Cassar, Lea; Fischer, Mira; Valero, Vanessa

Working Paper
Keep Calm and Carry On: The Short- vs. Long-Run
Effects of Mindfulness Meditation on (Academic)
Performance

IZA Discussion Papers, No. 15723

Provided in Cooperation with:

IZA – Institute of Labor Economics

Suggested Citation: Cassar, Lea; Fischer, Mira; Valero, Vanessa (2022) : Keep Calm and Carry On: The
Short- vs. Long-Run Effects of Mindfulness Meditation on (Academic) Performance, IZA Discussion
Papers, No. 15723, Institute of Labor Economics (IZA), Bonn

This Version is available at:

https://hdl.handle.net/10419/267460

Standard-Nutzungsbedingungen: Terms of use:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your personal
Zwecken und zum Privatgebrauch gespeichert und kopiert werden. and scholarly purposes.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial purposes, to
Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich exhibit the documents publicly, to make them publicly available on the
machen, vertreiben oder anderweitig nutzen. internet, or to distribute or otherwise use the documents in public.

Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen If the documents have been made available under an Open Content
(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, Licence (especially Creative Commons Licences), you may exercise
gelten abweichend von diesen Nutzungsbedingungen die in der dort further usage rights as specified in the indicated licence.
genannten Lizenz gewährten Nutzungsrechte.
DISCUSSION PAPER SERIES

IZA DP No. 15723

Keep Calm and Carry On:

The Short- vs. Long-Run Effects of
Mindfulness Meditation on (Academic)
Performance
Lea Cassar
Mira Fischer
Vanessa Valero

NOVEMBER 2022
DISCUSSION PAPER SERIES

IZA DP No. 15723

Keep Calm and Carry On:

The Short- vs. Long-Run Effects of
Mindfulness Meditation on (Academic)
Performance
Lea Cassar Vanessa Valero
University of Regensburg, CESifo and CEPR Loughborough University and CeDEx
Mira Fischer
WZB Berlin Social Science Center and IZA

NOVEMBER 2022

Any opinions expressed in this paper are those of the author(s) and not those of IZA. Research published in this series may
include views on policy, but IZA takes no institutional policy positions. The IZA research network is committed to the IZA
Guiding Principles of Research Integrity.
The IZA Institute of Labor Economics is an independent economic research institute that conducts research in labor economics
and offers evidence-based policy advice on labor market issues. Supported by the Deutsche Post Foundation, IZA runs the
world’s largest network of economists, whose research aims to provide answers to the global labor market challenges of our
time. Our key objective is to build bridges between academic research, policymakers and society.
IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper
should account for its provisional character. A revised version may be available directly from the author.

ISSN: 2365-9793

IZA – Institute of Labor Economics

Schaumburg-Lippe-Straße 5–9 Phone: +49-228-3894-0
53113 Bonn, Germany Email: publications@iza.org www.iza.org
IZA DP No. 15723 NOVEMBER 2022

ABSTRACT
Keep Calm and Carry On:
The Short- vs. Long-Run Effects of
Mindfulness Meditation on (Academic)
Performance*
Mindfulness-based meditation practices are becoming increasingly popular in Western
societies, including in the business world and in education. While the scientific literature has
largely documented the benefits of mindfulness meditation for mental health, little is still
known about potential spillovers of these practices on other important life outcomes, such
as performance. We address this question through a field experiment in an educational
setting. We study the causal impact of mindfulness meditation on academic performance
through a randomized evaluation of a well-known 8-week mindfulness meditation training
delivered to university students on campus. As expected, the intervention improves
students’ mental health and non-cognitive skills. However, it takes time before students’
performance can benefit from mindfulness meditation: we find that, if anything, the
intervention marginally decreases average grades in the short run, i.e., during the exam
period right after the end of the intervention, whereas it significantly increases academic
performance, by about 0.4 standard deviations, in the long run (ca. 6 months after the end
of intervention). We investigate the underlying mechanisms and discuss the implications
of our results.

JEL Classification: I21, C93, I12, I31

Keywords: performance, mental health, education, meditation, field
experiment

Corresponding author:
Mira Fischer
WZB Berlin Social Science Center
Research Unit on Market Behavior
Reichpietschufer 50
10785 Berlin
Germany
E-mail: mira.fischer@wzb.eu

* We are grateful to Michèle Belot, Alexander Cappelen, Ulrich Glogowsky, Uri Gneezy, Michael Kosfeld, Dorothea
Kübler, Mari Rege, Bertil Tungodden, Marie Claire Villeval, Roberto Weber and all the participants in seminars at Texas
A&M, UC San Diego, LUISS, LMU Munich, Humboldt University of Berlin, Erasmus School of Economics, University of
Stavanger, NHH, GATE Lyon and at the EALE and VfS Annual Conferences, the International Workshop on Applied
Economics of Education and the Field Days 2022 Workshop for useful comments. The study was pre-registered at the
AEA RCT registry (number AEARCTR-0004197. Financial support by Deutsche Forschungsgemeinschaft through CRC
TRR 190 (project number 280092119) is gratefully acknowledged.
1 Introduction

Mindfulness-based meditation practices are becoming increasingly popular in the

Western world, including in private and public organizations. 1 According to Lyddy
et al. (2021), more than half of all large companies o↵er their employees some form
of mindfulness training as of 2021. These include Aetna, Black Rock, Deloitte,
Deutsche Telekom, Ford, General Mills, Goldman Sachs, Google, HBO, McKin-
sey & Company, Nike, Procter & Gamble and SAP, among others. In addition
to firms, public sector and international organizations such as the United Nations,
the Australian government and the United States Postal Service o↵er similar pro-
grams to their employees. Mindfulness meditation is even part of the agenda of
the World Economic Forum. Similarly, more and more educational institutions
o↵er free mindfulness-based meditation courses to their students or consider inte-
grating them into their teaching curriculum. For instance, the British government
announced in February 2019 that in up to 370 English schools, students will begin
to practice mindfulness as part of a study to improve youth mental health. 2 Mind-
fulness programs are also o↵ered at universities where a large share of students
report experiencing stress and common mental health problems, such as anxiety
and depression. 3 Top universities, such as Cambridge, Harvard, LSE, MIT, Oxford,
Stanford, and Yale, are among those o↵ering free mindfulness meditation programs
to their students. 4
1
Even though consensus on an unequivocal definition of mindfulness is lacking so far, one of the
most commonly cited definitions was formulated by Jon Kabat-Zinn, who defined it as “a moment
to moment awareness that is cultivated by purposefully paying attention to the present-moment
experience, with a non-judgmental attitude” (Kabat-Zinn, 1994).
2
www.gov.uk/government/news/one-of-the-largest-mental-health-trials-launches-in-schools,
www.nytimes.com/2019/02/04/world/europe/uk-mindfulness-children-school.html
3
According to the American College Health Association Fall 2017 National College Health
Assessment (ACHA, 2018), 57% of college students in the US reported experiencing high levels of
stress in the past year. Moreover, 63% felt overwhelming anxiety and 41% felt so depressed that
they found it difficult to function. Note that these data are likely to underestimate the current
problem of poor mental health among college students. According to recent studies, the pandemic
sharply increased the risk of clinical depression among young US adults (Giuntella et al., 2021),
and these e↵ects persist even after easing of the pandemic (Barbieri et al., 2021).
4
Some of the top US business schools, such as HBS, NYU Stern, MIT, and the Rady School
of Management, have also begun o↵ering MBA and business courses on topics related to “mindful

1
 Extensive literature provides evidence on the e↵ectiveness of mindfulness med-
itation practices in achieving their main goals, namely, in reducing stress, anxiety
and depression (see, for instance, Khoury, 2015, for a review). However, little is
known about the potential spillovers of mindfulness meditation on other important
life outcomes, such as performance. This paper helps fill this gap by investigat-
ing the causal impact of a mindfulness meditation program on students’ academic
performance in higher education.
Mindfulness meditation may help improve academic performance by reducing
anxiety and depression (Owens et al., 2012), which are often associated with lower
academic performance (see Bernal-Morales et al. (2015) and Pascoe et al. (2020)
for a detailed literature review). As an example, among undergraduate university
students from the US, those with higher self-reported anxiety and depression levels
achieve poorer grades on examinations (Chapell et al., 2005; Hysenbegasi et al.,
2005). 5 Mindfulness practice may also improve academic performance by increasing
self-control and focus (Tang et al., 2015), which positively influence learning. Hence,
there are good reasons to believe that mindfulness meditation not only improves
mental health but may also help improve performance, which is arguably one of the
reasons why it has become so popular in education and business companies.6
However, other arguments point to the potential limitations or even negative ef-
fects of mindfulness meditation training on academic performance. One argument
concerns stress, whose relationship to performance is less straightforward than it
is for anxiety and depression. While excessive levels of stress are likely to impair
learning and memory, some level of stress could facilitate them (Vogel and Schwabe,
leadership”.
5
The worldwide economic costs of mental disorders, of which depression and anxiety are the
most common, are estimated to be US $2.5 trillion, due mostly to lost productivity. Mental
disorders thus account for higher economic costs than chronic somatic diseases such as cancer and
diabetes (Trautmann et al., 2016).
6
For instance, the University of Cambridge explicitly advertises its mind-
fulness training as a tool to support students to “study and thrive”.
https://www.cambridgestudents.cam.ac.uk/welfare-and-wellbeing/mindfulness-cam. In his
best sellers book, David Gelles argues that for companies, a more mindful workforce is more
productive (Gelles, 2015).

2
2016).7 In this case, the reduction in stress caused by mindfulness meditation could
possibly reduce academic performance. Furthermore, mindfulness meditation train-
ing may reduce students’ motivation to study by shifting attention away from future
states and academic goals towards the present moment and acceptance of the sta-
tus quo. In other words, mindfulness meditation could hinder goal-achievement
processes (Hafenbrack and Vohs, 2018). Finally, learning a new (mindfulness med-
itation) practice and adopting a new (mindfulness meditation) routine may divert
resources (in terms of cognitive e↵ort and time) from studying. 8 However, in this
last case, the negative e↵ect of the training should be short-lived, because as the
students repeat actions, the meditation practice is learned and habits are formed,
thereby requiring minimal decision-making and minimal regulatory control (and
probably less time) in the long run.
Taken together, these arguments suggest that there are potential countervail-
ing e↵ects of mindfulness meditation on academic performance. Furthermore, the
importance of each of these e↵ects may be changing over time. With the above dy-
namics and mechanisms potentially in play, it is difficult to predict how mindfulness
meditation ultimately a↵ects academic achievement. To the best of our knowledge,
our study is the first pre-registered and relatively large randomized controlled trial
to investigate the causal e↵ects of a mindfulness meditation training on short- and
long-term academic performance.
In collaboration with one of the largest German health insurance providers, we
o↵ered a free 8-week mindfulness meditation course to students at the University
of Cologne. 9 Interested students could apply to the course by registering and com-
7
This inverted U-shaped relationship between arousal and performance is known as the Yerkes-
Dodson law (Yerkes and Dodson, 1908). See Teigen (1994) for a discussion.
8
When initiating and implementing a new behaviour, people typically decide what to do
and how to do it in order to achieve certain outcomes and avoid others (Wood and Rünger,
2016). However, making conscious decisions and exerting willpower are cognitively demanding and
require the expenditure of some inner, limited resources that are depleted afterward (Muraven and
Baumeister, 2021). This would imply that learning a new practice and adopting new meditation
habits could reduce the amount of resources available to studying in the short term.
9
A pre-pandemic study by Grobe et al. (2018) reveals that in Germany about one in six
university students (almost half a million) experiences depression, anxiety, or panic attacks and

