Using Student Classification Specific Applications and Admissions Data to Forecast Enrollment

Lawrence J. Redlinger and Sharon EtheredgeThe University of Texas at Dallas

Presented at AIR 2004June 2, 2004

© 2003 Lawrence J .Redlinger, Sharon Etheredge and the Office of Strategic Planning and Analysis, The University of Texas at Dallas. No part of this presentation should be reproduced without permission. For questions, please contact OSPA at 972-883-6188 or spa@utdallas.edu

Our Purposes/Objectives

1. Discuss the assumptions and procedures to forecast enrollment.

2. Discuss some of the technical and contextual issues involved.

3. Discuss the use of the forecasts to establish applications and admissions targets that will keep student characteristics aligned with the strategic intentions of the university.

4. Discuss the use of forecasting as a targeting and policy tool for social action and for organizing the work of the university.

“Predictions are best made in a stable system where

trends are well established and rates of change for all

variables are known. It is even better if that stable

system is nested in a stable environment.”

FAT CHANCE

“The farther away the projected time is from the

present the more likely the projection will err by

some degree. This is especially so when the

environment is turbulent, the prediction involves

many variables, and/or the system is undergoing

continuous change.”

FAT CHANCE

FAT CHANCE’S DIMENSIONS OF CHANGE

The Sources of Change: “Internal” – “External”

The Duration of Change : Short to Long

Magnitude of Change : Small to Large

Frequency of Change : Single or Multiple

The Intensity of Change

The degree of Connectivity of Changes

The Threshold Where the Change makes a difference

Impact of Change (e.g., immediate or delayed)

And of course the

System’s Response to Change

E = I + (C- O)

Where I = INPUT STREAMS(First Time In College, Transfers, Non-degree Seeking Graduate

Students, Masters Students and Doctoral Students)

Where C = Continuing Students

Where O = OUTPUT STREAMS (Graduates, Drop-outs, Stop-outs, Transfers)

Three “Simple” Steps

1. Accurately estimate the Number of Continuing Students (C)

2. Estimate “Output and Loss” (O)3. Accurately estimate the Number of New

Students (I)

With enough lead time to allow organizational adaptations should they be

needed.

FOCUS: CONTINUING STUDENTS

1. Establish if persistence is a stable element or if there have been changes in persistence. In the illustration below, fall-to-fall and spring –to-fall persistence have been steadily climbing. Note also that the 7 Year Average difference between Fall-to-Fall and Spring-to-Fall is 4.37 percent. For the last three time periods, the difference is 4.5 percent.

2. Establish the appropriate data unit(s). For some institutions, that have active spring new enrollments and or staggered degree programs, spring-to-fall may provide a better measure of persistence.

Fall-to-Fall 1997-1998 1998-1999 1999-2000 2000-2001 2001-2002 2002-2003 2003-2004*Academic Persistence 53.4% 55.3% 55.1% 56.4% 58.9% 59.8% 60.8%

Spring-to-Fall 1998 1999 2000 2001 2002 2003 2004*Academic Persistence 59.7% 60.4% 59.2% 58.0% 63.3% 64.6% 65.1%

FOCUS: NEW STUDENTS

1. Establish Trends and/or Changes in Matriculation Rates (Admitted to Enrolled) for entering students by classification. Note that for some university’s the only meaningful classifications are freshmen and graduate students.2. The illustration below provides 6 years of data on the percent of admitted students by classification that actually enrolled. E.g., 49% of the admitted freshmen in fall 2003, actually enrolled.

