Issuu on Google+

Foundations of Enrollment Research

Ma#  S&mpson,  Ph.D.   Director  of  Enrollment  Research  and  Modeling   July  28,  2011  

© Performa Higher Education, LLC 2011 All Rights Reserved. Confidential Material: These materials may not be distributed without the consent of Performa Higher Education, LLC


ree Levels of Institutional Action

www.PerformaHE.com  


Operational •  Programs, People, and processes –  Right administrative system –  Right people responsible for administrative system –  Right processes to ensure data are accurate

www.PerformaHE.com  


Strategic •  Mining existing data –  Mapping inquiries, applied, admitted, and enrolled –  Pipelines –  Yield Rates –  Examine multiple points in the funnel •  Inquiry •  Applied •  Admitted •  Enrolled

www.PerformaHE.com  


Sample Map - Inquired

www.PerformaHE.com  


Sample Map – Enrolled

www.PerformaHE.com  


Sample Pipeline Comparison Top  10  An&cipated  Major  interests  for  all   inquiries:   –  Nursing  (2,689)   –  Business  Administra&on  (2,215)   –  Pre  Med/Vet/Dental  (2,128)   –  Other  (1,703)   –  Psychology  (1,423)   –  Communica&on  Arts  (1,384)   –  Educa&on  (1,374)   –  Music  (1,041)   –  Early  Childhood  Educa&on  (1,009)   –  Biology  (987)  

www.PerformaHE.com  

Top  10  An&cipated  Major  interests  for  all   enrolled  students:   –  Nursing  (213)   –  Early  Childhood  Educa&on  (93)   –  Business  Administra&on  (80)   –  Psychology  (46)   –  Biology  (45)   –  Accoun&ng  (40)   –  Sports  Management  (37)   –  Zoo  Biology  (37)   –  Middle  Childhood  Educa&on  (35)   –  Communica&on  Arts  (34)  


Sample Yield Rate Table

Source  Code  

Enrolled  

Not  Enrolled  

Total  

Yield  

Name  of   source  code  

Total  number   of  enrolled   students  with   source  code  

Total  number   students  who   did  not  enroll   with  source   code  

Total  number   of  students   enrolling  and   not  enroll  with   source  code  

(Enrolled/ Total)*100  

NRCCUA  

10  

1000  

1010  

0.9%  

Visit  

150  

985  

1135  

13.22%  

www.PerformaHE.com  


Leverage •  Predictive Modeling •  Admitted Student Research •  Benchmarking

www.PerformaHE.com  


What is Predictive Modeling •  Normally referrers to some type of regression analysis •  Regression is an attempt to predict an outcome based on one or more variables •  Predicting college GPA based on previous academic performance •  Predicting probabilities of contracting a disease based on a set of health factors •  Predicting probabilities of enrollment based on a set of criteria

www.PerformaHE.com  


Predictive Modeling •  Allows for a probability of enrollment to be assigned to every student in the inquiry pool •  Students with higher probabilities of enrollment are targeted more strategically –  Helps to streamline communications –  Significant cost savings can be achieved –  Staff are able to efficiently prioritize inquiries –  Output can be difficult to understand –  Requires a fairly complex understanding of statistics to accomplish www.PerformaHE.com  


How to Accomplish Predictive Modeling •  •  •  • 

Step 1: Identify variables Step 2: Fit a regression equation Step 3: Test the regression equation Step 4: Apply regression equation to current inquiry pool

www.PerformaHE.com  


Step 1: Identify Variables •  Consider the variables to be used –  Anticipated Major –  Counselor –  Demographic Information –  Market (in or out of market?) –  Qualities and characteristics of a student s zip code –  Source Code –  Territories

www.PerformaHE.com  


Probability of Enrollment

Step 2: Fit a regression equation

Independent Variable(s) www.PerformaHE.com  


Probability of Enrollment

Step 2: Fit a regression equation

Major

Source Code Gender Ethnicity

Independent Variable(s) www.PerformaHE.com  


Step 2: Fit a regression equation

Probability of Enrollment

Major (2.139)

Source Code (1.985) Intercept (0.989)

Ethnicity (1.012)

Independent Variable(s) www.PerformaHE.com  

Gender (1.345)


Step 2: Fit a regression equation Ln(p)=0.989+2.139(Major)+1.985(Source Code)+1.345(Gender)+1.012(Ethnicity)

Probability of Enrollment

Major

Source Code Intercept

Gender Ethnicity

Independent Variable(s) www.PerformaHE.com  


Step 3: Test the Regression Equation Rank  

Enrolled  

Predicted   Enrolled  

Not  Enrolled  

Predicted  Not   Enrolled  

1  

341  

394.65  

1856  

1802.35  

2  

233  

223.78  

1955  

1964.22  

3  

122  

140.28  

2080  

2061.72  

4  

82  

98.50  

2114  

2097.50  

5  

55  

59.65  

2139  

2134.35  

6  

35  

37.38  

2163  

2160.62  

7  

23  

28.44  

2172  

2166.56  

8  

15  

18.59  

2183  

2179.41  

9  

9  

9.50  

2186  

2185.50  

10  

1  

2.43  

2195  

2193.57  

www.PerformaHE.com  


Step 3: Test the Regression Equation Rank  

%  Enrolled  

%  Not   Enrolled  

%  Total   Enrolled  

Cum  %  of   Enrolled  

1  

15.5  

84.5  

37.2  

37.2  

2  

10.6  

89.4  

25.4  

62.6  

3  

5.5  

94.5  

13.3  

75.9  

4  

3.7  

96.3  

9.0  

84.9  

5  

2.5  

97.5  

6.0  

90.9  

6  

1.6  

98.4  

3.8  

94.7  

7  

1.0  

99.0  

2.5  

97.2  

8  

0.7  

99.3  

1.6  

98.8  

9  

0.4  

99.6  

1.2  

100.0  

10  

0.0  

100.0  

0.0  

100.0  

www.PerformaHE.com  


Step 4: Apply Regression Equation to Current Inquiry Pool 1 Probability of Enrollment in Log Odds

Sue Johnston, p = 89%

0

-1

Patrick Stuart, p = 45%

Bill Paul, p = 5% All Predictor Variables

www.PerformaHE.com  


Using a Model •  Predicted probabilities provide a path of engagement priority •  Engage the students most likely to enroll with the most expensive and resource heavy methods •  Order students by –  Actual probabilities –  Decile ranks –  Percentile ranks

www.PerformaHE.com  


Benefits •  Cost savings –  Significant savings can be achieved by modifying communication flow based on predictive model

•  Prioritize –  Counselor time –  Telecounseling –  Travel

www.PerformaHE.com  


Foundations of Enrollment Research