3
pleting an online questionnaire. Applicants were then randomly assigned to the
treatment or the control group, and students in the treatment group were o↵ered
a place in the course. We then measured the e↵ects of this intervention on the
short- and long-run grades of the students assigned to the treatment group relative
to students assigned to the control group, who did not have access to the course.
The program was based on the well-established “mindfulness-based stress reduc-
tion” (MBSR) course, developed by Jon Kabat-Zinn (1994) in the US in the 1970s.
The MBSR training has two great advantages: it is secular, and it is highly stan-
dardized. Thanks to these characteristics, the training has been subject to many
scientific studies (see our literature review in Section 2) and is now used worldwide
not only in education but also in health care and in companies and organizations.
The program consisted of eight weekly group sessions with an experienced and
certified MBSR teacher and daily individual exercises. To determine the impact of
the training on our primary outcome of interest – students’ academic performance
– we use grade information from the university’s administrative records. We have
access to all grades obtained in the semesters before the intervention, soon after
it ended, and up to six months after it. These grades allow us to investigate and
compare the intervention’s short- and long-term e↵ects on academic performance.
In addition, we measured students’ mental health, non-cognitive and cognitive skills,
study behavior, and health (self-care) behavior. These data were collected before
the intervention started and soon after it ended. They allow us to test whether
our intervention achieved its goals in terms of improving mental health and to
investigate channels through which it may a↵ect grades in the short run.
Our main findings can be summarized as follows. Consistent with previous
evidence, the mindfulness meditation program significantly improved students’ self-
reported mental health (stress, anxiety, and depression). It also significantly im-
proved self-reported non-cognitive skills (self-control, conscientiousness, and neu-
that the proportion of 18- to 25-year-old adults diagnosed with one of these mental disorders rose
by 38% between 2005 and 2016.

4
roticism) and very marginally increased cognitive skills (measured by performance
in an incentivized Stroop task). However, we find that, if anything, the inter-
vention marginally decreased students’ short-term academic performance. In a
“value-added specification” (VA) controlling for prior performance and program
and study-year fixed e↵ects, the intervention reduced grades by 0.26 sd (p = 0.055).
When using a “first-di↵erences specification” (FD), the e↵ect is still negative but
smaller and no longer significant.
We only observe robust positive e↵ects of the intervention on academic perfor-
mance in the long term. Specifically, the mindfulness training significantly improved
long-term academic performance by, on average, 0.386 (p = 0.017) sd in a VA spec-
ification and by 0.446 sd (p = 0.010) in a FD specification. No e↵ect was found on
the number of exams – more precisely, on the number of credit points – taken in
each semester, suggesting that the long-term improvement (short-term reduction)
in grades did not come at the expense (advantage) of taking fewer (more) exams.
These results survive additional robustness checks.
When exploring the underlying mechanisms behind our findings – and, in partic-
ular, behind our marginally negative e↵ect in the short run – our data indicate that
the intervention led to an increase in daily healthy routine and self-care practices,
such as sleeping more and relaxing more consciously, both of which were found to
be the best predictors of changes in short-term grades among all our pre-registered
potential channels (i.e., mental heath, cognitive and non-cognitive skills, and study
and health behavior). So overall, our evidence is consistent with the argument
above according to which adopting a new meditation practice and related healthy
habits reduces the amount of resources available for studying in the short (but not
the long) term. We also find that the long-term positive e↵ect of the intervention
on academic performance is driven by those students who practiced on their own
in addition to the course’s instructions and requirements – and thus most likely
continued practicing after the end of the intervention.

5
 Taken together, these findings reveal that, ultimately, mindfulness meditation
can have substantial positive spillover e↵ects on academic performance but that it
may take time before students can reap these additional benefits.
The paper proceeds as follows. The next section outlines our contribution to
the related literature in economics and briefly reviews related work in other fields.
Sections 3 and 4 describe the experimental design and the intervention, respectively.
Section 5 presents our empirical strategy. Section 6 reports the main findings and
robustness checks. Section 7 explores potential channels that may help explain our
results. Section 8 discusses the external validity and generalizability of our results.
Section 9 concludes.

2 Related Non-economic Literature

2.1 Literature in Economics

Our study contributes to di↵erent strands of literature in economics. First, we con-

tribute to the economics literature on the determinants of academic achievement.
A large body of literature examines the role of monetary and non-monetary incen-
tives in improving student achievement (see, e.g., Angrist et al., 2009; Leuven et al.,
2010; Levitt et al., 2016; Hvidman and Sievertsen, 2019; Lindo et al., 2010; Burgess
et al., 2021). 10 Compared with e↵ect sizes of about 0.2–0.4 standard deviations
commonly found in this literature, the e↵ects of our mindfulness intervention are
sizeable. More closely related to our study, Cappelen et al. (2017) investigate the
e↵ects of physical exercise on academic performance in a field experiment. They en-
courage a cohort of Norwegian college students to engage in more physical activities
by providing them free gym memberships and find that their intervention increases
gym attendance and academic performance by 0.15 standard deviations compared
with a control group. Whereas their paper investigates positive spillover e↵ects of
10
For a review of the experimental literature, see Schildberg-Hörisch and Wagner (2020).

6
physical exercise on academic performance, we examine spillovers of mindfulness
meditation on academic performance.
Second, our paper relates to an emerging but fast-expanding literature in eco-
nomics that evaluates the impact of meditation interventions on mental health,
economic behavior, and cognitive performance. Di Bartolomeo and Papa (2016)
find that subjects practicing meditation shortly before playing an investment game
exhibit more trust and pro-social behavior. Alem et al. (2021) study the impact of
an online MBSR program on stress and risk and time preferences among students
in the UK. Consistent with our findings, they find strong evidence that mindful-
ness training reduces perceived stress, measured using the Perceived Stress Scale.
However, they only find suggestive evidence that such programs can a↵ect attitudes
towards risk and time. Closely related to our paper are the studies by Charness
et al. (2022) and Shreekumar and Vautrey (2022). The former investigates the e↵ect
of a three-month training program based on principles of mindfulness and positive
psychology in three large firms on trainees’ cognitive performance in incentivized
decision-making tasks and psychological questionnaires. The authors find strong
evidence of a reduction in self-reported stress (which persists 3 months after the
end of the training) but only marginal and selective positive e↵ects of the training
on the incentivized tasks. Shreekumar and Vautrey (2022) conducted a four-week
experiment in which a large sample of US adults received free access to a popular
mindfulness meditation app. They find a reduction in symptoms of stress, anxiety,
and depression, and a 1.9% earnings increase on a proofreading task conducted two
weeks after the start of the intervention. They do not find, however, an e↵ect on the
Stroop task. Our study contributes to this literature by investigating the e↵ect of a
mindfulness training on performance in a di↵erent, but equally important, context
(education), and, crucially, it is unique in that it evaluates the e↵ects of mindfulness
training on a very important high-stakes outcome (academic performance) in addi-
tion to its e↵ects on self-reported health measures and an incentivized “laboratory”

7
task, thereby o↵ering greater external validity.
More broadly, our paper also contributes to an emerging literature in economics
investigating how well-being a↵ects performance. Oswald et al. (2015) is among
the first economic studies to address this question. They provide evidence from
a series of laboratory experiments for a positive causal link between human well-
being and performance. More recently, Bellet et al. (2019) and Coviello et al. (2021)
investigate this question in the field, with mixed results. While Bellet et al. (2019)
find a positive e↵ect of positive mood on workers’ sales performance, Coviello et al.
(2021) mainly find a negative e↵ect. Our paper di↵ers from – and thus complements
– these field studies in two crucial ways. First, while in the context of sales activities
an important determinant of performance comes from soft and social skills, in the
context of exams’ performance these channels are switched o↵. Thus, in a way, our
intervention provides the hardest test of the hypothesis that well-being increases
performance. Second, compared with these previous studies, which mainly rely
on short-lived shocks (like weather changes) that manipulate people’s mood, our
MBSR training introduced participants to well-being-enhancing practices that they
can apply any day for the rest of their lives. So in our intervention, participants’
well-being is likely to have been influenced in a more “fundamental” and persistent
way than in these previous studies.
Our paper is also related to the literature in economics evaluating the impact
of psychological interventions on important life outcomes. Heller et al. (2017) find
that interventions based on cognitive behavioral therapy (CBT) reduce delinquency
and increase school engagement and graduation. Blattman et al. (2017) also find
positive e↵ects of CBT on reducing crime, and Baranov et al. (2020) find that it
improves postpartum depression and parental decision-making. John and Orkin
(2022) find that simple psychological interventions can increase preventive health
behavior. We contribute to this literature by showing that mindfulness training
can a↵ect not only mental health but also important economic outcomes, such as

8
(academic) performance.
Finally, our paper contributes to the economic literature on the malleability of
non-cognitive skills. Existing studies usually concentrate on the ability to improve
non-cognitive skills of children (e.g., Heckman and Kautz, 2012; Alan et al., 2019). A
few exceptions, such as Blattman et al. (2017), focus on adults. These papers show
that such skills may be malleable at a later age. While their studies concentrate
on developing countries, our paper provides further evidence of such malleability in
young adults in a developed country. 11

2.2 Literature from other fields

In this section we briefly review closely related papers in the non-economic litera-
ture. First, we discuss the literature on the e↵ects of MBSR on variables that may
be relevant for academic achievement, and further evidence that these variables are,
indeed, related to academic performance. We then review the few existing studies
on the causal e↵ects of meditation on task performance and academic achievement.
While there is great interest in the application of MBSR in people experienc-
ing mental or physical illness (Goyal et al., 2014), the practice has similar benefits
in healthy people. Several studies review evidence about mindfulness-based tech-
niques in nonclinical populations. Chiesa and Serretti (2009) analyze 10 studies on
MBSR and conclude that it had a significant e↵ect on the reduction of stress levels
in healthy people. Eberth and Sedlmeier (2012) review evidence for mindfulness
meditation on various psychological variables. Analyzing 39 randomized controlled
trials with meditators in non-clinical settings, they conclude that mindfulness medi-
tation (including MBSR) had a significant e↵ect on the reduction of stress levels. In
addition, they found that it reduces anxiety and increases psychological well-being.
11
This is relevant because research in psychology has shown that in a✏uent families, IQ and
personality traits are to a larger extent determined by genes and are less likely to be influenced
by the environment than in poorer families (see review by Almlund et al. (2011)). Extending
this argument to countries’ di↵erences, one may reasonably conjecture personality traits to be less
malleable in developed than in developing countries.

