Examining the effect of social influence on student performance through network autocorrelation models

Abstract

The paper investigates the link between student relations and their performances at university. A social influence mechanism is hypothesized as individuals adjusting their own behaviors to those of others with whom they are connected. This contribution explores the effect of peers on a real network formed by a cohort of students enrolled at a graduate level in an Italian University. Specifically, by adopting a network effects model, the relation between interpersonal networks and university performance is evaluated assuming that student performance is related to the performance of the other students belonging to the same group. By controlling for individual covariates, the network results show informal contacts, based on mutual interests and goals, are related to performance, while formal groups formed temporarily by the instructor have no such effect.

Keywords:

• interpersonal network
• network autocorrelation
• network effects model
• social influence
• student academic performance

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. In analogy with economic literature, Dow [19] discussed an alternative estimation solution to deal with the presence of the endogenous term in network autocorrelation models framework based on instrumental variables and 2SLS estimation method. However, the choice of instruments is a task that can be hardly faced in studying peer effects [37], and it goes beyond the scope of this contribution.

2. Worthy of note is the recent 2012 feasibility study on the measure of student performance carried out by the OECD ‘Assessment of Higher Education Learning Outcomes (AHELO)’ project. For details visit OECD AHELO project website: www.oecd.org/edu/ahelo.

3. ‘Statements called intended learning outcomes, commonly shortened to learning outcomes, are used to express what it is expected that students should be able to do at the end of the learning period’ [27, p. 3].

4. The Dublin descriptors comprise ‘generic statements of typical expectations of achievements and abilities associated with qualifications that represent the end of each Bologna cycle’ [5, p. 65].

5. A first model was estimated to predict the graduate average grade of five students from their undergraduate grade. The model was also used to obtain values for five missing undergraduate grades (an inverse regression problem). The same procedure was adopted for the Dublin Descriptors, where each graduate level descriptor was regressed on the corresponding undergraduate one. In such a case, a total of six missing values were imputed using 10 variables.

6. We computed the Pearson correlation for all pairs of the five networks considered, and assessed the frequency of random measures as large as actually observed by using the dyadic QAP-correlation tool implemented in the UCINET software.

7. Several methods are available to obtain individual scores for latent variables, generally providing different results for each unit. We adopted the revised blockwise factor score regression procedure [42], that is based on the estimation of individual scores separately for the dependent (Bartlett scores) and independent (Regression scores) latent variables. It produces consistent estimators for all parameters in the case of a latent regression model.

8. This is probably due to the large number of parameters (if compared with the small sample size) that in turn provided large standard errors.