Elsevier

Economics of Education Review

Volume 55, December 2016, Pages 70-87
Economics of Education Review

Assessing the effect of school days and absences on test score performance

https://doi.org/10.1016/j.econedurev.2016.08.007Get rights and content

Highlights

  • Examines reduction of absences and extending the school year on test scores.

  • Uses student-level data to estimate three-way high-dimensional fixed effects model.

  • Reducing absences has a larger effect than a similar addition of school days.

Abstract

While instructional time is viewed as crucial to learning, little is known about the effectiveness of reducing absences relative to increasing the number of school days. Using administrative data from North Carolina public schools, this paper jointly estimates the effect of absences and length of the school calendar on test score performance. We exploit a state policy that provides variation in the number of school days prior to standardized testing and find substantial differences between these two effects. Extending the school calendar by ten days increases math and reading test scores by only 1.7% and 0.8% of a standard deviation, respectively. A similar reduction in absences would lead to gains of 5.5% in math and 2.9% in reading. We perform a number of robustness checks including utilizing flu data to instrument for absences, family-year fixed effects, distinguishing between excused and unexcused absences, and controlling for a contemporaneous measure of student disengagement. Our results are robust to these alternative specifications. In addition, our findings indicate considerable heterogeneity across student ability, suggesting that targeting absenteeism among low performing students could aid in narrowing current gaps in performance.

Introduction

During the last decade, the U.S. federal government and many states have taken a series of steps to improve educational outcomes in elementary, middle, and high school. To this end, many programs have been implemented1 whose primary aim is to hold schools accountable for the performance of their children. More recently, policy makers have renewed their attention towards2 the actual number of days that students spend at school. For example, while the federal government is aiming for an extension of the school calendar,3 many states and cities have already increased the number of school days.4 Despite these initiatives, little is known about the effectiveness of this type of intervention relative to other possible competing policies. For instance, reducing absenteeism may constitute an alternative avenue for policy as it would target specific students who would benefit the most from being in the classroom. Indeed, there is significant room for improving student attendance; Balfanz and Byrnes (2012) estimate that 5–7.5 million students are missing nearly a month of school. Absences are a pervasive problem present at all schooling levels, even the initial ones. For example, at least 10% of kindergartners and first graders miss 10% or more of the school year (see, Chang and Romero, 2008). The impact of these absences -even those in early grades- can have long term effects; chronic absenteeism constitutes an early predictor of dropping out of school (see, Romero, Lee, 2007, Connolly, Olson, 2012, Attendance Works & Healthy Schools Campaign). In addition, absences are costly for schools’ and districts’ budgets. For instance, in California, absences cost public schools $3.5 billion in state funding based on daily attendance between 2010/11 and 2012/13 (see, Harris, 2014). Given these large costs, recent initiatives have been designed to reduce chronic absenteeism such as “NYC Success Mentor Corps”5 and “WakeUp! NYC”6 which were recently launched in New York City.7

In this regard, the goal of this paper is to quantify the relative effectiveness of reducing absences with respect to extending the school calendar on test score performance based on a sample of elementary school students. While most studies have analyzed the importance of absences or days of class separately,8 this analysis provides an approach that allows for both effects to be examined simultaneously. We believe that, from a policy perspective this is key, given that extending the school year or reducing absences are likely to affect students at different margins. For example, missing a day of school due to an absence may be more detrimental to a student’s performance since they will need to make up for missed work later. Moreover, catching up is likely to be more difficult for low performing students, resulting in larger gaps in academic performance within the classroom. To examine possible heterogeneous effects of absences and days of class, we analyze whether children from relatively low income families, or those who perform poorly, benefit comparatively more from spending increased time at school. Similarly, we try to identify whether the loss of a school day has differential effects depending on the school grade, and whether absences’ effects persist in subsequent school years. We believe that providing a detailed analysis of heterogeneous effects will inform the policy discussion in terms of identifying specific groups of the population that may benefit the most from particular interventions. Finally, we investigate the effect of teacher and school quality on absences. We study to what extent attending (having) a better school (teacher) could lead to a decrease in the number of days absent.