9
Khoury et al. (2015) review 29 studies and confirm previous findings for stress and
anxiety. In addition, they conclude that MBSR is e↵ective in reducing depression.
Eberth and Sedlmeier (2012) identify a multitude of other positive e↵ects of
mindfulness training on cognitive (e.g., attention, memory) and non-cognitive (e.g.,
neuroticism) skills. Tang et al. (2015) review the literature investigating the e↵ects
of MBSR on self-regulation and find emerging evidence that it may positively a↵ect
the functioning of brain regions involved in the regulation of attention, emotion,
and self-awareness.
Few studies investigate the causal impact of meditation directly on performance.
Hafenbrack and Vohs (2018) report the results from a series of laboratory and
online experiments examining the e↵ects of a 15-minute mindfulness training on
task motivation and task performance. Together, their findings suggest that while
mindfulness meditation may impair participants’ motivation to complete cognitive
and performance tasks, it does not ultimately a↵ect performance on those tasks.
According to the study, the demotivation e↵ect of the intervention is mediated
by a reduction in future focus and arousal. However, the authors also find that
mindfulness meditation enables people to detach from stressors, which improves
task focus and may explain why, overall, mindfulness does not alter performance.
More similar to our study, Hall (1999) and Mrazek et al. (2013) implement longer
interventions and focus on academic performance. Hall (1999) randomly assigned 56
undergraduates to two study groups, one of which included meditation. Each study
group met one hour twice a week for the duration of the academic semester. In
the treatment group, students were instructed in natural breathing, relaxation and
attention-focusing techniques. The meditation process was practiced for 10 minutes
each at the start and conclusion of the study session. At the end of the semester, the
treatment group achieved a significantly higher grade point average (GPA) than the
control group. In Mrazek et al. (2013), 48 undergraduate students were randomly
assigned to either a mindfulness class or a nutrition class. Classes met for 45

10
minutes four times a week for two weeks. They found that the mindfulness training
improved GRE reading comprehension scores. Compared with these studies, our
sample size is four times as large and, importantly, is balanced not only along
past performance but also along all the channels through which the mindfulness
training may a↵ect performance, namely, mental health, cognitive and non-cognitive
skills, and studying and health behavior at the baseline. Hence, to the best of our
knowledge, our study is the first pre-registered and relatively large randomized
controlled trial to investigate the causal e↵ect of a mindfulness meditation training
on short- and long-term academic performance.

3 The Experiment

In collaboration with a major German health insurance provider, we o↵ered a free

eight-week mindfulness meditation course to students at the University of Cologne.
Interested students could apply for the course by registering and completing an
online questionnaire. Eligible applicants were then randomly assigned to the treat-
ment or the control group, and students in the treatment group were o↵ered a place
in the course. 12 We describe the design of the experiment in more detail below. The
experiment was pre-registered (AEARCTR-0004197) and received ethics approval
from the university’s institutional review board.

3.1 Recruitment

In the summer semester of 2019, a mindfulness meditation course was o↵ered to

students at the Faculty of Management, Economics and Social Sciences of the Uni-
versity of Cologne. The course was advertised beginning on April 8, 2019, via the
12
Given that we randomly assigned students who applied to the mindfulness training to either
a treatment or a control group, for obvious reasons the latter were not o↵ered any other course
in substitution. This design feature also has one disadvantage: having a passive control group
makes it harder to disentangle di↵erent mechanisms underlying a potential treatment e↵ect of the
meditation training (e.g., the e↵ect of the course’s content vs. time spent in the course). We try
to address this issue in section 7 where we investigate channels possibly explaining our results.

11
mailing list, newsletters, and social media accounts of the same faculty (see Figure
D.1 in Appendix D for an English translation of the announcement text).
Interested students could apply for a place in the course until April 24, 2019,
by registering and completing a baseline survey on a website implemented with the
survey software Qualtrics. 13 On the registration page, students were provided some
basic information about the origins and potential benefits of mindfulness meditation
and were asked to indicate which time slots would fit their schedule and which
exams they were planning to take at the end of the summer semester. They also
gave their consent to access and use the data on their grades and to link it to
their questionnaire answers. The baseline questionnaire included questions eliciting
students’ mental health, cognitive and non-cognitive skills, and study and health-
related behavior.
Students were informed that the course was financed by a large German health
insurance provider and would be free to them, but that places were limited and thus
not everyone could participate. We also informed students that the available slots
would be allocated through a lottery, taking into account their time availability
and the lectures they were planning to take. We highlighted that their answers to
the baseline questionnaire would not a↵ect their chance of getting a place. Finally,
applicants were asked to agree to answer a follow-up questionnaire, independently
of whether they would be placed in the course or not (see Figures D.2 in Appendix
D for an English translation of the registration page). They were paid 10 euros for
completing the survey.
13
Initially, the deadline was April 15, and the advertisement only targeted bachelor’s-level stu-
dents of the Faculty of Management, Economics and Social Sciences of the University of Cologne.
However, because by that time we had not attracted sufficient applicants, we extended the dead-
line to April 24 and decided to advertise the course to master’s-level students of the same faculty
as well.

12
3.2 Timeline

Table 1 provides an overview of the relevant dates and events of the study. The re-
cruitment of students started one week after the beginning of the summer semester
and lasted for two weeks, until April 24. At the beginning of May, students were
informed about whether they had received a place in the course or not. The medi-
tation course started on May 15 and lasted for nine weeks, with a break of one week
in the middle of June for the spring break holidays. The course was timed such that
it ended together with the official lecture period of the summer semester, on July
12. Most exams of the summer semester are written in the three weeks surrounding
the end of the lecture period (see Figure B.1 in the Appendix for the timing of
exams). Thus, our main analysis on the short-term e↵ects of the intervention on
academic performance uses all the grades of the exams written during the main
exam period of the summer semester, namely from July 6 until July 27. 14 Consis-
tent with this definition, our analysis on the long-term e↵ects of the intervention
on academic performance focuses on grades of the exams written during the main
exam period of the winter semester, namely January 25 through February 19: i.e.,
about half a year after the end of the intervention. As shown in Figure B.1 in the
Appendix, there are two secondary exam periods, in the second half of September
and in the end of November / beginning of December, in which fewer exams are
written. These consist of the retakes and voluntarily delayed exams of the summer
semester and the midterm exams of the winter semester, respectively. These can be
used to measure intermediate-term e↵ects of our intervention. However, since the
sample size is quite restricted, they are not the main focus of our analysis, although
14
As shown in Figure B.1, some exams are written in the middle of the summer semester, with
a peak around May 15: i.e., around the beginning of the meditation course. These exams are for
programs in which some lectures only last for the first half of the semester’s lecture period. As
they take place before the beginning of the meditation course, we do not expect any e↵ects of
the course on grades in those lectures. Also note that midterm exams are usually organized not
centrally by the faculty but by the responsible lecturers and that they have some discretion in
choosing around the middle of the lecture period instead of sticking to the dates they registered in
the system, which is why some of the exams are not properly dated and appear spread out during
the lecture period.

13
we still report results for them in additional analyses.

Table 1: Timeline

2019 April 1 Beginning of summer semester lecture period

April 8 - 24 Recruitment of participants and baseline questionnaire
April 25 - May 7 Applicants informed about treatment allocation
May 15 Beginning of meditation course
June 10 - 14 Pentecost holiday
July 12 End of meditation course and of summer semester lecture period
July 15 - 29 Follow-up questionnaire
July 6 - 27 Exam period of summer semester
2020 Jan 25 - Feb 19 Exam period of winter semester
Note: This table shows key dates of the field phase of the experiment.

3.3 Outcome Measures

We accessed students’ pre- and post-intervention grade information from the univer-
sity’s administrative records. Information on prior grades allows us to test whether
the treatments are balanced across academic performance prior to the intervention.
Grades at German universities vary from 1 to 6 following a descending order: i.e.,
1 represents the best possible grade and 6 the worst possible grade. Grades lower
than or equal to 4 are passing grades. For the sake of clarity, we invert the grades
so that higher grades correspond to better academic performance.
Our primary pre-registered outcome, the grade average, is derived from com-
puting the weighted arithmetic mean of the grades obtained by a student within
the di↵erent exam periods as defined in section 3.2. Each grade is multiplied by
the study points a student received for it and then divided by the sum of the study
points the student earned for the whole period:
Pn
k=1 (module gradei,t,k ⇥module pointsk )
gradei,t = Pn
k=1 module pointsi,t,k

where i denotes the student, t denotes the exam period, and k denotes the module.
This is the same formula the university uses to compute a student’s average grade.
Furthermore, as the means and standard deviations of grades di↵er significantly
across the 10 study programs our participants follow (e.g., the mean passing grade

14
is 1.8 (sd 0.48) for the MSc in Political Science and 2.6 (sd 0.78) for the BSc in
Economics), we standardize the grades to a mean of 0 and a standard deviation of
1 at the program level to make performance comparable across the programs. This
also makes our e↵ect sizes easily comparable with those found in other studies.
Mental health (stress, anxiety, and depression), non-cognitive skills (self-control,
conscientiousness, and neuroticism), and study and health behavior are self-reported
variables, whereas attention is measured with an incentivized Stroop task (all mea-
sured before and after the intervention) that we use to investigate the channels. All
variables were pre-registered. We chose these variables because we know from the
literature that they (i) are influenced by meditation, and (ii) are relevant for aca-
demic outcomes (see our literature review in Section 2). All items from the baseline
and follow-up questionnaires are listed in Table D.3 in Appendix D.
Stress is measured using the well-known Perceived Stress Scale (PSS; Cohen
et al., 1983). The PSS includes 10 items, each rated on a scale from 0 to 4, expressing
how often, in the last two weeks, the participant felt as described by the statement
(0=never, 4=very often). Our measure of stress is the sum of these 10 items, with
a higher value corresponding to greater stress.
We measure anxiety using the well-known seven-item Generalized Anxiety Dis-
order questionnaire (GAD-7; Spitzer et al., 2006). The test includes seven items,
each rated on a scale from 0 to 3, expressing how often, in the last two weeks, the
participant has been bothered by the problem described by the statement (0=not
at all, 3=nearly every day). The variable anxiety is then constructed by taking the
sum of these seven items, with a higher value corresponding to being more anxious.
Similarly, we measure students’ depression using the standard clinical screen for
depression, the nine-item Patient Health Questionnaire (PHQ-9; Kroenke et al.,
2001). The test contains nine items, each rated on a scale from 0 to 3, expressing
how often, in the last two weeks, the participant has been bothered by the prob-
lem described by the statement (0=not at all, 3=nearly every day). The variable

15
depression is then constructed by taking the sum of these nine items, with a higher
value corresponding to being more depressed.
We measure students’ self-control using the Brief Self-Control Scale (BSCS;
Tangney et al., 2004). This test contains 13 items, each rated on a scale from
1 to 5, expressing the extent to which the statement reflects how the participant
typically is (1=not at all, 5=very much). Our self-control variable is the sum of
these 13 questions, with a higher index corresponding to greater self-control.
To measure students’ attention, we use an incentivized Stroop task (Stroop,
1935), which requires participants to identify the color of a printed word when the
word’s meaning and color may be incongruent. The individual score is computed by
dividing the number of correct answers (of a total of 20) by the time a participant
needs to answer all tasks. We incentivized the task by paying 20 euros each to those
participants who were the three fastest among those with the most correct answers.
As for all the other survey items, the task was easily implemented using the survey
software Qualtrics.
We elicit students’ conscientiousness and neuroticism using a selected number
of questions from the Big Five Inventory (BFI; John et al., 1991). More specifically,
our questionnaire included nine items to elicit conscientiousness and seven items
to elicit neuroticism. Each item was rated on a scale from 1 to 5 expressing the
extent to which the participant agreed with the statement (1=disagree, 5=agree).
Our conscientiousness and neuroticism variables are then constructed by taking the
sum of these nine and seven items, respectively, with higher values corresponding
to a higher level of conscientiousness and neuroticism, respectively.
The questionnaire also included 15 questions related to study behavior, in par-
ticular to elicit concentration while studying (5 questions), learning strategies (4
questions), studying self-concept (3 questions) and exam behavior (3 questions).
Each question was answered on a scale from 1 to 5 expressing the extent to which
the participant agreed with the statement (1=disagree, 5=agree). Each of our study