Fall Semester 1998 1999 2000 2001 2002 2003

Freshman 0.5 0.5 0.53 0.55 0.45 0.49Sophomore 0.65 0.62 0.66 0.66 0.60 0.57

Junior 0.66 0.65 0.66 0.66 0.62 0.84Senior 0.67 0.66 0.67 0.67 0.59 0.72

Post-Bac. Non 0.68 0.66 0.65 0.68 0.65 0.81Masters 0.52 0.54 0.52 0.52 0.50 0.55Ph.D. 0.42 0.43 0.42 0.42 0.42 0.45

The next slide shows the input streams for fall 2003

Focus on New Students: Establish Input StreamsFall 2003 Streams

Fall 2003New

Applications Admitted

Number Applied

who Enrolled

Percent Applied

Percent of Applied who were Admitted

Percent of

Admitted Who

EnrolledFreshmen 5,402 2,348 1,151 100% 43% 49%

Sophomore 1,293 835 472 100% 65% 57%Junior 1,368 970 819 100% 71% 84%Senior 456 267 192 100% 59% 72%

Fall 2003New

Applications Admitted

Number Applied

who Enrolled

Percent Applied

Percent of Applied who were Admitted

Percent of

Admitted Who

Enrolled

Grad. Non-Degree Seeking 678 532 430 100% 78% 81%

Terminal Masters (e.g., MBA) 2,884 1,810 1,033 100% 63% 57%

Masters 224 122 63 100% 54% 52%Ph.D. 957 466 209 100% 49% 45%

Freshmen26%

Soph.11%

Junior19%

Senior4%

Post-Bac.10%

Masters25%

Ph.D.5%

Percent of Total New Student Applications of who enrolled Fall 2003 by Classification.

N= 4,369 new student enrollees

Transfer pools from other4-yr. institutions and Community colleges.

Appl i c a t i o ns by St ude nt Cl a s s i f i c a t i o n f o r Fa l l 2 0 0 2

0

100

200

300

400

500

600

700

800

900

2001-

06

2001-

07

2001-

08

2001-

09

2001-

10

2001-

11

2001-

12

2002-

01

2002-

02

2002-

03

2002-

04

2002-

05

2002-

06

2002-

07

2002-

08

2002-

09

2002-

10

F R S O J R S R G R S G M M T P H D

ESTABLISH THE PERIODICITY OF APPLICATIONS(as a means to staff and regulate work flow)

Continue to new slide

Appl i c a t i o ns by St ude nt Cl a s s i f i c a t i o n f o r Fa l l 2 0 0 2

0

100

200

300

400

500

600

700

800

900

2001-

06

2001-

07

2001-

08

2001-

09

2001-

10

2001-

11

2001-

12

2002-

01

2002-

02

2002-

03

2002-

04

2002-

05

2002-

06

2002-

07

2002-

08

2002-

09

2002-

10

F R S O J R S R G R S G M M T P H D

THE PERIODICITY OF APPLICATIONS(as a means to staff and regulate work flow)

Note that applications-by-date for each student classification has its own pattern. These patterns are quite consistent over time. For example, Junior transfers peak in two areas—after the fall semester ends and near the end of the spring semester. These patterns can be anticipated to manage office workflow, and processing.

NOISY DATA: Day-to-Day and Week-to-Week Activity Flows

Count of Freshman Applications Fall 2002

010

20304050

607080

90100

1999

0621

2001

0904

2001

0924

2001

1009

2001

1025

2001

1107

2001

1120

2001

1205

2001

1217

2002

0108

2002

0122

2002

0204

2002

0216

2002

0301

2002

0314

2002

0327

2002

0410

2002

0423

2002

0505

2002

0517

2002

0531

2002

0613

2002

0626

2002

0712

2002

0725

2002

0808

2002

0821

Before smoothing, the daily data appears quite noisy. The next slide shows this data smoothed by month.

