Day and time patterns in school behaviour
Amy Challen
Centre for Economic Performance, London School of Economics
This draft 04/04/13
Abstract
Using daily data on individual pupils, I find strong day of the week and time of day effects in behaviour incidents in secondary schools. Mondays have the highest rate of incidents, followed by Tuesdays, with substantially lower rates on Wednesdays-Fridays. The first lesson of the day has the lowest rate of incidents, with the rate rising through the day and the highest rate during lesson 5 (the last lesson). I investigate mechanisms, concluding that these patterns are not simply due to selective reporting or misreporting, and are not due to endogenous timetabling. Rather, they are likely to be related to pupils’ subjective wellbeing, particularly tiredness and boredom, and pupil absence and lateness, teacher wellbeing, and teacher absence may also contribute. This paper adds to the literature on time of day and day of the week effects. This is the first paper to show these patterns in schools, and my identification is not subject to many of the usual problems such as endogenous selection into activities or schedules, or variation through the week in activities. Poor behaviour in schools is also a major challenge to the effectiveness of schooling, and I suggest ways of improving behaviour through modifying schedules. Specifically, subjects critical to academic success (such as English and maths) and subjects with higher incident rates, should be scheduled during low-incident time slots. This could reduce the total number of incidents as well as improving attainment in the subjects which make most difference to pupils’ overall attainment.
Introduction
Rich countries have spent several decades increasing the resources provided to state schools, but this has largely failed to produce improvements in educational attainment, even when quality variables are taken into account (Hanushek, 2003; Wössmann, 2003). Public spending on education in the UK increased by over 5% per year in real terms during the 2000s, but is likely to decline substantially to 2014-15 (Chowdry & Sibieta, 2011). Improving the efficiency of time spent at school could therefore be an important means to improving attainment, particularly if this does not involve increased spending. One potential channel is through improving pupil behaviour at schools, which has been a focus of the UK coalition government’s education policy (DfE, 2011). Misbehaving pupils not only affect their own progress, but also impose negative externalities on other pupils, taking up teacher time and reducing the time that can be spent on learning (Lazear, 2001). Imposing discipline is also time consuming and unpleasant for teachers, who may leave schools with poor behaviour. Pupils with poor behaviour may also induce other pupils to behave poorly, further reducing the effectiveness of lesson times.
I present evidence of variation in pupil behaviour in secondary schools by time of day and day of the week. Differences in educational performance are usually analysed in terms of variation between individuals, schools, and countries. When differences within individuals through time are examined this is over long periods, e.g. on the development of cognitive and noncognitive skills over the life course (Borghans, Duckworth, Heckman ter Weel, 2008). Nevertheless, wellbeing and other psychological variables vary over the week and during the course of the day, so it is plausible that educational productivity might vary too. Investigating behaviour incidents in schools is interesting for a number of reasons. First, behaviour incidents are an indirect indicator of educational productivity: disruptive and problematic behaviour in the classroom makes teaching and learning more difficult (Lazear, 2001; Lavy, Passerman, Schlosser, 2012). Moreover, quite aside from any impact pupils’ behaviour might have on their peers, pupils’ tendency to misbehave may be an indication of their own capacity to learn at a point in time, both because this is a direct measure of not being engaged in educational activities, and because less engaging lessons may be causing pupils to misbehave more. We can therefore assume that behaviour incidents will be negatively related to educational productivity at an individual and a class level. Second, behaviour at school is an interesting measure of noncognitive skills. It has been shown to be related to the possession of other noncognitive traits such as self-control (Coolidge, DenBoer Segal, 2004; Petrides et al., 2004). The context and incentives pupils face, and the sort of behaviour valued, are also similar to those in the workplace (Segal, forthcoming; Bowles, Gintis Osborne, 2001; Roberts, Harms, Caspi Moffit, 2007). Thus observed daily and weekly patterns of behaviour within schools may be indicative of patterns also present in workplaces, and may therefore be informative on productivity during the working week. Much of the literature on time of day and day of the week effects relies on factory settings where output can be easily measured, but the school setting might better approximate productivity effects in the sort of office-based or service jobs which dominate developed economies. Third, behaviour can be seen as an indicator of wellbeing: poor behaviour in adolescence is often related to psychological distress (Dodge, 2006). Fourth, behaviour incidents may provide a better measure than is available in much of the literature on productivity over the week, as it is less subject to endogenous selection and observation.