16
behavior indexes is based on the sum of the respective variables, with a higher in-
dex corresponding to better study concentration, strategies, self-concept and exam
behavior, respectively.
Finally, the questionnaire included seven questions related to health behavior
and self-care practices. Given that there is no natural or obvious categorization of
such questions – unlike for the study behavior variables – we present the results both
by using each statement separately and by aggregating them in a single index based
on the sum of the seven variables, with a higher index corresponding to healthier
behavior.
In addition to the above-mentioned pre-registered variables, we collected a mea-
sure of participants’ mindfulness and of participants’ motivation to apply for the
training. Our mindfulness scale was adapted from the German Socio-Economic
Panel Study Innovation Sample (SOEP-IS). The index contains eight questions,
each answered on a scale from 1 to 6, expressing how often, in the last two weeks,
the participant experienced the situation described by the statement (1=almost
never, 6=almost always). The index is the sum of these eight questions, with a
higher index corresponding to greater mindfulness. To elicit students’ motivation
for applying to the course, they were asked the following question: “What moti-
vation is most important to your desire to learn mindfulness meditation? Please
choose an option: - I am curious; - I want to improve my concentration; - I want to
learn to relax better; - I want to learn to better deal with my emotions; - I want to
lose weight; - I want to be more productive.”

3.4 Sample Size and Randomization

The sample size of our experiment was determined by the number of applicants
and by the budget available to the health insurance provider, as follows. By April
24, 2019, we had received applications (completed surveys) from 282 candidates.
From this group, we excluded 58 because (i) they did not plan to write any exams

17
in the summer semester, (ii) they did not plan to take any exam that at least
one other applicant indicated they would write in the summer semester, and/or
(iii) they did not indicate any availability for any of the time slots o↵ered for the
meditation course. Eligible applicants were randomly assigned to the treatment or
to the control group at the individual level. Applicants who were only planning
to write less “popular” exams (i.e., exams that few other applicants would also
write) were randomized stratified along these exams such that, wherever possible,
for each student in the treatment group who is writing an exam, there is a student
in the control group writing the same exam. Finally, the budget provided by the
health insurance provider could cover the costs of the course for a maximum of
102 participants (6 meditation groups of 17 participants each). Hence, based on
this restriction and on the randomization procedure described, we allocated 102
applicants to the treatment group and 122 to the control group. Our sample size is
thus 224 observations.

4 The Intervention

Students allocated to the treatment were o↵ered a place in a mindfulness medita-

tion course based on the renowned “mindfulness-based stress reduction” (MBSR)
program. The MBSR program was developed by Jon Kabat-Zinn (1994) in the US
in the 1970s and is now used successfully worldwide in health care, psychotherapy,
and education, at companies and organizations, and for many people experiencing
stress. Mindfulness is a broad term that encompasses di↵erent meditation tech-
niques, cognitive skills, and attitudes and is commonly defined as “paying attention
in a particular way: on purpose, in the present moment, and non-judgmentally”
(Kabat-Zinn, 1994, p. 4). It thus involves practicing a skill (focusing attention on
a chosen object) and adopting a particular attitude (equanimity). As suggested by
its title, the main goal of the MBSR program is to reduce stress and improve mental

18
health more generally. Since its development 50 years ago, numerous international
studies have proved MBSR’s health-promoting, stress-reducing, and quality-of-life-
enhancing e↵ects (see the literature review in Section 2).
A standard MBSR course comprises eight weekly 150-minute group sessions and
one full-day retreat. Furthermore, participants in a standard course are asked to
use an audio recording to practice 45 minutes, six days a week. We designed our
intervention to adhere as closely as possible (given the funding and university set-
ting) to the standard MBSR course. More specifically, our course comprised eight
weekly 60-minute group sessions, which participants could attend in a room cen-
trally located on the university campus, in a building of the Faculty of Management,
Economics and Social Sciences, on Wednesday or Friday afternoons, depending on
their group assignment. The room was chosen to minimize the time students would
need to reach it, and most students’ lectures take place in the same or adjacent
buildings. 15 Each group included 17 participants, and there were six groups in to-
tal, three meeting on Wednesdays and three on Fridays. Similar to the standard
MBSR program, our participants also received audio recordings and handouts and
were asked to practice a 12-minute “body scan” or mindful movement exercise or a
30-minute sitting meditation alternating once a day. Furthermore, they were asked
to take a “time out” for three minutes and to reflect on their thoughts and feelings
in the present moment three times a day. Also in terms of content, our course was
closely oriented towards a standard MBSR course. Participants were taught formal
sitting mindfulness meditation, body awareness, and mindful movement exercises
and participated in group discussions of experiences. During the mindfulness ex-
ercises, they were asked to focus on sensory objects of awareness, such as sounds
and smells; breathing awareness; body sensations such as pressure, pain, itches, and
tensions; and thoughts and emotions. The course was designed and taught by two
experienced, certified MBSR teachers, who were selected and paid by the health
15
For the sake of external validity, note that most weekly trainings o↵ered by the top universities
listed in the introduction also last between 1 and 2 hours.

19
insurance provider. 16 Each group session was always taught by the same teacher.
Before the start of the intervention, the two teachers worked closely together to
design the course structure and the materials and content of each of the eight ses-
sions, so that each weekly session was as homogeneous as possible across the six
groups. For a detailed description of the structure and content of each of the eight
weekly sessions, see Table D.4 in Appendix D. Note that although the teachers and
participants knew that we would evaluate the program, they were not aware of the
main goal of the study.
On average, students who were randomly assigned a place in the course at-
tended just over half of the eight classes (4.5, sd: 2.6). Overall, 57% of assigned
students were present at their course during any week, but Figure B.2 shows that
mean attendance steadily decreased, from 81% in the first week to 35% in the final
week. Fifty-six percent of students attended more than half ( 5) the sessions. 17
Moreover, 59% of respondents of the post-intervention survey who were assigned to
the treatment group reported having done the exercises at least once a day at least
every other day during the first four weeks of the course, while 50% reported having
done so during the latter four weeks of the course, and 63% reported continuing to
practice at least some of the exercises at the time the survey was running (the first
and second week after the intervention ended). When asked whether they liked the
course, 89% of respondents reported that the course was “very good” or “good”.
Finally, 78% of respondents reported that they had learned “very much” or “much”
during the course, while 88% said that they would “definitely” or “probably” rec-
ommend the course to other students.
16
Note that in Germany the individual cost of attending a standard MBSR program outside
of the university is about 400 euros, which is equivalent to about $450 (as of the time of the
experiment).
17
See Figure B.3 for the distribution of participation by student. We exploit this variation
when analyzing the role of treatment intensity.

20
5 Empirical Strategy

The main goal of MBSR trainings is to improve mental health. Furthermore, pre-
vious studies have shown that this training can also benefit cognitive and non-
cognitive skills. Hence, we first test whether our training achieved its goals. We
begin by analyzing the e↵ects of the training on short-term mental health (stress,
anxiety, depression), cognitive skills (focus) and non-cognitive skills (self-control,
conscientiousness, neuroticism). Second, we test whether our intervention had any
spillover e↵ects on students’ academic performance, which was registered as our
primary outcome variable. Our data on students’ grades allows us to distinguish
between the e↵ects of the intervention on short- vs. long-term performance.
For each outcome, we report results of OLS regressions with and without control
variables. We refer to models that contain only a treatment dummy and no control
variable as “endline” (EL) specifications, as they constitute a simple comparison
between treatment and control groups of mean outcomes measured after the in-
tervention, and to models that control for the baseline measure of the respective
outcome as well as baseline-performance, study-program, and study-year fixed ef-
fects as “value-added” (VA) specifications. 18 VA specifications take the following
form:

yit = ↵ + T reatmenti + yit 1 + Xi + " (1)

where ↵ is a common intercept; T reatmenti is a binary variable equal to 1 if student

i was assigned to the treatment group, and 0 otherwise; and yit is the dependent
variable which, in the case of the secondary outcomes, takes the form of an index
18
The “value-added” terminology is borrowed from the literature estimating “educational pro-
duction functions”, which goes back to Ben-Porath (1967) and Hanushek (1979). The idea is that
a student’s human capital accumulation up to time period t-1 can be measured, so that the value
added to it by inputs invested during period t can be estimated (Koedel and Rocko↵, 2015). Since
in our study the treatment was randomized and none of the outcome measures were significantly
unbalanced at baseline, EL estimates can be interpreted as causal. The inclusion of control vari-
ables in the VA models is merely intended to improve the precision of the estimates and serves as
a robustness check.

21
over the questions used to elicit that specific outcome. All scales are standardized
over the whole sample to have a mean of 0 and a standard deviation of 1.
In the case of our primary outcome, yit is equal to gradei,t – as defined by
equation 3.3 – namely, the post-intervention performance of student i measured by
the weighted arithmetic mean of the grades obtained either (i) in the main exam
period of the 2019 summer semester right after the intervention, or (ii) in the main
exam period of the following winter semester, about half a year after the end of the
intervention. These two variables represent, respectively, students’ performance in
the short and in the long run. Because in the German grading system, smaller grades
are better grades, to avoid confusion we invert the grades such that larger grades
are better. To measure performance in comparable ways across study programs, we
standardize grades to a mean of 0 and a standard deviation of 1 at the program
level. 19
yit 1 is the baseline measure of the respective outcome. In the case of the
secondary outcomes, yit 1 is equal to the baseline scale elicited in the first question-
naire; in the case of the primary outcome, yit 1 is equal to the student’s prior grade:
i.e., the weighted average grade over all modules completed in the current program
before the semester in which the intervention took place. Finally, Xi denotes a row
vector of control variables, consisting of the following: prior ECTS credits (sum of
study points over all modules completed in the current program before the semester
in which the intervention took place), study-program dummies (10 categories; see
Table 2), and study-year dummies (4 categories: “1st year”, “2nd year”, “3rd year”,
“4th or higher year”). In this model, can be interpreted as the average treatment
e↵ect.
Additionally, for each outcome, we also report results of running alternative
19
See Section 3.3 for more details on how we compute the outcome.

