How do ability peer effects operate? Evidence on one transmission channel

ABSTRACT

Many (quasi-)experimental studies show that students tend to learn more in classes with better peers. However, the (presumably numerous) factors mediating the positive relationship between peer and own achievement have received less attention in the literature. I present evidence on one particular transmission channel: teachers tend to grade tougher in classes with better students, inducing parents to send their children to remedial tutorial lessons which have a positive effect on student outcomes. This study provides an example for negative peer ability externalities that are (over-)compensated by parental educational investments.

KEYWORDS:

• Ability peer effects
• transmission channels
• educational investments

JEL:

• I21

Acknowledgments

I want to thank Alejandro Cid and two anonymous referees for excellent comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author.

ORCID

David Kiss http://orcid.org/0000-0002-2475-8532

Notes

1 Empirical assessments of ability peer effects use several methods to minimize biases due to self-selection. Some researchers employ fixed-effects frameworks (e.g., Hanushek et al. 2003; Ding and Lehrer 2007; Burke and Sass 2013). Others analyze data from controlled randomized trials (Boozer and Cacciola 2001; Duflo, Dupas, and Kremer 2011) or exploit natural experiments (Ammermüller and Pischke 2009; Sund 2009; Imberman, Kugler, and Sacerdote 2012; Kiss 2013). These studies suggests that, on average, a one standard deviation increase in peer achievement raises own achievement by 0.1 standard deviations within a year. Throughout, (i) the terms (academic) achievement, skills, and (test) scores are used interchangeably, and (ii) ‘peer’ is simply a synonym for ‘classmate’.

2 The seminal paper by Black (1999) shows that houses are more expensive if they are located in catchment areas of higher-quality schools. As emphasized by Bayer, Ferreira, and McMillan (2007) and Ries and Somerville (2010), though, it is less clear whether better schools increase the valuation of houses granting access to them, or simply results from spatial concentration of households with both high incomes and favorable socioeconomic backgrounds. Taken together, these findings point towards the direction that parents presumably value both, good neighborhoods and high-quality schools.

3 In this context, Cicala, Fryer, and Spenkuch (2016) and Elsner and Isphording (2017) show that ordinal rank losses can lead to behavioral problems and lower high school graduation rates. Bonesrønning (2004) and Figlio and Lucas (2004) further find that tougher grading practices do have a positive effect on student achievement – this paper provides one explanation for their findings.

4 Simple reduced form estimates provided by Altonji (1995) indicate that returns to additional math classes are higher than those from other subjects like languages or social studies. Betts and Rose (2004) complement this finding by showing that returns to math courses are positively related to their level of difficulty. Findings by Joensen and Nielsen (2009) support the notion that improvements in labor market outcomes are actually caused by math classes.

5 See Bela, Pink, and Skopek (2012) for details. Blossfeld, Rossbach, and von Maurice (2011) provide a general introduction to the NEPS.

6 Let i,j index students, c classes, and let denote both class size and the set of student IDs that constitute class c. Then, the peer math achievement level student i is exposed to equals , with denoting the math test score of classmate j.

7 Elementary school usually takes four years in Germany. Upon completion, students are then tracked by ability into various types of secondary schools. The most common secondary schools are upper-, middle-, and lower-secondary school (Gymnasium, Realschule, and Hauptschule). The highest educational degree of the parents is classified according to the International Standard Classification of Education (ISCED) guidelines, see OECD (1999) for details.

8 However, parents were asked in every assessment about their child's math mark and tutorial decisions.

9 As will be shown in detail in Section 4, the results from Table 4 are replicated in Table 8 with matched data where two students are considered as ‘statistical twins’ if (i) their math scores are similar while (ii) they are enrolled in different classes.

10 In addition to that, because (i) the data are cross-sectional, and (ii) end-of-year marks denote the dependent variable, the composition of classes should be relatively stable over time. To assess the fluctuation in classes, I compute the share of fifth-graders who changed their class after one year. In 60% of cases, the composition of classes remained unchanged after a year. In 75% (90%) of classes, the fluctuation was less than 10% (25%).

11 Hof (2014), Martorell and McFarlin (2011), and Zhang (2013) contain detailed literature reviews on both remedial and self-initiated tutoring.

Additional information

Funding

This project is funded by the Deutsche Forschungsgemeinschaft (German Research Foundation (DFG)), Priority Program SPP1646, which is gratefully acknowledged. It employs data from the Progress in International Reading Literacy Study (PIRLS) and the National Educational Panel Study (10.5157/NEPS:SC3:3.1.0 and 10.5157/NEPS:SC4:6.0.0). The PIRLS data were produced by the Institut für Schulentwicklungsforschung (IFS) Dortmund and provided to me by the research data center of the Institute for Educational Quality Improvement (IQB) Berlin. NEPS data were collected as part of the Framework Program for the Promotion of Empirical Educational Research funded by the German Federal Ministry of Education and Research (BMBF). Since 2014, NEPS is carried out by the Leibniz Institute for Educational Trajectories (LIfBi) at the University of Bamberg in cooperation with a nationwide network.