This is the first paper to demonstrate a link between pupils’ behaviour incidents and scheduling patterns. Day of the week effects have been observed in a range of areas of human activity, including labour productivity (Yao, Dresner & Zhu, 2010); injury rates (Card & McCall, 1996); absence from work (Ichino & Moretti, 2009); cardiovascular mortality (Evans et al., 2000; Chenet, McKee, Leon, Shkolnikov, & Vassin, 1998; Willich et al., 1994; Anson & Anson, 2000); stock market returns (Pettengill, 2003); job satisfaction (Taylor, 2006); and wellbeing (Csikszentmihalyi Hunter, 2003; Bryson & MacKerron, 2012; and for a review of the field see Bryson Forth, 2007). These tend to find that Mondays (or the first day of the working week, which varies across countries) are worse on many dimensions: productivity is lower, absence is higher, injury rates are higher, stock market returns are lower, subjective wellbeing is lower, and cardiovascular mortality is higher. Time of day effects have also been observed: injury rates at work are higher at night (Fortson, 2004); while subjective wellbeing is low at the start of the day and rises to reach a peak in the evening (Csikszentmihalyi Hunter, 2003; Stone et al., 2006; Bryson & MacKerron, 2012). However, it is difficult to infer causality from this literature for a number of reasons.
First, many of the surveys of productivity and affect suffer from potential endogeneity of the day and time of the observation, with people able to choose when to respond. For instance, we would expect the days respondents were available for a survey interview to be related to their working patterns (Taylor, 2006). My dataset covers every day when a school is open, so there is little selective reporting of this kind: pupils cannot choose whether their incidents are recorded. Second, many papers investigating time and day effects are unable to disentangle the impact of time or day from the impact of activities typically performed at those times. Activities and conditions of work may vary over the day or week (e.g. there may be a higher workload at the beginning of the week), and controlling for activities typically attenuates the observed time patterns (Stone et al., 2006; Csikszentmihalyi Hunter, 2003). Relatedly, workers or firms may already be organising the working day around employees’ preferences or capacities, which would lead to an underestimate of the impact of time and day on productivity. Workers may also have selected into firms based on their time preferences. By comparison, pupils in my dataset have no control over their timetables; there is no variation in timetable or workload through the school week; and during compulsory schooling pupils have little choice over what subjects they study, who teaches them, and who their classmates are.[1] They are also unlikely to have had much choice over the school they attend.[2] Thus in most cases the only margin on which they can affect their activities is through absenting themselves altogether. This makes the measurement context importantly different from most workplaces, where employees have sorted into that specific employment (and been selected by employers); workload and tasks may differ through the week; and workers may have discretion over the hours they choose to work (even if this is simply a choice amongst shifts). Likewise, once teachers have chosen to work full time or part time and timetables have been set, they will not have much discretion over the lessons they teach. There is plausibly little endogeneity in the timetabling of subjects: schools need to staff every lesson in the week, and block timetabling by yeargroup and department makes strategic manipulation of timetabling difficult. In any case, the patterns I observe hold for each subject considered individually, so it is unlikely that subject scheduling is driving the overall results. We can therefore take the estimates of time and day effects as evidence of the impact of the time slot itself, rather than of the subjects being taught. Third, it may be difficult to compare different times of day in other contexts because of the different durations of activities and timing of breaks. In my data I can compare the rates of behaviour incidents across lessons because I know that pupils are engaging in similar activities in equally long blocks of time. Finally, I have associated data available which allows me to conduct robustness checks and explore the causal mechanisms for the patterns I find. Thus my identification of day of the week and time of day effects may be more robust than those reported in other contexts.
