‘Bad Apple’ peer effects in elementary classrooms: the case of corporal punishment in the home

ABSTRACT

This paper provides the first empirical evidence on the existence of negative spillover effects from children exposed to corporal punishment in the home (CPH). We find that interactions with peers suffering from CPH depress achievement in both math and language among Vietnamese fifth graders. Specifically, a one standard deviation increase in the Peers' Violence Index is associated with a reduction in the math and the language test scores by 0.11 and 0.14 standard deviations, respectively. These adverse impacts could potentially be attributed to the unfavorable changes in student academic aspirations, student actual learning efforts, and the inter-student relationships.

KEYWORDS:

• Corporal punishment
• student achievement
• peer effects

JEL CODES:

• I20
• I21
• J18

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 These provinces include Ben Tre, Da Nang, Hung Yen, Lao Cai, and Phu Yen.

2 In the original questionnaire, the possible responses to the question on the method of allocating student to classrooms are ability in math, ability in Vietnamese, general ability, age, residency, ethnicity, and random allocation.

3 On average, there are 2.2 fifth-grade classes per school. There are 26 schools with one class (26), 14 schools with two classes (28), 13 schools with three classes (39), three schools with four classes (12), one school with five classes (5), two schools with six classes (12), and one school with eight classes (8).

4 All test scores are standardized. The average presented in Table 1 is not equal to zero because our sample only consists of classrooms where students were randomly assigned. Re-standardizing these test scores does not change our results.

5 We also construct other peer measures of CPH: Peers' Violence-Original Response and Fraction Exposed to Violence. Peers' Violence-Original Response is the peer measure constructed from taking the average of Own Violence-Original Response, excluding the student himself/herself. Fraction Exposed to Violence, the peer measure of Exposed to Violence, is the proportion of a student's peers whomever experience corporal punishment by parents, excluding the student himself/herself.

6 We plot the distribution of the Peers' Violence-Original Response along with the residualized Peers' Violence-Original Response in Figures A1 and A2 in Appendix A. Figures A1 and A2 confirm that there is enough variation in the peers' exposure to CPH, regardless of the way we construct it.

7 These four variables are 0/1 indicators.

8 Teacher Education is a dummy variable taking the value of one if the teacher obtains at least a four-year university degree and zero otherwise. Teacher Qualification is an indicator taking the value of one if the teacher earns any qualifications at a university and zero otherwise. Excellent Teacher Award is a dummy that equals one if the teacher receives the ‘Excellent Teacher’ Award granted by the provincial administration (the highest level) and zero otherwise. Teacher Experience is the number of years working as a teacher.

9 This is also known as the left-hand side balancing test (Pei, Pischke, and Schwandt 2019).

10 Own Violence Index is constructed as described in Section 2.1. Grade Repetition is an indicator taking the value of one if the student repeats any grade and zero otherwise. Student gender is represented by a Being Female dummy that equals one if the student is female and zero otherwise. Being Minority is an indicator that takes the value of one if the student belongs to an ethnic minority group. Parental education is reflected by two dummies respectively representing whether the mother and the father have a college degree. We also include in $S{C}_{ics}$ three measures of student baseline characteristics obtained from the first wave of the data: Level of family support, academic ability, and motivation to succeed at school. The responses range from 1-very high to 5-very low.

11 Female Teacher is a dummy indicating if the teacher is female. Teacher Education is a dummy variable taking the value of one if the teacher obtains at least a four-year university degree and zero otherwise. Teacher Qualification is an indicator taking the value of one if the teacher earns qualifications at a university and zero otherwise. Excellent Teacher Award is a dummy that equals one if the teacher receives the ‘Excellent Teacher’ Award granted by the provincial administration (the highest level) and zero otherwise. Teacher Experience enters the regressions as various indicators representing different year ranges to account for the nonlinear returns of teacher experience (Ost 2014).

12 Average peer achievement in math (language) is the class-level mean of the beginning-of-year test scores in math (language), excluding student i. Peer gender composition is measured by the fraction of female students in a class, excluding student i. In constructing peer parental education, we calculate the mean of the two indicators, Mother Has College Degree and Father Has College Degree for each student, then take the class-level average of that measure excluding that of student i.

13 We replicate Column 1 and 5 of Table 5 but replace the Peers' Violence Index with the average peer baseline achievement, the fraction of female students and peer parental education as explanatory variables in Column 1 through 3 and 5 through 7 of Table A4. Unlike the Peers' Violence Index, we do not find consistent and statistical evidence on the impacts of these peer measures.

14 We also estimate the effects of CPH-inflicted peers using the same specification as our baseline model on different samples: (i) schools where all classrooms have students randomly assigned, (ii) schools where students are not randomly allocated, and (iii) schools with only one fifth-grade classroom. We still observe the negative spillover effects of CPH (Table A6).