Halo effects in grading: an experimental approach

Figures & data

Table 1. Results subject A and subject B by manipulation and student gender.

		Subject A				Subject B
	N	M	SD	Min	Max	M	SD	Min	Max
Overall	107	8.7	4.83	0	15	9.13	1.44	5	13
Weak Condition
Paul	17	3.13	2.00	0	9	8.56	1.83	5	11
Paula	19	2.84	0.90	1	4	8.68	1.16	7	12
Overall	35 (32.7%)	2.97	1.49	0	9	8.63	1.48	5	12
Average Condition
Paul	16	8.19	1.38	5	10	8.94	1.18	7	11
Paula	19	8.68	1.11	7	11	9.21	1.55	6	12
Overall	35 (32.7%)	8.46	1.25	5	11	9.09	1.38	6	12
Strong Condition
Paul	17	14.41	0.51	14	15	9.59	1.62	6	12
Paula	20	14.30	0.80	13	15	9.65	1.53	7	13
Overall	37 (34.6%)	14.35	0.68	13	15	9.62	1.55	6	13

Participants graded the vignettes on a 15-point scale (15 = very good).

Table 2. Results subject A and subject B by expert status of the participants.

		Subject A				Subject B
Weak Condition	N	M	SD	Min	Max	M	SD	Min	Max
Novice	27	3.19	1.52	1	9	8.81	1.47	6	12
Expert	8	2.25	1.17	0	4	8.00	1.41	5	9
Overall	35 (32.7%)	2.97	1.49	0	9	8.63	1.48	5	12
Average Condition
Novice	26	8.85	0.93	7	11	9.19	1.45	7	12
Expert	9	7.33	1.41	5	9	8.78	1.30	6	10
Overall	35 (32.7%)	8.46	1.25	5	11	9.09	1.38	6	12
Strong Condition
Novice	27	14.30	0.67	13	15	10.04	1.43	6	13
Expert	10	14.50	0.71	13	15	8.50	1.35	7	11
Overall	37 (34.6%)	14.35	0.68	13	15	9.62	1.55	6	13

Participants graded the vignettes on a 15-point scale (15 = very good).

Table 3. Standardised and unstandardised slope estimates and R2 of the multiple regression analyses predicting the grade in subject B.

		Model 1				Model 2
	B	β	S.E.	p	B	β	S.E.	p
Intercept	9.12	6.04	.42	<.001	9.37	6.21	.43	<.001
A1 (strong)	0.51	.34	.13	<.001	0.51	.35	.11	<.001
A2 (weak)	−0.48	−.32	.13	.01	−0.48	−.32	.12	.01
Gender					0.30	.20	.19	.28
Status					−1.01	−.67	.19	<.001
Gender Stu.					−0.24	−.16	.18	.36
	R² = .07				R² = .16

A1 and A2 contrast variables. Gender participant: female = 0. Status: novice = 0, expert = 1. Gender stud.: gender of the rated student: female = 0. Est. = Estimate. S.E. = Standard error.

Data availability statement

The data used in this study as well as the materials (vignettes in German and an ad-hoc translated version) are open and available online https://osf.io/4dwkc/?view_only=d61c56cf48cf4fcd839130d20dd16712