The Higher Education Enrollment Decision: Feedback on Expected Study Success and Updating Behavior

Figures & data

Figure 1. Timing of events from application moment until end of field experiment.

Figure 2. Visual feedback received by students in intervention condition.

Table 1. Probit estimations future study success.

	(1)	(2)	(3)
Test score	0.315***	0.251***	0.304***
	(0.064)	(0.066)	(0.073)
Constant	1.321***	−4.078***	1.127
	(0.155)	(1.231)	(1.161)
Baseline controls	✓	✓	✓
GPA exit exam		✓	✓
Student controls			✓
Pseudo R^2	0.094	0.142	0.193
N	501	501	501

Note. Robust standard errors in parentheses. */**/*** denote significance at a 10/5/1 percent confidence level (two-sided). Baseline controls are application moment and a faculty dummy.

Figure 3. Comparing predictions of Model (1) and (3).

Table 2. Descriptive statistics.

	Experimental group			Applicants 2015
	Range	Mean	SD	Range	Mean	SD
Female	{0,1}	0.61	0.51	{0,1}	0.59	0.51
Age	{17,58}	20.13	4.75	{17,31}	20.08	2.48
GPA secondary school	{5.8,7.9}	6.58	0.38	{5.6,8.3}	6.60	0.43
Faculty of law	{0,1}	0.49	.50	{0,1}	0.57***	.50
Application date	{3,200}	98.14	32.24	{0,223}	100.74	32.34
Model prediction	{31,97}	75.46	13.62	{37,97}	73.68***	10.16
Matching score	{0,10}	6.64	1.75	{0,10}	6.63	1.59
Number of students	313			501

Note. Application date is coded as the number of days since January 1st of that year.

Table 3. Balancing tests.

	Intervention		Control		Nonexperimental
	Mean	SE	Mean	SE	Mean	SE
Female	0.61	(0.04)	0.60	(0.04)	0.58	(0.02)
Age	19.73	(0.26)	20.53	(0.46)	20.31	(0.10)
Faculty of law	0.48	(0.04)	0.49	(0.04)	0.55**	(0.02)
Application date	101.14	(2.46)	97.18	(2.49)	110.43**	(1.99)
Model prediction	75.18	(1.54)	75.73	(1.08)
N	155		158		713

Note. Robust standard errors in parentheses. */**/*** indicate that the mean value differs significantly at the 10/5/1 percent confidence level from the intervention mean.

Figure 4. Self-serving bias: Student versus model predictions. Note. The y-axis represents the density, the x-axis of the top panel represents the difference between the student expectations and the model predictions, and the x-axis of the lower panel represents the student expectations before enrollment.

Figure 5. Prevalence of self-serving bias by prediction groups.

Table 4. Estimation results: Initial enrollment effects.

	(1)	(2)	(3)	(4)
Received information	0.105*	0.129**	0.121**	0.115**
	(0.057)	(0.052)	(0.051)	(0.0510)
Constant	0.449***	0.277*	−0.244*	−0.242
	(0.039)	(0.116)	(0.140)	(0.160)
Prediction model covariates		✓	✓	✓
Student prediction			✓	✓
Student controls				✓
y-control	0.449	0.449	0.449	0.449
R^2	0.011	0.170	0.220	0.249
N	313	313	313	313

Note. Robust standard errors in parentheses. */**/*** denote significance at a 10/5/1 percent confidence level (two-sided). The predication model covariates are the proficiency test score, application moment, and a faculty dummy. The student control variables are gender and age.

Figure 6. Enrollment status and model prediction by experimental group. Note. The y-axis represents the density, the x-axis of the top panel represents the difference between the student expectations and the model predictions, and the x-axis of the lower panel represents the student expectations before enrollment.

Table 5. Initial enrollment effects by level of self-serving bias and gender.

	(4)	(5)	(6)
Received information (RI)	0.115**	0.117**	0.117**
	(0.051)	(0.050)	(0.052)
Self-serving bias⋅RI		0.000
		(0.002)
Female⋅RI			−0.010
			(0.010)
Self-serving bias		−0.007
		(0.005)
Female	−0.095*	−0.107*	−0.050
	(0.053)	(0.055)	(0.075)
Constant	−0.242	0.267*	0.242*
	(0.160)	(0.162)	(0.160)
Prediction model covariates	✓	✓	✓
Student prediction	✓	✓	✓
Student controls	✓	✓	✓
y-control	0.449	0.449	0.449
R^2	0.249	0.249	0.251
N	313	313	313

Note. Robust standard errors in parentheses. */**/*** denote significance at a 10/5/1 percent confidence level (two-sided). The predication model covariates are the proficiency test score, application moment, and a faculty dummy. The student control variable is age.

Table 6. Estimation results: Subsequent enrollment effects.

	(1)	(2)	(3)	(4)
Received information	−0.004	0.026	0.016	0.006
	(0.075)	(0.071)	(0.072)	(0.071)
Constant	0.690***	0.257	0.126	0.050
	(0.055)	(0.205)	(0.326)	(0.349)
Prediction model covariates		✓	✓	✓
Student prediction			✓	✓
Student controls				✓
y-control	0.690	0.690	0.690	0.690
R^2	0.000	0.165	0.174	0.212
N	157	157	157	157

Note. Robust standard errors in parentheses. */**/*** denote significance at a 10/5/1 percent confidence level (two-sided). The predication model covariates are the proficiency test score, application moment, and a faculty dummy. The student control variables are gender and age.

Table 7. Estimation results: Belief updating.

	Respondents				Nonresponse
	12 Credits		≤6 Credits
	Coeff	SE	Coeff	SE	Coeff	SE
Expectation June 2016	87.84	(2.00)	89.38	(1.81)	83.95*	(1.25)
Predicted success June 2016	80.39	(2.28)	77.06	(2.47)	.	.
Expectation November 2016	93.61	(1.32)	79.58	(4.10)	71.71**	(1.41)
Observed success February 2017	100.00		62.50		59**
N	36		16		105
	157

Note. */**/*** denote significant differences with the mean outcome of the respondents at the 10/5/1 percent confidence level.

Table

Netherlands

Scale

Maximum

91–100

81–90

71–80

61–70

51–60

41–50

31–40

21–30

11–20

1–10

Percentage equivalent

96–100

86–95

76–85

66–75

56–65

46–55

36–45

26–35

16–25

6–15

National scale

10

9

8

7

6

5

4

3

2

1

Description

Uitmuntend

Zeer Goed

Goed

Ruim Voldoende

Voldoende

Bijina Voldoende

Onvoldoende

Zeer Onvoldoende

Slecht

Zeer Slecht

Excellent

Very good

Good

Highly satisfactory

Satisfactory

Almost satisfactory

Unsatisfactory

Very unsatisfactory

Poor

Very poor

Note. Reprinted from Pan‐European Grading Scales: Lessons from National Systems and the ECTS, by Terence Karran. Retrieved from http://www.tandfonline.com/doi/abs/10.1080/03797720500087949 Copyright 2004 by Dr. Terence Karran, University of Lincoln, October 20, 2004.