Applications by Student Classification for Fall 2002

0

100

200

300

400

500

600

700

800

900

2001-06 2001-07 2001-08 2001-09 2001-10 2001-11 2001-12 2002-01 2002-02 2002-03 2002-04 2002-05 2002-06 2002-07 2002-08 2002-09 2002-10

FR

SMOOTHING BY USING MONTH

ESTABLISH RELATIVE STABILITYFocusing on Incoming Freshmen

Freshmen Applications for Fall Semesters, 2001, 2002, 2003 and Fall 2004 to April 2004, by Date of Application

0

200

400

600

800

1000

1200

Jan. Yr.1

Feb.Yr. 1

Mar.Yr. 1

Apr.Yr. 1

May Yr.1

JuneYr. 1

July Yr.1

Aug.Yr. 1

Sept.Yr. 1

Oct. Yr.1

Nov.Yr. 1

Dec.Yr. 1

Jan. Yr.2

Feb.Yr. 2

Mar.Yr. 2

Apr.Yr. 2

May Yr.2

JuneYr. 2

July Yr.2

Aug.Yr. 2

Date

Nu

mb

er

F-01 F-02 F03 F04

Note the consistency in the peaks

In this case, variations in amplitudeare due to modifications in recruitmentStrategy.

Establish Cumulative Application TrendsFocusing on Incoming Freshmen

Freshmen Applications for Fall Semesters 2001 to 2003 and Estimated for Fall 2004

0

1000

2000

3000

4000

5000

6000

Jan. Yr.1

Feb.Yr. 1

Mar.Yr. 1

Apr.Yr. 1

May Yr.1

JuneYr. 1

July Yr.1

Aug.Yr. 1

Sept.Yr. 1

Oct. Yr.1

Nov.Yr. 1

Dec.Yr. 1

Jan. Yr.2

Feb.Yr. 2

Mar.Yr. 2

Apr.Yr. 2

May Yr.2

JuneYr. 2

July Yr.2

Aug.Yr. 2

Time

Num

ber o

f App

licat

ions

F 2001 F 2002 F 2003 F 2004 actual & estimated

These cumulative trend lines will be used to construct an applications-admissions-enrollment target, and a forecasting line. See next slide

Application Targets Based on Prediction Line and a Target Enrollment of 1,300 based on a matriculation rate of 21% for

Freshmen for Fall 2004

448

2134

1452

6213

6160598657005333

46904078

3353

0

1000

2000

3000

4000

5000

6000

7000

Oct. Yr.1

Nov. Yr.1

Dec. Yr.1

Jan. Yr.2

Feb. Yr.2

Mar. Yr.2

Apr. Yr.2

May Yr.2

June Yr.2

July Yr.2

Aug. Yr.2

Time

Num

ber o

f Cum

ulat

ive A

pplic

atio

ns

This chart represents: a need for 6,213 applications to achieve 1,300 enrolled new freshmen based on the assumption that the 6,213 applications will yield 21% actual enrollment of freshmen with the student characteristics desired by the university (6,213 x .21 = 1304). Continue to next slide for more information

Application Targets Based on Prediction Line and a Target Enrollment of 1,300 based on a matriculation rate of 21% for

Freshmen for Fall 2004

448

2134

1452

6213

6160598657005333

4690

4078

3353

0

1000

2000

3000

4000

5000

6000

7000

Oct. Yr.1

Nov. Yr.1

Dec. Yr.1

Jan. Yr.2

Feb. Yr.2

Mar. Yr.2

Apr. Yr.2

May Yr.2

June Yr.2

July Yr.2

Aug. Yr.2

Time

Num

ber o

f Cum

ulat

ive A

pplic

atio

ns

This chart also represents a means to track the application process to measure whether or not recruitment efforts are meeting, lagging or exceeding time relevant targets. The chart will be used to create a forecasting on the next slide. Continue to next slide for more information.

ESTABLISH RELATIVE STABILITY“SMOOTHING INTERNAL ALTERATIONS”

1. The forecasting line SHOULD attempt account for INTERNAL changes in recruitment practices. This is easier said than done!

2. An accepted method for smoothing minor alterations in processesis to use a weighted average to establish the prediction lines and transitional probabilities.

3. By monitoring activity patterns in the recruitment area, we can contextually establish “weights” for the variables in the average.

4. The weighted average plus the performance target enrollment, assuming constant applications-admissions-matriculation probabilities, yields a prediction line for applications needed.

