Teachers’ perceptions of students’ school performance: the impact of classroom composition. Evidence from a survey experiment

Figures & data

Figure 1. Violin plot indicating the connection between the share of poor students and Roma students within the schools:school principals’ estimates.

Note: Data: NABC 2021, 8th grader sample, own calculation.

Table 1. The characteristics of sample schools, relative to all Hungarian primary schools.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
	Rich students^(a)	Poor students^(b)	Roma students	Math test score, 8^th grade	Reading test score, 8^th grade	SES	Village	Town	County seed	Budapest (capital city)	Very small school^(c)	Small school	Medium-sized school	Large school
Mean in school, not in the sample	0.08	0.18	0.20	0.00	0.00	0.01	0.40	0.33	0.15	0.13	0.02	0.31	0.33	0.35
Mean difference	−0.01	0.03*	0.07**	−0.19**	−0.21**	−0.32**	0.16**	−0.07**	−0.05**	−0.03+	−0.00	0.06*	0.02	−0.08**
	(0.01)	(0.01)	(0.02)	(0.06)	(0.06)	(0.08)	(0.03)	(0.03)	(0.02)	(0.02)	(0.01)	(0.03)	(0.03)	(0.03)
Standardized mean difference	−0.07	0.16	0.25	−0.19	−0.21	−0.33	0.33	−0.16	−0.16	−0.10	−0.02	0.12	0.05	−0.17
N of schools	2,530	2,531	2,522	2,702	2,702	1,676	2,812	2,812	2,812	2,812	2,812	2,812	2,812	2,812

Robust standard errors, clustered at the school level, in parentheses.

**p < 0.01, *p < 0.05, + p < 0.1

Own calculation using data from the Hungarian National Assessment of Basic Competencies (NABC); the cohort of 8^th graders in 2021.

The NABC assesses students’ mathematics and reading comprehension achievement with a PISA-like test. Mathematics and reading comprehension achievement are assessed by two 90-minute long in-class tests supervised by the classroom teachers and marked centrally. The NABC has a take-home supplementary questionnaire regarding students’ social backgrounds and home environments. The NABC also has a school questionnaire for the school principal, with questions about the school context.

The source of data in Columns 1-3 and 7–14 is the NABC school questionnaire, answered by the school principal.

The source of data in Columns 4 and 5 is students’ individual test scores, aggregated to the school level.

The source of data in Column 6 is individual students’ background data from the NABC background questionnaire, aggregated to the school level.

^(a) The translated version of the original Hungarian question asked principals to rate the share of students in the school who had ‘above average material circumstances’ (átlag feletti anyagi körülmények között élő).

^(b) The translated version of the original Hungarian question asked principals to rate the share of students in the school who ‘are living under extraordinarily bad material circumstances’ (nagyon rossz anyagi körülmények között élő).

^(c) A very small school is a school with a maximum of five teachers (as defined by the NABC).

Standardized mean differences are expressed in standard deviation units.

Figure 2. Sample vignettes to rate (non)-poor and (non)-Roma students.

Note: Words that appear in the [squared brackets] are the alternative word used in other vignettes.

Figure 3. Teacher’s perceptions of students’ school performance.

Figure 4. Teachers’ perceptions of (non)-poor and (non)-Roma students’ average school performance under different classroom compositions.

Table 2. Compositional effects in students’ teacher-perceived school performance.

	(1)	(2)
	Direct compositional effect Teacher-perceived school performance of disadvantaged students	Indirect compositional effect Teacher-perceived school performance of non-disadvantaged students
Roma students [${\beta }_2$]	−0.07	0.03
	(0.17)	(0.15)
Social-status-based compositional effect [${\beta }_3$]	−0.03	−0.07
	(0.18)	(0.16)
Ethnicity-based compositionaleffect [${\beta }_3$+${\beta }_4$]	−0.37**	−0.51**
	(0.15)	(0.17)
Difference between Roma/poor compositional effects [${\beta }_4$]	−0.35	−0.44*
	(0.25)	(0.22)
N of teachers	562	562

Robust standard errors (clustered at the school level) in parentheses

**p < 0.01, *p < 0.05, + p < 0.1

Note: coefficients in the table are estimated with Equation (1).

All models contain a constant and the following control variables: teachers’ gender, teachers’ age, dummy variables signaling the grade in which the teacher is a homeroom teacher, and fixed effects for the taught subjects. Models also contain school fixed-effects.