I use a dataset of behaviour incidents in four schools over six academic years to examine the impact of scheduling on the number of behaviour incidents. I find that pupil disruption differs by day of the week and time of day, with Mondays having significantly more behaviour incidents than later days, and the last lesson of the day having significantly more incidents than earlier lessons. There are a number of possible explanations for this pattern, which I investigate with the data available, including a companion dataset on pupils’ psychological state collected for the UK Resilience Programme Evaluation. I consider the following explanations:
1. Pupil wellbeing varies by time of day and day of the week, and this is reflected in differences in behaviour.
2. Conditional on being in school, pupils behave uniformly across the week and the day. However, pupil lateness, absences, and exclusions vary by day of the week, and this drives the impact on behaviour incidents (composition effects).
3. There is no actual difference in behaviour over the week, but teachers’ ability or willingness to report incidents varies by day and time, and this reporting difference generates the patterns I see.
4. Teacher wellbeing has a strong weekly and daily pattern which affects their ability to engage pupils and control classes, and this generates differences in pupil behaviour through the week.
5. Teacher absences have a strong weekly and daily pattern; pupils are more likely to misbehave with supply teachers than with regular staff; so this generates weekly and daily differences in the behaviour incident rate.
6. Schools do not schedule subjects arbitrarily (endogenous timetabling), so the patterns I observe relate to other factors occurring over the school week, e.g. the hardest subjects or the weakest teachers are usually scheduled on Mondays or the last lesson of each day.
Note that these explanations are not mutually exclusive, and several could be contributing to the pattern I observe. I provide evidence that pupils’ wellbeing varies across the week, so it is plausible that their behaviour might vary too. I suggest that it is unlikely that the patterns are entirely due to differences in reporting or endogenous scheduling, but that pupil absence and lateness, teacher and parent wellbeing, and teacher absence may all contribute to the observed pattern. I suggest that pupil behaviour and academic attainment could be improved by reorganising pupils’ schedules. Specifically, the number of incidents could be reduced by exploiting differences in the incident patterns and rates over the week for different subjects. I cannot show that this would increase academic performance, but an improvement in behaviour would be valuable even without an accompanying improvement in attainment. Overall, my findings suggest that modifying the organisation of time within schools could be an important way of improving conditions for staff and pupils and might increase attainment.
Behaviour in English schools
Whether there are gains to be had from improvements in behaviour in English schools depends on whether it is currently a problem. Media reports of school-based behaviour paint a grim picture of rampant dangerous behaviour, but these are probably not representative.[3] A survey of headteachers from G8 countries found classroom disturbance to be the most frequently occurring behaviour problem across countries, with 54% of pupils in England attending schools with weekly or more frequent classroom disruption, and 23% with intimidation or verbal abuse of other students at least weekly (Miller et al., 2009). A survey by a UK teachers’ union found that two-thirds of teachers had had to deal with a violent pupil in the current academic year, with the majority of respondents saying that they thought behaviour had got worse over the previous 5-10 years (ATL, 2012). Two-thirds of secondary school teachers responding to the 2012 Teacher Voice Omnibus survey said that they believed negative pupil behaviour was driving teachers out of the profession (NFER, 2012). However, 84% of secondary schools are rated as having good or outstanding standards of behaviour (Ofsted, 2012), and the most common form of poor behaviour is low-level disruption of lessons, with serious violence very rare and carried out by a small number of pupils (Ofsted, 2005; DfE, 2012a; ATL, 2012). Nevertheless, such disruption can be irritating and can obstruct learning. The impact of poor behaviour on a pupil’s own performance is difficult to measure. Gutman and Vorhaus (2012) use a cohort study which allows them to control for a range of pupil characteristics, finding that pupils who misbehave frequently achieve 178.80 points lower at age 16 then those with no behaviour problems, equivalent to three extra GCSEs at grade A* - a substantial difference, although even rich controls will not account for the endogeneity of misbehaviour and attainment.