Contrary to most of the literature that has considered countries, states, counties, or schools as the unit of analysis,9 we make use of detailed longitudinal data at the individual level from North Carolina public schools. This allows us to control for students’, teachers’, and schools’ observable and unobservable characteristics. Therefore, this paper is able to analyze the importance of time spent at school from several perspectives (i.e. absences and days of class), as well as to implement a rigorous econometric strategy to address problems of endogeneity in a number of ways. In order to deal with the various threats to identification, such as health shocks, disengagement effects, and omitted variable bias, we employ several different identification strategies. First, we use previous year test scores, student, teacher and school fixed effects, and measures of overall school performance to control for heterogeneity. Second, we control for a contemporaneous measure of student disengagement (i.e. school suspensions). Third, we utilize flu data at the county level to instrument for absences. Fourth, we employ family-year fixed effects to account for any time-varying family specific shocks. Finally, we examine unexcused absences to take into account any illnesses or other excused events that may affect both absences and grades.10 Reassuringly, our results are consistent across specifications.

Estimating models with three-way high-dimensional fixed effects when the sample size is large is not a trivial matter. In our case, it requires to estimate an extremely large number of parameters11 (i.e. 701,166 students; 33,051 teachers; and 1364 schools), therefore an iterative algorithm is implemented in order to overcome computational issues.

Results show substantial differences between the effect of absences and days of class on test score performance. Our preferred specification indicates that extending the school calendar by ten days would increase math and reading test scores by 1.7% and an insignificant 0.8% of a standard deviation, respectively,12 while a similar reduction in absences would lead to an increase of 5.5% in math and of 2.9% in reading scores.13 Estimation results show that absences have an even larger negative effect among low performing kids, suggesting that the costs of catching up are higher among those who show greater difficulties at school. In addition, we analyze whether spending more time at school (i.e. fewer absences or a longer academic calendar) has a larger effect on students in later grades. While being ten days absent in grade 3 leads to a decrease of 2% of a standard deviation in math test scores, in grade 5 the effect is 8.1%. Moreover, we present evidence indicating a persistent effect of absences in subsequent grades, particularly in math. Finally, we show that attending (having) a school (teacher) in the 75th percentile of the fixed effect distribution decreases absences by 0.14 (0.14) days relative to the 25th percentile; a relatively large result given that the average number of absences is 6.28 days.14 Overall, the results point towards the presence of an important asymmetry between the effects of expanding total time spent at school through a reduction of absences or through an extension of the school calendar. However, we should emphasize that expanding or missing instructional time in elementary school are likely to have different implications than, for example, in college. Therefore, these findings should be circumscribed to students that are attending the first years of their schooling careers.

The financial resources required to extend the school calendar are large due to the high fixed cost of operating a school.15 This fact combined with our findings showing a larger effect of absences than days of class on test score performance are likely to have important policy implications. While policymakers in many countries are discussing changes to the school calendar,16 making better use of the existing time schemes by reducing absences constitutes an avenue for policy that deserves greater attention. A possible (inexpensive) example of an intervention that reduces absences is the Education Act of 1996 in the United Kingdom. This policy empowers head teachers to issue Penalty Notices for unauthorized absences from school. This means that when a pupil has five or more days of unauthorized absences in any term (where no acceptable reason has been given for the absence), or if their child persistently arrives late for school after the close of registration, their parents or guardians may receive a Penalty Notice of £60 if paid within 21 days, rising to £120 if paid within 28 days. A report on the effectiveness of these fines (Crowther and Kendall, 2010), found that 79% of local authorities said penalty notices were “very successful” or “fairly successful” in improving school attendance.17 Finally, an alternative path to reduce absenteeism has been suggested by Cuffe, Bignell, and Waddell (2014), who show that increasing participation in extracurricular activities (e.g. athletic competitions) leads to a reduction in unexcused absences for high schoolers.18

The remainder of this paper is organized as follows. Section 2 places our work in context with the related literatures on student absences and school length. Section 3 details the data used in the empirical analysis. Section 4 outlines the econometric strategy. Section 5 describes the results. Section 6 presents a series of robustness checks. Section 7 examines the heterogeneous effects of absences and days of class by several student characteristics. Section 8 concludes.

Section snippets

Background

The length of the school year and a student’s total number of absences combine to determine the total amount of instructional time a student receives in a given year. Despite this, their effects on student performance have largely been examined independently; likely due to the lack of available data on absences and the limited variability on school year length.

Data

The North Carolina education data is a rich, longitudinal, administrative data set that links information on students, teachers, and public schools over time. This database is maintained by the North Carolina Education Research Data Center (NCERDC), which is housed at Duke University.

