Gaming Over: Quasi-Experimental Evidence on the Effect of Competition on Gaming

Abstract

Does competition increase gaming? Gaming is a well-known problem in the public sector, and a strong logic suggests that increased competition should lead to more gaming. However, issues with reverse causality challenge research aiming to obtain causal estimates of the impact of competition. To deal with this problem, we exploit an exogenously induced change in competition in Danish schools. Following a municipal consolidation reform, some public schools were exposed overnight to increased competition. Using difference-in-differences models to estimate the effect of competition on gaming, we find no evidence of teacher-assessed grade inflation post-reform. Although gaming is widespread, particularly in public schools, increased competition is not bound to exacerbate the problem.

Keywords:

• Competition
• gaming
• public schools

Acknowledgements

The authors would like to thank Simon Calmar Andersen for excellent comments on a very different earlier draft of this paper and for comments at the Midwestern Political Science Association Conference and the SPSA Conference.

Notes

1 About 2% of respondents are unaware of the school type.

2 Since no numbers exist on between-municipality school choice prior to the reform, we cannot entirely exclude the possibility that between-municipality school choice happened more frequently prior to the reform. However, this seems less of a concern bearing in mind the hassles involved for parents and the priority given to students from the home municipality.

3 The description is based on the years 2006 and 2007 that constitute the main years of interest in our empirical analysis. Requirements have since been tightened considerably with the introduction of grade thresholds.

4 Since schools differ in their students’ characteristics, knowledge of external exam results or teacher-assessed grades alone or in combination would not be enough to assess the quality of schools (in terms of their "value added"), or whether there are valid grounds for gaps between both. But the argument is that aside from some underlying quality of schools, families will have at least some incentives to choose schools where their students are likely to obtain better grades for the same level of performance.

5 The possibility of directly investing greater resources in coaching students on the style of the tests, however, remains. In so far as this directs resources from more productive use, this will be a concern, albeit less salient than that of gaming which we evaluate here.

6 As discussed at the beginning of this section, it may be driven by other reasons. That is why we use changes (rather than levels) in competition to identify changes in gaming since these other reasons are unlikely to fluctuate with these particular municipality consolidation events.

7 We chose 10 kilometers because municipalities are required to provide free school transportation for children with more than approximately five kilometers (on average, depending on age) from their residence to school.

8 We use the Ministry’s Institution Register or, alternatively, Kommunalhåndbogen (2005). Addresses from the Institution Register were validated and, in a few cases, corrected based on information from Kommunalhåndbogen (2007) or Kommunalhåndbogen (2005).

9 Google Maps offers distances measured by car, on foot, or by public transportation. We chose walking distance. This data was collected several years after the reform. This may induce some measurement error because of road changes, etc. However, the error is probably unsystematic and negligible, given that we use walking distance.

10 An example calculation appears in Table A3.

11 This minimum detectable effect is calculated taking the standard error of the estimated coefficient of ΔSC × 2007 from column 10 in Table 2 (0.0218, corresponding to the quotient between the displayed coefficient and t-statistic 0.00371/0.17 = 0.0218). At a 5% threshold for statistical significance, this corresponds to a MDE of 0.0218 × 1.96 = 0.043. The relative size is similar for other outcomes. For example, for the summary outcome (the GPA measure), it is .0223 or 2.0 percent of the SD for the outcome.

12 As Figure A2 shows, the competition measure is skewed to the right. As a robustness check, we have log transformed the competition measure (Table A4). The results are substantially similar.

13 The differences between internal and external assessments pre-reform (column 4 of Table 1) are consistent with there being some gaming pre-reform. However, they are not definitive evidence of it. For example, they may be justified by the differences betweenstudents’ differences in performance in a continuous teacher assessment and exam-based external assessments.

Additional information

Notes on contributors

Martin Baekgaard

Martin Baekgaard is a professor at Department of Political Science, Aarhus University. His research interests include administrative burden, performance information use and politic-administrative relations and has been published in journals like Journal of Public Administration Research and Theory and Public Administration Review.

Carlos X. Lastra-Anadón

Carlos X. Lastra-Anadón is an assistant professor at the School of Politics, Economics, and Global Affairs at IE University, Madrid (Spain). His research interests lie at the intersection of political economy and education policy, as well as health and environmental policies. His work has been published in journals such as Political Science Research and Methods, Electoral Studies or Governance.

Søren Serritzlew

Søren Serritzlew is a professor in Political Science at the Department of Political Science at Aarhus University. His research interests currently focus on fiscal federalism, local government, and public policy decision making. His work has been published in journals such as American Political Science Review, American Journal of Political Science, and Journal of Public Administration Research and Theory.

Kim Mannemar Sønderskov

Kim Mannemar Sønderskov is a professor in Political Science and in the Center for the Experimental-Philosophical Study of Discrimination at Aarhus University. His current research interests include political behavior, discrimination, and neighborhood effects, and his work has been published in journals such as American Political Science Review, American Sociological Review, and Nature Human Behavior.