22
OLS regressions of the following form:

yi = ↵ + T reatmenti + " (2)

where yi is the di↵erence between the endline and baseline (yit - yit 1 ) out-
comes. In this model, can be interpreted as the average di↵erence-in-di↵erences
between the treatment and control groups. 20 We refer to this model as a “first-
di↵erence specification” (FD). It is an individual-level fixed-e↵ects model that only
uses variation at the level of the individual and is intended to eliminate bias from
unobserved, time-invariant variables. We show these results as a robustness check,
acknowledging that neither the VA nor the FD models are clearly superior in our
case. 21
Our empirical strategy can be interpreted as follows. The univariate regression
of the endline score on the treatment dummy (i.e., equation 1 without controls) does
not control for any possible imbalances at baseline and simply compares outcomes.
The multivariate regression of the endline score on the treatment dummy and the
baseline score, as well as baseline performance and study-program and study-year
fixed e↵ects (i.e., equation 1), identifies e↵ects from di↵erences between treatment
and control groups within the cells resulting from the control variables. And, fi-
nally, the univariate regression of the first di↵erence on the treatment dummy (i.e.,
equation 2) identifies the treatment e↵ect from changes within individuals. Thus,
these models control for possible imbalances at baseline with increasing strictness.
Note, however, that none of our baseline measures are significantly unbalanced. We
20
The first-di↵erence specification produces exactly the same coefficient and standard error for
the treatment dummy as a di↵erence-in-di↵erences model with standard errors clustered on the
individual level (using the ”cluster” option in Stata that is based on Liang and Zeger (1986)) of
the form yi = ↵ + T reatGroupi + T imei + (T reatGroupi ⇤ T imei ) + ".
21
The FD model assumes very high predictive power of the baseline for the endline measure.
However, if the baseline is based on di↵erent measures than the endline and thus is a noisy control
variable, the VA model might be preferred. For a discussion, see McKenzie (2012). In our case,
the grades are based on di↵erent exams in the pre- and post-intervention periods, and correlations
are rather low: 0.49 for the baseline and short-term grade and 0.58 for the baseline and long-term
grade.

23
show all specifications throughout. 22

6 Results

6.1 Balance Checks, Attrition, and Representativeness

We begin this section by investigating whether the treatment and control groups
were balanced at baseline. Table A.1 in Appendix A presents mean values of the
(pre-registered) variables and demographics elicited in the baseline survey and the
baseline performance from the university’s registry, separately for the treatment and
control groups, using the overall sample. The scales, grades, and demographics were
all balanced for the overall sample. Of the 224 students on our overall sample, 181
(84 (82%) in the treatment group and 97 (80%) in the control group) took at least
one exam in the main examination period of the summer semester, immediately
after the end of the program. These grades constitute our short-term performance
outcomes. Table A.2 shows also that this reduced sample was balanced at baseline,
although the p-value for the prior grade becomes rather small at 0.205, indicating
that the grade is not as well balanced as it is for the overall sample. Of the 224
students in our overall sample, 124 (56 (55%) in the treatment group and 68 (56%)
in the control group) took at least one exam in the main examination period of the
winter semester, half a year after the end of the program. These grades constitute
our long-term performance outcomes. Table A.3 shows that this reduced sample
was also balanced at baseline.
The follow-up questionnaire was answered by 94 students in the treatment group
(92%) and 93 students in the control group (76%) after a maximum of four per-
sonalized reminders were sent by email. Despite this di↵erence, we find no bias in
22
Note that because we only have a few clusters in the treatment group and no clusters in
the control group, these specifications do not take into account potential dependencies between
observations belonging to the same meditation group. Importantly, however, group assignment
was independent of exams taken.

24
attrition. Table A.4 shows that the treatment and control groups in this reduced
sample were balanced at baseline, except for the dummy indicating that a student
was in their 4th or higher year of study, where the proportion is somewhat larger
in the treatment group, significant at the 10% level.
All students of the Faculty of Management, Economics and Social Sciences of the
University of Cologne were eligible to apply for a place in the meditation course.
Table 2 compares our experimental sample with the overall student body of the
faculty in order to shed light on whether certain students were more likely to se-
lect into the experiment. As the table shows, our sample, at 24 years of age on
average, is the same age as the average student of the faculty. In our sample,
the proportion of women, at 54%, tends to be slightly higher than for the faculty
overall (45%). The proportion of students pursuing di↵erent education programs
matches approximately the distribution of students across the di↵erent programs
at the overall faculty. Only the proportion of students matriculating into the BSc
in Social Sciences program tends to be somewhat higher in our sample than for
the faculty overall (19% vs. 13%), while the proportion of students in the BSc in
Business Administration tends to be somewhat lower in our sample than at the
overall faculty (21% vs. 31%).
Furthermore, we find that students in our sample are moderately stressed, with
an average score on the PSS of 21 (over a range of 0 to 40). The latter compares
very well with the average level of stress of the German student population: in
a representative sample of 18,000 university students in Germany in 2016, Herbst
et al. (2016) find an average score of 20 on the PSS. We also find that our sample
of students experience mild depression and mild to moderate anxiety – scoring on
average 8.5 on the PHQ-9 and 9.5 on the GAD-7. 23
23
We could not find a study reporting the average depression and anxiety for a representative
sample of the German student population. However, according to the representative sample of
7,500 German university students surveyed by (Kroenke et al., 2009) in 2017, 16% and 17% of
students are categorized as experiencing depression (using the GAD-2) and anxiety (using the
PHQ-2) respectively. This is consistent with our study, where 15% of students experience severe
anxiety (using the GAD-7) and 17%, moderately severe depression (using the PhQ-9).

25
 Table 2: Sample Comparison - Demographics
Faculty Our Sample
Program Students (%) Female (%) Age Students (%) Female (%) Age
BSc Business Administration 31.4 47.0 22.9 21.4 54.2 22.7
BSc Economics 13.3 32.3 23.2 11.2 40.0 23.4
BSc Economics w/ Social Sciences 2.7 44.0 22.4 6.7 33.3 22.3
BSc Social Sciences 12.9 53.0 23.8 18.8 59.5 23.3
BSc Health Economics 4.4 80.6 23.9 4.5 100 24
BSc Information Systems 6.7 19.9 23.4 10.3 34.8 24.1
MSc Busines Administration 11.0 46.6 25.3 10.7 54.2 25.3
MSc Economics 5.8 32.9 25.8 7.6 64.7 25.8
MSc Political Science 3.8 50.2 26.6 4 77.8 28.2
MSc Other Program 8.2 52.2 26.6 4.9 45.5 26.6
Overall 100 45.1 24.0 100 53.6 24.0
N 8181 224
Note: This table compares the proportion of students overall, the proportion of women, and the mean age in di↵erent
fields of study in the experimental sample and the universe of students at the Social Science Faculty of the University
of Cologne.

Overall, this section shows that while not all students in our sample answered the
endline questionnaire and that there are many for whom we do not observe a grade
for a specific exam period, this attrition does not seem to follow any systematic pat-
tern, and the sample remains balanced. Furthermore, our sample’s characteristics
closely match those of the faculty’s student body, indicating that the meditation
program attracted students of di↵erent genders, ages, and programs of study al-
most equally, speaking to the generalizability of our results for students of this
social science faculty, one of the largest of its kind in Germany. We also find that
our sample compares well with the German student population in terms of mental
health speaking also in favor of the generalizability of our results for students in
Germany more generally.

6.2 E↵ects of Mindfulness Meditation on Secondary Out-

comes

6.2.1 E↵ects on mental health

We first examine whether our mindfulness meditation training improves mental

health, the original intent of such programs. Table 3 presents OLS regressions

26
with stress, anxiety, and depression as dependent variables. All models include the
binary treatment variable, Treat, taking the value of 1 if a student was assigned
to the treatment group and 0 if the student was assigned to the control group.
Stress, anxiety, and depression are reduced between 0.4 and 0.7 standard deviations
(sd). All results are significant at the 1% level and very similar across the value-
added and the first-di↵erence specifications. 24 Columns 1-3 in Table C.1 show
that models without control variables and merely comparing the endline scores
give very similar results. Overall, these results show that, consistent with previous
studies, the mindfulness training improves students’ mental health and thus o↵er a
validation of the e↵ectiveness of our intervention.

Table 3: Mental Health Outcomes

(1) (2) (3) (4) (5) (6)

Stress (VA) Stress (FD) Anxiety (VA) Anxiety (FD) Depression (VA) Depression (FD)
Treat -0.663⇤⇤⇤ -0.704⇤⇤⇤ -0.439⇤⇤⇤ -0.448⇤⇤⇤ -0.602⇤⇤⇤ -0.592⇤⇤⇤
(0.124) (0.144) (0.119) (0.132) (0.114) (0.126)
Stress (BL) 0.481⇤⇤⇤
(0.070)
Anxiety (BL) 0.605⇤⇤⇤
(0.065)
Depression (BL) 0.579⇤⇤⇤
(0.065)
Prior Grade -0.055 0.054 0.010
(0.072) (0.070) (0.064)
Prior ECTS Credits -0.002 -0.002 -0.002
(0.002) (0.002) (0.002)
Constant 0.224 0.293⇤⇤⇤ 0.411 0.188⇤ 0.521 0.270⇤⇤⇤
(0.346) (0.104) (0.329) (0.100) (0.330) (0.084)
Study Program FE Yes No Yes No Yes No
Study Year FE Yes No Yes No Yes No
R2 0.371 0.115 0.453 0.059 0.504 0.107
N 182 187 182 187 182 187
Note: This table shows OLS regressions on mental health outcomes elicited in the survey on the treatment dummy.
Models (1), (3), and (5) use the post-intervention outcome as dependent variable and control for the baseline measure
of the respective outcome, prior mean grade, prior total ECTS credits, study-program fixed e↵ects (9 dummies),
and study-year fixed e↵ects (3 dummies). Models (2), (4), and (6) use the change in outcome (endline baseline)
as dependent variable and include no control variables. All models contain a constant. Heteroskedasticity robust
standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

24
The results remain highly significant even after correcting for multiple hypotheses testing.

27
6.2.2 E↵ects on non-cognitive skills

Table 4 reports coefficient estimates of OLS regressions of non-cognitive skills on

the treatment dummy. Self-control and conscientiousness are both improved by
about 0.5 sd, whereas neuroticism is reduced by 0.4 sd. All results are significant
at the 1% level, and results are very similar across the value-added and the first-
di↵erence specifications. 25 Columns 4-6 in Table C.1 show that models without
control variables and just comparing the endline scores give very similar results.

Table 4: Non-Cognitive Skills Outcomes

(1) (2) (3) (4) (5) (6)
Self-Control (VA) Self-Control (FD) Conscient. (VA) Conscient. (FD) Neuroticism (VA) Neuroticism (FD)
Treat 0.517⇤⇤⇤ 0.490⇤⇤⇤ 0.522⇤⇤⇤ 0.470⇤⇤⇤ -0.353⇤⇤⇤ -0.403⇤⇤⇤
(0.086) (0.089) (0.095) (0.101) (0.112) (0.116)
Self-Control (BL) 0.758⇤⇤⇤
(0.043)
Conscientiousness (BL) 0.711⇤⇤⇤
(0.047)
Neuroticism (BL) 0.732⇤⇤⇤
(0.058)
Prior Grade -0.033 0.075 -0.068
(0.049) (0.054) (0.062)
Prior ECTS Credits 0.003⇤ 0.002⇤ -0.001
(0.001) (0.001) (0.002)
Constant -0.345 -0.210⇤⇤⇤ -0.680⇤⇤⇤ -0.208⇤⇤⇤ 0.179 0.150⇤
(0.252) (0.061) (0.254) (0.070) (0.244) (0.079)
Study Program FE Yes No Yes No Yes No
Study Year FE Yes No Yes No Yes No
R2 0.701 0.140 0.655 0.104 0.537 0.061
N 182 187 182 187 182 187
Note: This table shows OLS regressions of non-cognitive skills outcomes elicited in the survey on the treatment dummy. Models (1), (3),
and (5) use the post-intervention outcome as dependent variable and control for the baseline measure of the respective outcome, prior mean
grade, prior total ECTS credits, study-program fixed e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies). Models (2), (4), and
(6) use the change in outcome (endline baseline) as dependent variable and include no control variables. All models include a constant.
Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

6.2.3 E↵ects on concentration

Table 5 reports coefficient estimates of OLS regressions of the incentivized Stroop

task, our measure of concentration. The treatment improves performance in the
Stroop task by about 0.2 standard deviations. This e↵ect is not significant in
the value-added specification but is marginally significant in the first-di↵erence
specification. Column 7 in Table C.1 shows that models without control variables
and just comparing the endline scores show a very small and insignificant e↵ect,
25
The results remain highly significant even after correcting for multiple hypotheses testing.