Students in North Carolina have to take (standardized) exams in math and reading at the end of the school year, with the aim to assess whether they have met grade-level expectations.30

Methodology

The data enables us to observe the EOG test scores, the number of class days, and the absences of students in each year for grades 3 to 5 until the day of the test. Our primary aim is to estimate the causal effect of both an absence and an additional day of instruction on performance. The number of instructional days prior to the exam varies across schools and years and therefore enables the identification of the effect of absences separately from additional instructional time.

In analyzing the

Baseline results

Table 3 presents the regression results for math and reading, based on Eq. (1). Specification (1) is a simple OLS regression of standardized test scores52 without any fixed effects or controls for student ability. The coefficients on absences for both math and reading are negative, significant and large in magnitude. However, since there are no controls for unobserved individual characteristics, which are in all probability negatively correlated

Robustness checks

Despite the set of controls that have been included in Table 3 (i.e. student, teacher, and school fixed effects, previous year test score, school lagged performance, peer absences, and free/reduced price lunch status), our results may still be driven by confounding effects. For example, student engagement may not be fixed over time or family/health shocks could affect absences and test score performance in a way that may not be captured by our extensive set of controls. In this section we

Heterogeneous effects

On average, absences have a negative effect on test scores, while the positive impact of an additional day of class within the observed range is much smaller. However, these effects may differ based on student characteristics. As noted earlier, catching up after an absence is likely to be more difficult for a low performing student. Understanding the heterogeneous effects of an absence will help to inform the policy discussion by identifying groups of the population that are likely to

Conclusions

This paper jointly estimates the relative effectiveness of reducing absences to extending the school calendar on test score performance. Despite the fact that many policy makers have focused their attention on extending the school calendar, the evidence presented in this manuscript indicates that targeting absenteeism could constitute a more effective intervention. First, our empirical strategy shows that the effect of reducing absences relative to extending the number of school days is

References (44)

  • G. Caetano et al.

    Quasi-experimental evidence of school choice through residential sorting

    (2013)
  • S.J. Caldas

    Reexamination of input and process factor effects on public school achievement

    The Journal of Education Research

    (1993)
  • D. Card et al.

    Does school quality matter? Returns to education and the characteristics of public schools in the United States

    Journal of Political Economy

    (1992)
  • Chalkboard Project (2008). A review of research on extended learning time in K-12 schools. Online....
  • H.N. Chang et al.

    Present, engaged, and accounted for: The critical importance of addressing chronic absence in the early grades

    Technical Report

    (2008)
  • R. Chetty et al.

    Measuring the impacts of teachers I: Evaluating bias in teacher value-added estimates

    American Economic Review

    (2014)
  • F. Connolly et al.

    Early elementary performance and attendance in Baltimore City Schools? Pre- kindergarten and kindergarten

    Technical Report

    (2012)
  • K. Crowther et al.

    Investigating the use of parental responsibility measures for school attendance and behaviour: Final report

    Research Report, DFE-RR041

    (2010)
  • H. Cuffe et al.

    Too busy for school? The effect of athletic participation on absenteeism

    Technical Report, 8426

    (2014)
  • J. Currie et al.

    Early test scores, school quality and socioeconomic status: Long run effects on wage and employment outcomes

  • S. Devadoss et al.

    Evaluation of factors influencing student class attendance and performance

    American Journal of Agricultural Economics

    (1996)
  • C. Dobkins et al.

    Skipping class in college and exam performance: Evidence from a regression discontinuity classroom experiment

    Economics of Education Review

    (2010)
  • Cited by (130)

    • Residential sorting, local environments, and human capital

      2024, Regional Science and Urban Economics
    View all citing articles on Scopus

    We thank Peter Arcidiacono, Marta De Philippis, Marc Gurgand, Joseph Hotz, Monica Langella, Hugh Macartney, Alan Manning, Marjorie McElroy, Guy Michaels, Steve Pischke, Seth Sanders, Douglas Staiger, and participants at the CEP and Tibergen Institute labor seminars, Duke Labor Lunch Group, the applied micro seminars at University of Bristol and University of Oslo, and participants of the NBER education meeting. We are grateful to the Center for Child and Family Policy for access to the NCDPI dataset and to North Carolina Division of Public Health for the NC DETECT data. The NC DETECT Data Oversight Committee does not take responsibility for the scientific validity or accuracy of methodology, results, statistical analysis, or conclusions presented. All errors and omissions are our own.

    View full text