![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
Suspension from school is a commonly used, yet controversial, school disciplinary measure. This paper uses unique survey data to estimate the impact of suspension on the educational outcomes of those suspended. It finds that while suspension is strongly associated with educational outcomes, the relationship is unlikely to be causal, but rather likely stems from differences in the characteristics of those suspended compared to those not suspended. Moreover, there is no evidence that suspension is associated with larger educational penalties for young people from disadvantaged family backgrounds compared to those from more advantaged family backgrounds. These results hold regardless of whether self-reported suspension or mother-reported suspension is considered. The absence of a clear negative causal impact of suspension on educational outcomes suggests that suspension may continue to play a role in school discipline without harming the educational prospects of those sanctioned.
Acknowledgements
This paper is based on research commissioned by the Australian Government Department of Education, Employment and Workplace Relations (DEEWR) under the Social Policy Research Services Agreement (2010–2012) with the Melbourne Institute of Applied Economic and Social Research. The paper uses data from the Youth in Focus Project which is jointly funded by DEEWR, the Australian Government Department of Families, Housing, Community Services and Indigenous Affairs and the Australian Research Council (Grant Number LP0347164) and carried out by the Australian National University. The views expressed in this paper are those of the authors alone and do not necessarily represent the views of DEEWR, the Australian Government, or the Melbourne Institute. We would like to thank Anastasia Sartbayeva and Chris Ryan for sharing their programming code, Mike Helal for help with data on academic disruptions, and DEEWR staff, the editor and two anonymous referees for helpful comments on previous drafts. Any remaining errors are those of the authors.
Supplemental data
Supplemental data for this article can be accessed in the online version http://dx.doi.org/10.1080/09645292.2014.980398.
Notes
1 One of the underlying arguments for sanctioning disruptive behavior by suspension is the spillover effects that such behavior might have on other students. Estimated average peer effects in educational achievement tend to be small, but there is growing evidence that they may be heterogeneous (e.g. Sacerdote Citation2011). Such differences could influence the extent to which the suspension trade-off facing school principals might vary across different groups of students.
2 See Kinsler (Citation2011) and the references cited therein.
4 The same document reports that the number of short suspensions (1-4 days) was over three times the number of long suspensions, but it does not report on the number of short suspended students.
5 See, for example, http://www.thechronicle.com.au/story/2012/07/19/school-incidents-spark-call-for-more-discipline and http://www.theaustralian.com.au/national-affairs/education/suspend-judgment-keep-kids-at-school/story-fn59nlz9-1226398136400, respectively.
6 At best, this literature examines the association between suspension and achievement matching on or otherwise controlling for a small number of observable characteristics (e.g. gender, race) either at the individual or at the aggregate level (e.g. Rausch and Skiba Citation2005; Arcia Citation2006).
7 In contrast, he argues that suspension does have positive impacts both as a deterrent to misbehavior and on the achievement of other class members via peer effects. The implication is that school principals do not face a trade-off between outcomes for the individual and outcomes for the rest of the class in decisions regarding the use of suspension.
8 In a related study, Karakus et al. (Citation2012) estimate a recursive bivariate probit model to control for direct effects of behavioral problems on employment as well as the indirect effects through endogenous high school graduation. They show that middle-school behavior problems (although not explicitly suspension) impact negatively high school graduation but, conditional on graduation, have no significant impact on adult employment.
9 Nonresponse to particular items in the questionnaire means that sample size is usually lower than these two figures. Sample size is further reduced where we analyze outcomes whose occurrence depends on the occurrence of another outcome.
10 University entrance scores are known by different names in different States and Territories in Australia. Queensland uses a different system called the Overall Position (OP). The OP score ranges between 1 and 25, where 1 is the highest and 25 is the lowest possible score. In all other states, the score ranges from 1 to 100 (highest). We transform the OP score to match the other scores using the conversion factors that university administrators use when comparing Queensland school leavers with those from other states for the purpose of university admission. Scores under 30 are reported as being 30 to the student, so reported scores of under 30 must be data errors and we recode these to missing here.
11 The mean university entrance score is inflated, as scores below 30 are reported as 30.
12 For example, the proportion reporting a suspension is higher in the unmatched sample than in the matched sample. We cannot rule out that there is also selection into the matched sample on relevant unobservables, such as frequency or severity of suspensions among those reporting suspension. If it is the more serious suspension cases in the unmatched sample and/or in the original population that are dropping out of the matched sample, then this could potentially limit the extent to which our conclusions apply to such cases.
13 As a sensitivity check, we also estimated a tobit model in order to account for the fact that university entrance scores are censored at 30 and 100. The results are virtually identical to the OLS estimates as data censoring affects very few respondents.
14 We do not apply a Heckman selection model as it is not feasible to find a valid exclusion restriction that influences one educational outcome (such as high school completion), but not another closely related educational outcome (achieving a university entrance score, given high school completion).
15 Note, however, that standard errors in column 2b are higher than in column 2a, indicating that the smaller sample size results in less precise estimates.
16 We also estimate an IV version of the model where self-reported suspension is instrumented by mother-reported suspension. Although arguably not an ideal instrument, the IV results are very similar to those in with suspension associated with an estimated 19 percentage point reduction in the probability of completion. We thank an anonymous referee for this suggestion.
17 Because two of our three models are non-linear, and because conclusions in the remaining linear model (for university entrance score) are highly sensitive to the value assumed for Rmax, we do not implement the related Oster (Citation2013) approach to infer selection bias by examining coefficient stability and movements in R2 as additional controls are added to the model.
18 In a further robustness check, we redefine ‘disadvantage’ to be having intensive welfare receipt versus no welfare receipt, dropping everyone else from the sample. Once again, the interaction term is small and insignificant for all outcomes.
19 We also use occupational ranking (the ANU4 scale) of the mother as an alternative proxy for socioeconomic disadvantage. The ANU4 scale is a continuous measure developed at the Australian National University (for more details see Jones and McMillan Citation2001). Here, the ANU4 scale is calculated from the current or most recent occupation. Our conclusion remains the same. Results are available upon request.