28
suggesting that with respect to this outcome, controlling for imbalances at baseline
and comparing changes rather than endline scores yields di↵erent results.

Table 5: Stroop Task

(1) (2)
Stroop Task (VA) Stroop Task (FD)
Treat 0.174 0.228⇤
(0.120) (0.136)
Stroop Task (BL) 0.546⇤⇤⇤
(0.077)
Prior Grade 0.058
(0.077)
Prior ECTS Credits -0.001
(0.002)
Constant -0.712⇤⇤ -0.157
(0.289) (0.101)
Study Program FE Yes No
Study Year FE Yes No
R2 0.405 0.015
N 182 187
Note: This table shows OLS regressions of the incentivized Stroop
task elicited in the survey on the treatment dummy. Model (1) uses
the post-intervention Stroop outcome as dependent variable and con-
trols for the baseline Stroop outcome, prior mean grade, prior total
ECTS credits, study-program fixed e↵ects (9 dummies), and study-
year fixed e↵ects (3 dummies). Model (2) uses the change in outcome
(endline baseline) as dependent variable and does not include con-
trol variables. All models include a constant. Heteroskedasticity
robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤
p < .01

6.3 E↵ects of Mindfulness Meditation on Short-Term Aca-

demic Performance

Table 6 presents results from OLS regressions of students’ grades obtained during
the main exam period for the 2019 summer semester immediately after the interven-
tion. Model (1) has the endline grade as dependent variable and includes the binary
treatment variable, Treat. Model (2) repeats this analysis but additionally controls
for prior performance, as well as study-program and study-year fixed e↵ects, making

29
this a value-added model. Model (3) has the first-di↵erenced grade (endline grade
– baseline grade) as dependent variable and includes the binary treatment vari-
able, making this an individual-level fixed-e↵ects model that only detects changes
within individuals. While Model (1) gives a treatment e↵ect of -0.31 sd (p=0.035),
Model (2) gives a treatment e↵ect of -0.26 sd (p=0.055), and Model (3) gives an in-
significant treatment e↵ect of -0.10 sd (p=0.491). 26 While all three models suggest
that the e↵ect of the meditation program on short-term academic performance was
negative, controlling for imbalances in prior performance with increasing strictness
reduces the treatment e↵ect size to the degree that the null hypothesis of no e↵ect
can no longer be rejected. We interpret these findings as evidence that, if anything,
the program harmed students’ short-term performance.

6.4 E↵ects of Mindfulness Meditation on Long-Term Aca-

demic Performance

We next investigate the e↵ects of the intervention on long-term academic perfor-

mance. Table 7 presents results from OLS regressions of students’ grades obtained
during the main exam period of the 2019/20 winter semester, half a year after the
intervention. The table, again, shows results for a model with the endline grade as
dependent variable and without any control variables, a value-added model addi-
tionally controlling for prior performance and study-year and study-program fixed
e↵ects, and a model with the first-di↵erenced grade as the dependent variable and
no further control variables. All three models show that the meditation intervention
had a significantly positive e↵ect on long-term academic performance. While Model
(1) gives a treatment e↵ect of 0.28 sd (p=0.099), Model (2) gives a treatment e↵ect
of 0.39 sd (p=0.017), and Model (3) gives a treatment e↵ect of 0.45 sd (p=0.010). 27
26
Ex-post power calculations on the specifications without controls reveal that Models (1) and
(3) reach a power of 66% and 17%, respectively. We are thus underpowered to determine whether
the insignificant coefficient of -0.102 on the treatment dummy in Model (3) is a real null e↵ect.
27
Ex-post power calculations on the specifications without controls reveal that Models (1) and
(3) reach a power of 47% and 78%, respectively.

30
 Table 6: Grade (short term)

(1) (2) (3)

Grade (E) Grade (VA) Grade (FD)
Treat -0.308⇤⇤ -0.260⇤ -0.102
(0.145) (0.135) (0.148)
Prior Grade 0.458⇤⇤⇤
(0.080)
Prior ECTS Credits 0.005⇤
(0.002)
Constant 0.143 0.003 0.044
(0.095) (0.683) (0.094)
Study Program FE No Yes No
Study Year FE No Yes No
R2 0.025 0.294 0.003
N 181 177 177
Note: This table shows OLS regressions of the mean grade of the main
summer-semester exam period on the treatment dummy. Model (1) does
not include any control variables. Model (2) uses summer-semester grade
as dependent variable and controls for prior mean grade, prior total ECTS
credits, study-program fixed e↵ects (9 dummies), and study-year fixed ef-
fects (3 dummies). Model (3) uses the change in grade (summer-semester
grade baseline grade) as dependent variable and does not include any
control variables. All models include a constant. Heteroskedasticity ro-
bust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

We interpret these findings as robust evidence that the program positively a↵ected
students’ long-term performance.
Finally, similar results emerge when looking at the distributions of changes in
short-term and long-term grades compared with the baseline (see Figure 1). The
exact p-value for the combined Kolmogorov-Smirnov (KS) test for the left-hand
panel (delta grade short) of Figure 1 is 0.148. The exact p-value for the combined
KS test for the right-hand panel (delta grade long) is 0.097.

31
 Table 7: Grade (long-term)

(1) (2) (3)

Grade (E) Grade (VA) Grade (FD)
Treat 0.284⇤ 0.386⇤⇤ 0.446⇤⇤⇤
(0.171) (0.159) (0.170)
Prior Grade 0.515⇤⇤⇤
(0.105)
Prior ECTS Credits 0.003
(0.002)
Constant -0.128 -0.152 -0.159
(0.121) (0.644) (0.122)
Study Program FE No Yes No
Study Year FE No Yes No
R2 0.022 0.318 0.053
N 124 121 121
Note: This table shows OLS regressions of the mean grade of the main
winter-semester exam period on the treatment dummy. Model (1) does
not include any control variables. Model (2) uses the winter-semester
grade as dependent variable and controls for prior mean grade, prior total
ECTS credits, study-program fixed e↵ects (9 dummies), and study-year
fixed e↵ects (3 dummies). Model (3) uses the change in grade (winter-
semester grade baseline grade) as dependent variable and does not in-
clude any control variables. All models include a constant. Heteroskedas-
ticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤
p < .01

32
 Figure 1: Distributions of changes in short-term and long-term grade

Notes: The left-hand panel shows changes in the the short-term and the right-hand panel
shows changes in the long-term grade by treatment group. Grades are non-standardized.

6.5 E↵ects on other performance measures: Medium-run

grades and ECTS study points

Besides the two main exam periods which constitute our performance outcomes of
primary interest, some students took exams after the 2019 summer break and as
mid-term exams of the 2019/20 winter semester before the Christmas break. The
former are either exams that students voluntarily moved to the secondary exam
session of the summer semester so as to spread the burden of exam writing across a
longer period, which is possible for some of the larger exams written in the faculty,
or retaken exams that students had previously failed. The latter are exams written
in some classes that last only half a semester. In both cases, we only observe small
samples of 91 and 41 students, respectively, whose grades are not well balanced. As
shown in Table C.2, value-added (columns 2 and 5) and first-di↵erence (columns
3 and 6) models give positive e↵ect sizes between 0.21 and 0.38 sd. When all
medium-term grades of the two exam periods are combined in a first-di↵erence
model (column 9), we get a positive e↵ect size of 0.34 sd, significant at the 5%
level. This finding is consistent with the results we find for the long-term grade
outcomes, which give us very similar e↵ect sizes. It suggests that the positive e↵ect

33
of the meditation intervention on performance already exists several weeks after the
end of the intervention.
Next, we check whether the intervention had any e↵ect on the “quantity” of
exams taken, measured as the number of ECTS study points taken by a student
during each exam period. ECTS study points indicate the workload of a module in
which a student is writing an exam and determines its weight in the student’s overall
grade. Thus, the higher the total number of ECTS study points completed during an
exam period, the higher the overall workload completed. Table C.3 shows that the
number of ECTS points is not a↵ected by the treatment, in either the short or the
long run. Thus, the observed quality changes in performance were not accompanied
by quantity changes: the long-term improvement (short-term reduction) in average
grades did not come at the expense (advantage) of completing fewer (more) ECTS
points.
Finally, when combining the short-, medium-, and long-term grades, we find
that they cancel each other out, giving an overall e↵ect that is not significantly
di↵erent from zero (see columns 7-9 in Table C.4).

6.6 Robustness Checks, Heterogeneity Analyses and Role

of Treatment Intensity

As described in section 6.1, the sample sizes of students used in our short-run
vs. long-run analyses do not coincide. Some students finished their studies in the
summer semester and thus took no exams in the following winter semester. Hence,
there are 70 students whose grades we observe in the short run but not in the
long run. Furthermore, a few students did not take any exam during the summer
semester but only later, in the winter semester. Hence, there are 13 students whose
grades we observe in the winter but not in the summer semester. To test whether
our results are influenced by this di↵erence in samples, we run again our short-term
and long-term analyses described in sections 6.3 and 6.4, but this time we restrict

34
the sample to those students for whom we observe both a short-term grade and a
long-term grade. Results are reported in Table C.5. As can be seen, all the results
remain unchanged.
One might wonder whether our e↵ect might be caused by students in the treat-
ment group taking di↵erent exams than the ones originally planned and which we
used for our randomization strategy (see section 3.4). For example, one might
worry that students in the treatment group decided to switch their planned exams
towards more difficult ones in the summer semester and to easier ones in the winter
semester. While we cannot directly test for this conjecture, it is reassuring to know
that we find no treatment di↵erence in (i) the number of planned exams (p = 0.76),
(ii) the number of actual exams taken (p = 0.54), or (iii) the number of exams not
taken: that is, the number of planned exams minus the number of exams actually
taken (p = 0.35). In both the treatment and the control groups, students took on
average about 1 exam less than planned, namely about 1.3 exams instead of 2.4.
Furthermore, note that most courses are mandatory and are only o↵ered once a
year during a specific semester. Therefore, students do not have a lot of freedom to
move them around.
Next, we investigate whether there were any heterogeneous e↵ects of the inter-
vention on short-term and long-term grades by baseline scales or baseline grades –
keeping in mind the limited sample size. Tables C.6 and C.7 report the results. As
can be seen, when performing a median split of the sample, we identify no signif-
icant e↵ects by baseline mental health, concentration (Stroop task), or grade. We
merely find that the negative e↵ect of the intervention on short-term performance
is concentrated on people with below-median conscientiousness in our sample.
Table C.8 investigates heterogeneities in the e↵ect on the scales. We find no sig-
nificant heterogeneities for non-cognitive skills, concentration, or health behavior.
However, we find that the intervention has a significantly more beneficial e↵ect on
people with worse mental health at baseline. Students with above-median stress

