Scheduled to Gain: Short‐ and Longer‐Run Educational Effects of Examination Scheduling*

• Date: 01 July 2020
• DOI: http://doi.org/10.1111/sjoe.12363
• Author: Simon Søbstad Bensnes
• Published date: 01 July 2020

Scand. J. of Economics 122(3), 879–910, 2020

DOI: 10.1111/sjoe.12363

Scheduled to Gain: Short- and Longer-Run

Educational Effects of Examination

Scheduling*

Simon Søbstad Bensnes†

Norwegian University of Science and Technology, NO-0131 Oslo, Norway

simon.bensnes@ssb.no

Abstract

In this paper, I present findings concerning the effect of examination scheduling on high-stakes

exam scores and longer-run outcomes. I show that random variations in examination schedules,

which increase the time students haveto prepare, have positive effects on exam scores. The effect

is highly concave, and stronger for females and in quantitative subjects. I trace the effects of

preparation time into tertiary education outcomes, finding significant effects for female students

on the extensiveand intensive margins. I show how easily exam scores and,consequently, longer-

run outcomes are affected by a random institutional factor unrelated to student ability.

Keywords: Examination schedules; exam scores; preparation time; upper secondary

JEL classification:I20; I21

I. Introduction

Each spring, millions of students across the world take high-stakes

examinations at the end of upper secondary schooling. Examples include

the General Certificate of Secondary Education in the United Kingdom,

the Baigrut in Israel, and the SAT and Advanced Placement exam in the

United States. The scores that students receive on these tests are then used

to sort students into jobs and tertiary education. When exam scores are

utilized in this manner, the underlying assumption is that exam scores

are reliable proxies for student ability. However, recent evidence indicates

that random disturbances during exams, such as ambient air pollution and

pollen proliferation on the day of the exam, have relatively large effects

on both exam scores and longer-run outcomes (Lavy et al., 2015; Bensnes,

2016). Furthermore, institutional factors can also affect exam conditions

*I am grateful for input from Chris VanKlaveren, Bjarne Strøm, TorbergFalch, Magne Mogstad,

Sarah Bana, Chang Lee, and participants at various seminars and workshops, and for thorough

comments from three anonymous referees. I am also grateful for institutional insights from

Øyvind Kvanmo. Any remaining errors are myown.

†Also affiliated with Statistics Norway.

C

The editors of The Scandinavian Journal of Economics 2019.

880 Scheduled to gain

(e.g., Pope and Fillmore, 2015). These findings show that: (i) the accuracy

with which exams reflect students’ abilities depends in part on how sensitive

exam grades are to random variations in exam conditions; and (ii) the more

sensitive this relationship is, the less useful exams are as a placement tool.

In this paper, I focus on an institutional factor that is unrelated to cognitive

ability but affects exam scores: examination scheduling. The aim of this

paper is, first, to estimate the effect of random variations in examination

scheduling on exam scores, and then to trace this effect to longer-run

outcomes.

Identification in this paper is achieved by using a specific feature of

the national examination system for Norwegian upper secondary schools,

combined with very detailed administrative data. Each year, on a specific

date shortly before their exams, students in the Norwegian upper secondary

school system are informed when their written end-of-year exams will

be held and which subjects they will be examined in. The time between

the date of the announcement and the date of each exam serves as the

measure of preparation time. The identification used stems from subject-

specific random variation in the examination schedules across years. Results

suggest that increasing the preparation time from 5–8 days to 9–12 days

increases test scores by 5.3 percent of a standard deviation. Interestingly, the

marginal return to preparation time approaches zero when preparation time

is increased beyond this point. Overall, the results indicate that students

only study for a limited number of days, regardless of the amount of

preparation time they are granted. Furthermore, the effect is stronger in

quantitative subjects and for females, who are also more strongly affected

in the longer run: an increase in the share of exams with relatively long

preparation periods by one standard deviation increases the probability of

a female upper secondary school student enrolling in a tertiary education

program by half a percentage point.

While, in this paper, I use the direct variation in the conditions

associated with exams, the underlying cause of this variation stems from

the institutional framework rather than from environmental factors, as is

the case in Lavy et al. (2015), for example. The effects of institutional

factors on students’ outcomes, per se, is not an unstudied topic. Areas that

have received some attention include school start times (e.g., Carrell et al.,

2011; Edwards, 2012) and course scheduling (e.g., Dills and Hernandez-

Julian, 2008). However, most studies are confined to measuring short-run

outcomes. An exception to this restriction is the literature on school starting

age. Black et al. (2011) show that starting school later leads to better test

scores, but that these effects are driven by age. They also find that starting

school later leads to better mental health at age 18, but lower earnings at age

30. Landersø et al. (2017) find that starting school later causes students to

score better at exams and to study in fields with higher entry requirements.

C

The editors of The Scandinavian Journal of Economics 2019.

S. S. Bensnes 881

Fredriksson and ¨

Ockert (2014) find that starting school later negatively

affects life-cycle earnings.1As with school starting age, the institutional

framework around exit exams can randomly affect students’ performances

and later opportunities, regardless of their innate ability or human capital.

Identifying and remedying such effects can potentially improve the match

quality between individuals and education paths.

The effect of institutionally driven exam timing has previously been

studied by Pope and Fillmore (2015). They explore how the time between

exams affects the exam scores of US students taking the Advanced

Placement exams, and they have found a linear relationship up to 10 days

(the maximum number of days in their data), with a more pronounced

effect for white students, Asian students, and female students. Their findings

indicate that only the score on the second exam is affected by how many

days pass between exams. They argue that this effect is likely to be a

result of fatigue. However, they admit that the heterogeneity patterns that

emerge are more likely to be caused by “cramming” (i.e., an intense

period of studying just before exams). The current paper differs from Pope

and Fillmore (2015) in four key ways. First, the variation in preparation

time is approximately twice as large, allowing me to identify a non-linear

effect over a longer period than 10 days. Second, Norwegian students take

exams in more subjects, allowing me to uncover heterogeneous effects

across subjects. Third, students taking the Advanced Placement exams are

generally more academically talented than the average upper secondary

school population (Pope and Fillmore, 2015). It is therefore of interest

to explore, as I do in this paper, whether the effects that Pope and Fillmore

(2015) have uncovered hold for a more typical student population. Finally,

I can follow students through longer-run outcomes. This is an important

contribution as it is not clear, ex ante, whether increased preparation time

only increases test scores, or whether it also affects longer-run outcomes,

such as university enrollment.

This paper makes two main contributions. First, I estimate the

causal effect of examination scheduling and preparation time on exam

performance; this expands our current understanding of how institutionally

driven shocks to exam conditions affect exam grades while human capital

levels remain very similar. Second, and importantly, I show how these

random variations in exam conditions affect longer-run outcomes. From

this, we can see how the sensitivity of exam scores to these conditions can

significantly affect students’ future opportunities and educational paths. The

more general lesson from these findings is that the institutional framework

1The papers included here are in no way a comprehensive list of previous studies. For further

contributions, see references in the papers cited above.

C

The editors of The Scandinavian Journal of Economics 2019.