5. In the next slide, the forecasting line is based on a weighted average of fall 2002 at .33 and fall 2003 at .66. This is based on what we know about recruitment practices, scholarship packages, etc…

Fall 2004 Prediction Line with State Multipliers Based on a Weighted Average Model:

F02 + 2(F03) 3

13.871

11.0091.0381.0901.1651.3251.5241.853

2.9124.278

0

2

4

6

8

10

12

14

16

Oct. Yr.1

Nov. Yr.1

Dec. Yr.1

Jan. Yr.2

Feb. Yr.2

Mar. Yr.2

Apr. Yr.2

May Yr.2

June Yr.2

July Yr.2

Aug. Yr.2

Date

Stat

e Mul

tiplie

rs

Application Targets Based on Prediction Line and a Target Enrollment of 1,300 based on a matriculation rate of 21% for Freshmen for Fall 2004

448

2134

1452

6213

6160598657005333

4690

4078

3353

0

1000

2000

3000

4000

5000

6000

7000

Oct. Yr.1

Nov. Yr.1

Dec. Yr.1

Jan. Yr.2

Feb. Yr.2

Mar. Yr.2

Apr. Yr.2

May Yr.2

June Yr.2

July Yr.2

Aug. Yr.2

Time

Num

ber o

f Cum

ulat

ive A

pplic

atio

ns

Fall 2004 Prediction Line with State Multipliers Based on a Weighted Average Model:

F02 + 2(F03) 3

13.871

11.0091.0381.0901.1651.3251.5241.853

2.912

4.278

0

2

4

6

8

10

12

14

16

Oct. Yr. 1 Nov. Yr. 1 Dec. Yr. 1 Jan. Yr. 2 Feb. Yr. 2 Mar. Yr. 2 Apr. Yr. 2 May Yr. 2 June Yr. 2 July Yr. 2 Aug. Yr. 2

Date

Stat

e M

ultip

liers

ΣAf = (at1,2,…n)(mt1,2,…n)Where Af = total applications; a = applications at time 1…n, and m = state multiplier at time 1…n

Prediction Lines based on Rolling Probabilities for Freshman, Fall Semesters 2001-2003 and Target Line for Fall 2004

-1

4

9

14

19

24

Oct. Yr. 1

Nov. Yr. 1

Dec. Yr. 1

Jan. Yr. 2

Feb. Yr. 2

Mar. Yr. 2

Apr. Yr. 2

May Yr. 2

June Yr. 2

July Yr. 2

Aug. Yr. 2

Date

Expe

cted

Add

ition

al Vo

lum

e

F 2001 F 2002 F 2003 F 2004*

100% of all Applications

“Take-off”

Convergence

Here we see how the real data for past semesters and the prediction line coincide.

The Forecasting line againFall 2004 Prediction Line with State Multipliers Based on a Weighted

Average Model:F02 + 2(F03)

3

13.871

11.0091.0381.0901.1651.3251.5241.853

2.912

4.278

0

2

4

6

8

10

12

14

16

Oct. Yr. 1 Nov. Yr. 1 Dec. Yr. 1 Jan. Yr. 2 Feb. Yr. 2 Mar. Yr. 2 Apr. Yr. 2 May Yr. 2 June Yr. 2 July Yr. 2 Aug. Yr. 2

Date

Stat

e M

ultip

liers

ΣAf = (at1,2,…n)(mt1,2,…n)Where Af = total applications; a = applications at time 1…n, and m = state multiplier at time 1…n