35
levels at baseline experience a stress reduction more than three times the size of the
stress reduction experienced by less stressed students (-1.13 sd vs. -0.29 sd, column
1). Students with above-median levels of depression experience a reduction in de-
pression twice the size of the reduction in depression experienced by less depressed
students (-0.84 vs. -0.42, column 3). Results indicate similar heterogeneities for
anxiety, but the latter are not significant.
Next, we test whether the e↵ects may di↵er based on the motivation for applying
to the meditation course. When we ask whether students who applied for a place
in the meditation course because they wanted to improve their performance expe-
rienced di↵erent e↵ects than students who applied because they wanted to improve
their mental and physical health, we find, again, no significant heterogeneities for
either grades or mental health (see Table C.9). We also do not find any significant
heterogeneities by gender (see Table C.10).
Finally, we compare the e↵ect sizes for people who attended the meditation
course more often (i.e., attendance rate above the median) with those for people
who attended the meditation course less often (attendance rate below the median) –
keeping in mind that this is not a clean measure of intensity, as it is also determined
by endogenous factors such as students’ motivation and other obligations. As shown
in Table C.11, we do find some significant di↵erences: the short-term negative e↵ect
on the grade tended to be stronger for students who attended more often, while we
find no significant di↵erences for the long-term positive e↵ect. The improvements
of mental health, health behavior, non-cognitive skills and concentration also all
tended to be concentrated on students who attended the meditation course more
often.

36
7 Channels

We have shown that mindfulness meditation training significantly increased short-

term mental health and non-cognitive skills and marginally increased cognitive skills
as measured by performance in the Stroop task. As explained in the introduction,
aside from stress, all of these variables are unequivocally expected to be, if anything,
positive determinants of academic performance. And, in fact, consistent with these
expectations, we find positive and significant correlations between baseline grades
and self-control, conscientiousness, performance in the Stroop task and a negative
and significant correlation between baseline grades and depression at baseline (Ta-
ble C.12). However, if anything, the intervention marginally decreased short-term
performance. How to reconcile these results?
As argued in the introduction, there could be three (non-mutually) exclusive
explanations for the observed negative short-term e↵ect of the intervention on the
grades. First, it could be that the relationship between stress and performance is
positive in our sample and therefore by reducing the former, the meditation course
also reduced the latter. However, our data speak against this explanation. As
shown in Table C.12, we find no correlation between stress and grades at baseline.
Consistent with this evidence, we also find that the short-term variation in grades
is not associated with the short-term variation in stress (Table 10). Second, it
could be that the intervention changed students’ study behaviour by reducing their
motivation to study and perform well academically. Third, it could be that the
intervention, by requiring students to initially invest time and e↵ort in learning a
new self-care practice and creating new healthy habits, took resources away from
studying in the short run – even though the students remained equally motivated
to perform well academically. We use pre-registered variables to investigate these
two additional potential channels in the sub-sections below. For both channels, we
report the coefficients of the treatment dummies from VA regressions containing

37
the full set of control variables and robust standard errors. 28

7.1 Study Behavior

The questionnaire included 15 questions related to study behavior. Each question

was answered on a scale from 1 to 5 expressing the extent to which the participant
agreed with the statement (1=disagree, 5=agree). These questions were meant to
elicit concentration while studying (5 questions), learning strategies (4 questions),
study self-concept, such as enjoyment and confidence when studying (3 questions)
and exam behavior (3 questions). Following this natural categorization, we created
four indexes of study behavior, each based on the sum of the respective variables,
with a higher index corresponding to better study concentration, strategies, self-
concept, and exam behavior.
Table 8 reports the results. As can be seen, the intervention significantly im-
proved students’ overall concentration, learning strategies and study self-concept
by 0.34, 0.23 and 0.18 standard deviations respectively (p < 0.01). However, we
find no significant e↵ects for exam behavior. 29
Taken together, these results show that, if anything, the intervention positively
a↵ected study behavior suggesting that the reduction in motivation to study and to
perform well academically is probably not the channel for our short-term negative
e↵ects. Furthermore, there is also no reason why such a negative e↵ect, if it was
due to students’ change in motivation, should disappear in the long run. We thus
turn to our third and last candidate channel in explaining our findings.

7.2 Health Behavior and Self-Care Practices

The questionnaire included seven questions related to health behavior and self-
care practices. Given that there is no natural or obvious categorization for these
28
The results do not change when using the first-di↵erence specification. The latter are available
upon request.
29
Note that we did not elicit exam behavior at baseline.

38
 Table 8: Study Behaviors

(1) (2) (3) (4)

Concentration Learning Strategies Study Self-Concept Exam Behavior
Treat 0.335⇤⇤⇤ 0.232⇤⇤⇤ 0.184⇤⇤⇤ 0.158
(0.071) (0.073) (0.068) (0.117)
Prior Grade 0.015 0.017 0.071⇤ 0.187⇤⇤⇤
(0.039) (0.038) (0.042) (0.067)
Prior ECTS Credits 0.002 0.003⇤⇤ 0.002⇤⇤ 0.003
(0.002) (0.001) (0.001) (0.002)
Concentration (BL) 0.477⇤⇤⇤
(0.056)
Learning strategies (BL) 0.655⇤⇤⇤
(0.051)
Study self-concept (BL) 0.667⇤⇤⇤
(0.062)
Constant 0.106 0.090 -0.137 -0.126
(0.222) (0.240) (0.271) (0.180)
Study Program FE Yes Yes Yes Yes
Study Year FE Yes Yes Yes Yes
R2 0.440 0.565 0.568 0.123
N 182 182 182 182
Note: This table shows OLS regressions of study behaviors elicited in the endline survey on the treatment dummy. All
models use the post-intervention outcome as dependent variable and control for the baseline measure of the respective
outcome, prior mean grade, prior total ECTS credits, study-program fixed e↵ects (9 dummies), and study-year fixed e↵ects
(3 dummies). All models include a constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤
p < .05, ⇤⇤⇤ p < .01

questions, we first present the results using each statement separately and then
aggregate them in a single index based on the sum of the scores for each of the
questions, with a higher index corresponding to a healthier behavior and/or to
taking more care of oneself. Table 9 reports the results. Columns 1 to 7 show
that alcohol consumption, smoking, taking medication and rising at the same time
each day are not significantly a↵ected by the intervention. However, students in
the treatment group report drinking less co↵ee or tea to stay awake ( : -0.20,
p = 0.031), seeking significantly more conscious relaxation ( : 0.51, p = 0.000) and
being significantly less likely go to bed late and be tired the next day than students
in the control group ( : -0.44, p = 0.001). Column 8 shows that aggregating all the
items into a single index yields similar results. The intervention improved students’
overall health behavior and self-care practices as measured by an increase in the
index ( : 0.42, p < 0.01).
As already argued in previous sections, establishing new healthy habits and

39
 Table 9: Health Behaviors
(1) (2) (3) (4) (5) (6) (7) (8)
Co↵ee/Tea Alcohol Smoking Medication Getting Up Sleeping Late Relaxation Health Index
Treat -0.203⇤⇤ -0.062 -0.034 -0.102 0.104 -0.435⇤⇤⇤ 0.505⇤⇤⇤ 0.423⇤⇤⇤
(0.093) (0.120) (0.085) (0.152) (0.127) (0.130) (0.126) (0.114)
Prior Grade -0.045 -0.114⇤ -0.023 0.018 -0.001 -0.051 -0.034 0.064
(0.054) (0.067) (0.042) (0.071) (0.067) (0.068) (0.064) (0.059)
Prior ECTS Credits 0.001 -0.001 -0.001 0.001 -0.003 -0.003⇤ 0.001 -0.001
(0.002) (0.002) (0.001) (0.002) (0.002) (0.002) (0.002) (0.002)
Co↵ee/Tea (BL) 0.788⇤⇤⇤
(0.052)
Alcohol (BL) 0.680⇤⇤⇤
(0.063)
Smoking (BL) 0.832⇤⇤⇤
(0.047)
Medication (BL) 0.387⇤⇤⇤
(0.140)
Getting up (BL) 0.530⇤⇤⇤
(0.077)
Sleeping late (BL) 0.504⇤⇤⇤
(0.067)
Relaxation (BL) 0.495⇤⇤⇤
(0.070)
Health index (BL) 0.656⇤⇤⇤
(0.050)
Constant 0.425 0.263 0.130 0.266 0.021 0.694⇤⇤ 0.127 -0.427⇤
(0.356) (0.315) (0.110) (0.468) (0.268) (0.337) (0.379) (0.257)
Study Program FE Yes Yes Yes Yes Yes Yes Yes Yes
Study Year FE Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.652 0.493 0.738 0.179 0.325 0.362 0.399 0.530
N 182 182 182 182 182 182 182 182
Note: This table shows OLS regressions of health behaviors elicited in the endline survey on the treatment dummy. All models use the post-intervention
outcome as dependent variable and control for the baseline measure of the respective outcome, prior mean grade, prior total ECTS credits, study-
program fixed e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies). All models include a constant. Heteroskedasticity robust standard errors
are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

40
learning new self-care practices takes resources (such as time and cognitive e↵ort)
away from studying, which could explain why grades marginally decreased in the
short term despite the improvement in mental health, cognitive and non-cognitive
skills, and study behavior. 30 Consistent with this argument, column 12 of Table
C.11 further shows that the increase in the health behavior index is driven by
students who attended the course more often ( =0.569, p < 0.01). As we have
seen, these are also the students whose grades marginally decreased in the short
term (see columns 1 and 2). Students who rarely attended the training did not
improve their health behavior and did not see any reduction in their short-term
grades.
To shed further light on the validity of this explanation, we also look at the
correlations between changes in short-term grades and changes in all other relevant
(pre-registered) variables in our study. Consistent with our conjecture, columns
1 and 2 in Table 10 show, respectively, that among all pre-registered variables,
variation in short-term grades is only (negatively) associated with variation in the
health behavior index ( =-0.140, p = 0.098) – and in particular, with the variables
a↵ected by the intervention, namely, with relaxing more consciously ( =-0.235,
p = 0.010) and with being less likely to go to bed late ( =0.144, p = 0.050). 31
Thus, overall, the analysis in this sub-section shows that while the intervention
may have improved study behavior, which – if anything – should increase perfor-
mance, it also required learning new self-care (meditation) practices and adopting
a healthier routine, both of which may be competing with studying. In particular,
the tendency to relax more, to drink less tea or co↵ee to stay awake, and to go
30
For instance, instead of spending time relaxing or sleeping, students could be studying. Fur-
thermore, even just the act of making conscious decisions and exerting willpower in establishing
this new routine demands some cognitive e↵ort and, in turn, the expenditure of some inner lim-
ited resources that may become depleted and thus not available for studying. (Muraven and
Baumeister, 2021).
31
An exception is conscientiousness. We find that variation in conscientiousness is negatively
associated with variation in short-term grades (significant in column 2, =-0.249 and p = 0.031,
but not in column 1, =-0.161 and p = 0.110 i). However, we can think of no reasonable argument
for why an increase in conscientiousness should reduce performance.

