School cheating and social capital

Abstract

In this paper we investigate the relationship between social capital and cheating behaviour in standardized tests. Given the low-stakes nature of these tests, we interpret the widespread presence of cheating as a signal of low trust towards central education authorities and as lack of respect for the rule of law. We find that cheating is negatively correlated to several social capital proxies in the local environment where a school is located. We also distinguish between different dimensions of social capital: contrasting universalistic and particularistic social values, cheating appears to be negatively correlated only to measures of universalistic social values.

Keywords:

• cheating
• social capital
• standardized tests

Acknowledgements

We would like to thank Patrizia Falzetti, Giuseppe Albanese, Claudio Rossetti and Diego Scalise for generous help with the data. We are indebted to Rosario Ballatore, Marco Bertoni, Alessio D'Ignazio, Colin Green, Giovanna Labartino, Pietro Tommasino, Marco Tonello, three anonymous referees and seminar participants at the Bank of Italy, Invalsi, University of Salerno, International Workshop on Applied Economics of Education and University of Roma Tre for valuable comments. At the time this project started and the paper was prepared, Paolo Sestito was also affiliated to Invalsi. The views expressed in this paper are solely those of the authors and do not involve the responsibility of the Bank of Italy or Invalsi. The usual disclaimer applies.

Notes

1 Although Satyanath, Voigtländer, and Voth (2013) show that membership in both ‘bridging’ and ‘bonding’ associations positively predicts entry in the Nazi Party.

2 More details on the estimates of cheating produced by Invalsi are provided in Section 2.1.1.

3 Tests are high-stakes only in the eighth grade, where they contribute to the national lower-secondary school leaving exam, accounting for about one-sixth of the total mark.

4 At most, principals can decide to change the allocation of teachers to classes within the same school.

5 Teachers’ unions fiercely oppose the administration of such tests, arguing that they represent very poor indicators of schools’ or teachers’ performance: therefore, their eventual usage in order to inform personnel decisions and the allocation of resources between schools would lead to severe misallocations of resources and to unjust punishments or rewards. In a way, they were behaving as if the tests were indeed high-stakes.

6 In this respect, our work is similar in spirit to Fisman and Miguel (2007), who show that parking violations by UN diplomats (protected by diplomatic immunity) are strongly related to the degree of corruption in their country of origin. Interestingly, when enforcement authorities acquired the right to confiscate diplomatic licence plates of violators, unpaid violations dropped sharply. This is consistent with the fact that, in classes randomly monitored by an external observer sent by Invalsi, cheating is virtually absent.

7 Actually, the very first use of the term ‘social capital’ can be found in Hanifan (1916), a sociological analysis of the impact of parents’ involvement into schools’ life upon schools’ performance.

8 Bertoni, Brunello, and Rocco (2013) and Lucifora and Tonello (2012) explicitly exploit this ‘natural experiment’.

9 Actually, Invalsi (which since 2012 has started to give back the results to each individual school both gross and net of the cheating estimates) is currently revising its procedure in order to reduce the possible incidence of false positives. The new procedure is iterative. The first step is the standard procedure just described. In the second step, the plausibility of the gross results (in terms of class averages and within-class variances) is verified by comparing them to what one could expect on the basis of a model whose parameters are estimated using the subsample of classes randomly selected to receive external inspectors and by verifying whether the within-class distribution of the gross results is correlated with the within-class distribution of the marks attributed by the teachers. The final cheating estimate is then obtained by combining the results of the two steps.

10 In Italy there are currently 8092 municipalities, 110 provinces and 20 regions.

11 Italy is routinely divided in four macro-regions: North-West, North-East, Centre, South and Islands.

12 We also experimented with fractionalization indices based on father's education and occupation, obtaining similar results.

13 Marks are measured on a scale from 1 to 10, and are usually given at the end of term as a synthetic measure of student performance. Terms may span either a quarter or a semester, depending on the school.

14 Invalsi is actually envisaging to exploit such a fact by moving towards a system in which fifth- and sixth-grade tests are ‘unified’, conducting a test at the very beginning of sixth grade, to get an overall measure of learning accomplishment in primary school. See http://banner.orizzontescuola.it/Rilevazione_apprendimenti_as_2013_2014.pdf.

15 Results for other grades are available upon request and are broadly consistent with what was shown previously.

16 For primary schools, we were able to match 5540 schools, which is approximately 73% of the total.