Establish Cumulative Application Trends and Performance Targets

Focusing on Incoming FreshmenFreshmen Applications for Fall Semesters 2001 to 2003

and Estimated for Fall 2004

0

1000

2000

3000

4000

5000

6000

Jan. Yr.1

Feb.Yr. 1

Mar.Yr. 1

Apr.Yr. 1

May Yr.1

JuneYr. 1

July Yr.1

Aug.Yr. 1

Sept.Yr. 1

Oct. Yr.1

Nov.Yr. 1

Dec.Yr. 1

Jan. Yr.2

Feb.Yr. 2

Mar.Yr. 2

Apr.Yr. 2

May Yr.2

JuneYr. 2

July Yr.2

Aug.Yr. 2

Time

Num

ber o

f App

licat

ions

F 2001 F 2002 F 2003 F 2004 actual & estimated

Use of the forecasting line to establish targets

Anticipating and Accounting for Variations in Matriculation Rates

The following slide shows the number of needed applications based on a change in the assumed rate of matriculation.

If, for example, as a result of a tuition increase the rate of matriculation declines from 21% to 20%, an additional 219 applications will be needed to meet the goals of 1,300 new, enrolled freshmen.

On the other hand, if, for example, newly instituted financial aid programs and enhanced aid packages, raise matriculation to say 30%, a lower target can be established.

In general, matriculation rates vary by classes of institutions—from highly selective to open admissions– but for any institution, a forecasted rate can be established based on historical data and altered based on local knowledge.

SEE NEXT SLIDE.

Application Targets Based on Prediction Line and a Target Enrollment of 1,300 Freshmen for Fall 2004 Assuming Constant Admissions Rate (43%)

and Variable Matriculation Rate

2134

4350

448

33534078

4690

53335700

5986 6160 6213

1452

3471

4222

4856

55215901

6197 6377 6432

464

2209

1504

431341913991

32843734

2855

2347

1494

314

1017

0

1000

2000

3000

4000

5000

6000

7000

Oct. Yr.1

Nov. Yr.1

Dec. Yr.1

Jan. Yr.2

Feb. Yr.2

Mar. Yr.2

Apr. Yr.2

May Yr.2

June Yr.2

July Yr.2

Aug. Yr.2

Time

Num

ber o

f Cum

ulat

ive

Appl

icat

ions

E= App(.21) E= App(.20) E= App.(.30)

If the assumed matriculation rate is 20% (perhaps due to increases in tuition), the number of applications needed, jumps from 6,213 to 6,432. If for example, because of increases in student aid, the matriculation rate is 30%, the number of applications needed is reduced to 4,350. The point is that one must understand the contextual issues underlying yields in any academic year.

The Procedures we have been describing can be used on other student application categories. Below are charts of Junior applications.

Junior Applications by Application Date

0

50

100

150

200

250

300

Jan-Yr 1

Feb-Yr 1

Mar-Yr 1

Apr-Yr 1

May-Yr 1

June-Yr 1

July -Yr 1

Aug -Yr 1

Sept -Yr 1

Oct-Yr 1

Nov-Yr 1

Dec-Yr 1

Jan-Yr 2

Feb-Yr 2

Mar-Yr 2

Apr-Yr 2

May-Yr 2

June-Yr 2

July -Yr 2

Aug-Yr 2

Sept -Yr 2

Oct-Yr 2

F01 F02 F03

Cumulative Junior Applications For Fall Semesters 2001, 2002 and 2003

0

200

400

600

800

1000

1200

1400

1600

Jan-Yr 1

Feb-Yr 1

Mar-Yr 1

Apr-Yr 1

May-Yr 1

June-Yr 1

July -Yr 1

Aug -Yr 1

Sept-Yr 1

Oct-Yr 1

Nov-Yr 1

Dec-Yr 1

Jan-Yr 2

Feb-Yr 2

Mar-Yr 2

Apr-Yr 2

May-Yr 2

June-Yr 2

July -Yr 2

Aug-Yr 2

Sept-Yr 2

Oct-Yr 2

Time

Nu

mb

er

F01 F02 F03

Cumulative Junior Applications For Fall Semesters 2001,2002, 2003 and F04 Target

14 23 4290

156

385

685

891

1192

14071497 1499 1500

0

200

400

600

800

1000

1200

1400

1600

Oct- Yr 1 Nov- Yr 1 Dec- Yr 1 Jan- Yr 2 Feb- Yr 2 Mar- Yr 2 Apr- Yr 2 May- Yr2

June -Yr2

July -Yr2

Aug- Yr 2 Sept -Yr2

Oct- Yr 2

Time

Num

ber o

f App

licat

ions

F01 F02 F03 F04

The Procedures we have been describing can be used on other student application categories. Below are charts for Masters applications.

