variables are all less than 0.2. This means we did not have to calculate a test statistic for any of
the independent variable’s correlations or perform significance testing.
Multicollinearity
The correlation matrix does not show that multicollinearity is present between any of the
variables because none of the correlations are above a 0.8 or below a -0.8.

Relevance of Findings
In making our final model, we chose to include the variables Percent of Total Enrollment that are
Caucasian and Public(d) since both were statistically significant. The dummy variable of Public(d)
represents public vs. privately owned universities with public=1 and private=0. This variable
lowered our standard error and increased our adjusted R2 for the final model. We chose not to
include the variables Percent of First-Time Undergraduates In-State and Percent of First-Time
Undergraduates Out-of-State since they were not statistically significant. We also found it

-5-
surprising that as the percentage of freshman receiving financial aid increased, graduation rate
decreased.

Recommendations
We recommend for the United States Department of Education to use this model when planning
new institutions and to help determine the success of the institution based on the plans USDE
intends to follow. Such as location, type of institution, average tuition, and fees, and even how
many students they intend to accept for part-time and full-time enrollment. We also believe that
pursuing testing further on location could be very beneficial as urbanization location had the
largest impact on graduation rate. Having more information on how location affects graduation
rate could also help USDE make more informed decisions on where to place new institutions.

Conclusion
Based off the dataset and our findings, we can conclude that location has a large impact on
graduation rate; and part-time enrollment and percent of freshman receiving any financial aid
negatively affects graduation rate. Next steps will be to further evaluate and explore how
location affects graduation rate. For any question on our finding or next steps please contact our
team at group71@email.USDE.com.

-6-
Appendix
Figure 1- Final Regression Output

(Click here to go back findings)

Figure 2- % of First-Time Undergraduates In-State & Out-of-State Regression Output

(Click here to go back to findings)

-7-
Figure 3- % of Total Enrollment that are Caucasian Regression Output

(Click here to go back to findings)

Residual Plots

Figure 4- Average Tuition and Fees Residual Plot

Avg Tuition and Fees, 2010-2014 Residual Plot


70

20
Residuals

0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000
-30

-80
Avg Tuition and fees, 2010-2014

-8-
Figure 5- Percent of Freshman Receiving any Financial Aid Residual Plot

Percent of Freshmen Receiving Any Financial


Aid Residual Plot

20
Residuals

-30 10 20 30 40 50 60 70 80 90 100

-80
Percent of freshmen receiving any financial aid

(Click here to go back to findings)

Figure 6- Full-Time Enrollment Residual Plot

Full-time Enrollment Residual Plot


70

20
Residuals

-2500 2500 7500 12500 17500 22500 27500 32500 37500 42500
-30

-80
Full-time enrollment

(Click here to go back to findings)

Figure 7- Part-Time Enrollment Residual Plot

Part-time Enrollment Residual Plot


70

20
Residuals

-2500 2500 7500 12500 17500 22500


-30

-80
part-time enrollment

(Click here to go back to findings)

-9-
Figure 8- Percent of Total Enrollment that are Caucasian

Percent of Total Enrollment that are White


Residual Plot

40
Residuals

-10
0 10 20 30 40 50 60 70 80 90 100

-60
Percent of total enrollment that are white

Check For Multicollinearity and Endogeneity

Figure 9- Correlation Matrix

- 10 -

You might also like