Applications Masters F01-F03 and Target F04

392139033701

32282803

2425

19961641

79192

5521235

0500

10001500200025003000350040004500

Oct- Yr1

Nov- Yr1

Dec- Yr1

Jan -Yr2

Feb -Yr2

Mar -Yr2

Apr -Yr2

May -Yr2

June -Yr 2

July -Yr2

Aug- Yr2

Sept -Yr2

Time

Num

ber

F01 F02 F03 F0 4

Forecast Line For F04 Masters Applications

1

20.42

49.60

7.10

3.18 2.39 1.96 1.62 1.40 1.21 1.06 1.000 .00

10 .00

20 .00

30 .00

40 .00

50 .00

60 .00

Oct- Yr1

Nov- Yr1

Dec- Yr1

Jan -Yr2

Feb -Yr2

Mar -Yr2

Apr -Yr2

May -Yr2

June -Yr 2

July -Yr2

Aug- Yr2

Sept -Yr2

Time

Phas

e M

ultip

liers

This slide shows the prediction lines for 5 student classifications at their individual “take-off” points and their “convergence” points.

Prediction Lines for Selected Applications by Student Classifications, Fall Semester 2003

0

20

40

60

80

100

120

Oct. Yr. 1 Nov. Yr. 1 Dec. Yr. 1 Jan. Yr. 2 Feb. Yr. 2 Mar. Yr. 2 Apr. Yr. 2 May Yr. 2 June Yr. 2 July Yr. 2 Aug. Yr. 2Date (Transitional Probabilities)

Add

ition

al V

olum

e

Freshmen Sophomores Juniors Masters Ph.D.

100% of all Applications

“Take-off”

Convergence

Does it work?

Predicted Versus Actual Enrollment for Fall Semesters 1998-2003

9,27

8

9,84

6

10,7

17 12,1

55

12,9

88

13,6

87

9,518 10,101

13,725

10,1

76

9,49

7

13,8

00

13,0

83

12,6

25

10,8

37

12,45510,945

13,229

0

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

1998 1999 2000 2001 2002 2003Fall Semester

Enro

llmen

t

Predicted Feb/March Predicted July Actual Enrollment

0

200

400

600

800

1,000

1,200

1,400

1,600

Num

ber o

f Stu

dent

s

SoftwareEng.

Interdis. Psy. MBA Acct. Bio. Elec.Eng. UND BA ComputerSci.

Majors

Application to Enrollment Sequences for Top Ten Majors listed by Applicants, Fall 2003

Applied Admitted Enrolled

The Top Ten Areas Account for 65% of all Applications and 32% of the New Fall Semester Enrollees

One can use the same procedures to establish target for specific areas.

Three “Simple” Steps

1. Accurately estimate the Number of Continuing Students (C)

2. Estimate “Output and Loss” (O)3. Accurately estimate the Number of New

Students (I)

With enough lead time to allow organizational adaptations should they be needed.

E = I + (C- O)1. Establish if persistence is a stable element or if there have been

changes in persistence; establish the best data unit for estimation (fall-spring).

2. Establish Trends and/or Changes in Matriculation Rates.

3. Establish Input Streams,their periodicity and relative stability.

4. Account for internal changes in recruitment and retention strategies.

5. Establish Cumulative Application Trends. Smooth when appropriate.

6. Set Performance Targets. Create alternative targets based on differing assumptions about persistence, admissions, and rates of matriculation. Monitor the capacity of key majors.