41
 Table 10: Channels

(1) (2)
Grade (short) Grade (short)
Stress -0.096 -0.081
(0.103) (0.107)
Anxiety 0.065 -0.019
(0.113) (0.127)
Depression 0.102 0.126
(0.109) (0.119)
Conscientiousness -0.161 -0.249⇤⇤
(0.100) (0.114)
Neuroticism -0.011 -0.004
(0.125) (0.123)
Stroop Task -0.027 0.032
(0.085) (0.089)
Study Behavior (Overall) 0.028
(0.129)
Health Behavior (Overall) -0.140⇤
(0.084)
Study Behavior (Concentration) 0.026
(0.085)
Study Behavior (Strategy) 0.178
(0.112)
Study Behavior (Self Concept) -0.010
(0.135)
Health Behavior (Co↵ee/Tea) -0.148
(0.103)
Health Behavior (Alcohol) 0.029
(0.086)
Health Behavior (Smoking) -0.029
(0.106)
Health Behavior (Medication) 0.014
(0.065)
Health Behavior (Getting Up) 0.043
(0.069)
Health Behavior (Sleeping Late) 0.144⇤
(0.073)
Health Behavior (Relaxation) -0.235⇤⇤⇤
(0.090)
Constant -0.017 -0.018
(0.075) (0.076)
R2 0.043 0.108
N 149 149
Note: This table shows OLS regressions of the change in grade in the main summer
semester exam period (summer semester grade baseline grade) on the change in stress,
anxiety, depression, conscientiousness, neutroticism, and performances in the Stroop
task. Model (1) contains the change in the overall study behavior index and the overall
health behavior index as additional regressors, while Model (2) contains the change
in the disaggregated health and study behavior measures as additional regressors. All
models include a constant. Heteroskedasticity robust standard errors are in parentheses.
⇤
p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

42
to bed earlier at night may have reduced the time and e↵ort left for studying and
thereby may have contributed to decreasing academic performance in the short run.
However, as already mentioned in previous sections, most of these negative e↵ects
should be short-lived, because as people repeat actions, the meditation practice
is learned and habits are formed, thereby requiring minimal decision making and
minimal regulatory control (and probably less time) in the long run (Evans and
Stanovich, 2013). This is consistent with our long-term results.

7.3 Channels for the Long-Term E↵ect

We now try to dig deeper into the potential channels that led to the long-term
positive e↵ect of the intervention on students’ grades. The most important channel
we want to rule out is that the observed long-run positive e↵ect is just a student’s
compensatory response for the reduction in grades in the short run compared to
their average grades at baseline. That is, we want to rule our that students in our
treatment group studied harder in the long run in order to make up for the short-
run reduction in grades compared to their average baseline grades. If this were
the case, the positive treatment e↵ect on long-run grades should be concentrated
on students who experienced a reduction in grades in the short run compared to
baseline. To test for this possibility, we look at the e↵ect of the treatment on long-
term grades separately for students whose average short-run grades were higher
than at baseline and for students whose average short-run grades were lower than
at baseline. The results are reported in Table C.13. As can be seen, the opposite is
true: The positive long-run treatment e↵ect is much larger and only significant for
students who already experienced an increase in grades in the short run compared
to baseline. The students who experienced a reduction in grades in the short run
do not experience a significant increase in grades in the long run (although the
coefficient remains positive).
Having ruled out that our positive long-term treatment e↵ect is caused by stu-

43
dents’ compensatory responses to the short-run negative changes, we want to test
whether it is indeed the students who continued practicing mindfulness meditation
after the end of the course who got the highest benefits in terms of higher grades in
the long run. Our problem in investigating this question is that the ex-post ques-
tionnaire was only administered immediately after the intervention ended. Hence,
we do not directly observe our secondary outcomes in the long run. To address this
issue, the best we could do is to look for a variable in our ex-post questionnaire
that might indicate how motivated, or how likely, a student is to continue perform-
ing the meditation practice in the long run, i.e., six months after the end of the
intervention. The following item seemed the most reasonable to us. Students were
asked: ”In the last two months, how often have you meditated on your own (i.e.,
beyond or independent of the course exercises)? [treatment group]; In the last two
months, how often have you meditated on your own? [control group].” Students
could answer by choosing one of the following options: 1- (Almost) every day; 2-
(Almost) every week; 3- Sometimes; 4-Never. 32
We conjecture that students who practiced meditation beyond the training re-
quirements are more likely to have continued practicing in the long run and thus to
have benefited from the practice’s e↵ects on their performance. To test this con-
jecture, we divide subjects into those who reported practicing on their own at least
sometimes (71.0%) and those who reported never practicing on their own (29.0%).
We then look at how the interaction between the treatment and having practiced
on one’s own a↵ected grades in the long run. As shown in column 3 in Table
C.14, the treatment e↵ect for those who did not practice on their own is -0.269
(p=0.517) whereas the treatment e↵ect for those who did practice on their own is
0.624 (-0.269+0.893). A joint F-test for coefficient of the treatment dummy and
32
Note that a plausible conjecture would also be that the students who experienced a reduction
in their short-run grades compared with their baseline decided to stop their practice whereas
students who experienced an increase in their short-run grades compared with their baseline
decided to continue practicing and, thereby, benefited the most from the practice in the long run.
To directly test for this conjecture, however, we would need a measure of “own practice” that was
elicited after the students received the information of their short-run grades rather than before.

44
the coefficient of the interaction of the treatment dummy with the own practice
dummy gives a p-value of 0.000. In column 4, the e↵ect for those who did not
practice on their own is -0.310 (p=0.422) whereas for those who did practice, it is
0.659 (-0.310+0.969, joint F-test: p=0.000). This shows that the long-term positive
e↵ect of the intervention on academic performance is driven by those students who
practiced (and most likely continued practicing) on their own, beyond the course’s
instructions.
Finally, columns 1-2 and 5-12 of Table C.14 also report e↵ects by own practice
on short-term grades and on our (short-term) secondary outcomes, respectively. As
can be seen, the interaction term is never significant, suggesting that exercising on
one’s own while also attending the meditation course and doing the course exercises
does not produce any additional benefit. This suggests that having (likely) contin-
ued practicing on one’s own after the end of the intervention rather than having
practiced on one’s own during the intervention likely drives our long-term results.

8 Discussion of External Validity

In this section, we evaluate the external validity of the results of our field experiment
by discussing the SANS conditions (selection, attrition, naturalness, and scalability)
outlined by List (2020). In terms of selection, we have shown that our student
sample compares well with the overall student body of the Faculty of Management,
Economics and Social Sciences of the University of Cologne (one of the largest
universities in Germany) in terms of age, gender, and programs of study. We also
found that our sample of students is comparable to a representative sample of the
German student population in terms of their mental health, speaking in favor of
the representativeness of our sample. Regarding attrition, we did not find that it
was related to treatment status and the samples for which we observe grade and
questionnaire outcomes remain balanced in terms of all relevant baseline variables as

45
discussed in Section 6.1. Considering the naturalness of the setting, our intervention
mirrors the type of meditation courses students are o↵ered at some universities (e.g.,
Cambridge, LSE, MIT or Yale). Carrying out our field experiment at a university
allows us to have a setting that is natural to students. In addition, the stakes of
performing well academically are so high for students that they are unlikely to adapt
their academic behaviour to what they might believe the researchers want them to
do. Hence, our finding that mindfulness meditation can have substantial positive
spillover e↵ects on academic performance but that it may take time for students to
reap these benefits is likely to replicate and generalize to other student populations.
The last consideration for external validity from List’s SANS conditions concerns
scalability. Scaling our intervention to large numbers of students should be quite
straightforward. The MBSR training is highly standardized, can be done in a
class setting and could be made available as part of the curriculum in any typical
university. One important consideration for the implementation of such programs
is the costs. For our experiment, we collaborated with a health insurance company
that hired two certified MBSR teachers who received approximately 5,000 euros each
for teaching eight 60-minute group sessions to three groups each (they also prepared
audio recordings and handouts). This amounts to a price per class of roughly 1,700
euros and a per-student cost of about 100 euros. 33 As the mechanism through which
our intervention a↵ects academic performance is likely individual behavior change
due to learning and applying a new skill individually, no equilibrium e↵ects that
might reduce e↵ectiveness when large populations are treated are to be expected.
Although we o↵er evidence from a university, our research design can be adapted
to firms. MBSR trainings are o↵ered worldwide not only in education but also in
companies and organizations (e.g., Deloitte, Ford, Google and Nike to name a few).
We expect that the impact of such a training on performance in the work con-
text depends on what drives this performance. Performance in firms does not only
33
Note that universities in some countries might also approach health insurance providers to
pay for the course, similarly to what we have done.

46
depend on individual ability but cooperative behavior and likely has a di↵erent
production function than individual academic performance. We know, however,
that collective performance can be raised by improving social skills (Weidmann and
Deming, 2021) which in turn can be promoted by mindfulness meditation (Jones,
2018). This suggests that mindfulness meditation programs could improve team-
work besides individual skills and thus enhance work performance in the long term
to a larger extent than we observe in an academic setting. Indeed, measuring indi-
vidual academic performance is likely to capture a lower bound e↵ect of mindfulness
meditation training on performance in many settings. While there are, of course,
other e↵ects (including negative) susceptible to influence the impact of such pro-
grams on performance in firms, a full account of such e↵ects is beyond the scope of
this paper. We hope that our research design will be replicated in other universities
and in firms around the world to shed light on where mindfulness meditation can
be beneficial not only for mental health but also for performance.

9 Conclusion

To the best of our knowledge, we conducted the first pre-registered and relatively
large randomized controlled trial investigating potential spillovers of a mindfulness
meditation course on academic achievement in higher education. We advertised
the course to all students of the social science faculty in a German university.
Among the students who applied to the program, we randomly assigned 102 to the
treatment group, who were o↵ered a place in the course, and 122 to the control
group, who were not. We also investigated channels through which our intervention
a↵ects grades.
Consistent with previous studies, the intervention improved students’ mental
health by reducing their stress, anxiety and depression levels, and raised students’
non-cognitive skills, such as self-control, conscientiousness and neuroticism. We

47
also find a marginal positive e↵ect on cognitive skills measured by performance in
the incentivized Stroop task. However, our main result reveals that it takes time
for the students to benefit from the positive spillovers of the meditation course on
their performance. Our analysis shows that, if anything, the intervention worsened
students’ academic performance in the short run. We only observe robust significant
positive e↵ects of the intervention on academic performance in the long run, i.e.,
six months after the end of the intervention. In exploring channels for these results,
we show that the intervention significantly improved daily health behavior and self-
care practices (such as relaxing more often and avoiding going to bed late), both of
which may take resources away from studying in the short run.
Taken together, these findings reveal that, ultimately, mindfulness meditation
can have substantial positive spillover e↵ects on academic performance but that
it may take time for students to reap these benefits. As with any other type of
investment, in which costs are borne in the short run in order to gain positive
returns in the long run, a mindfulness meditation training may be initially costly
in terms of performance because it requires students to put time and e↵ort into
learning a new (self-care) practice and creating new (healthy) habits but in later
stages, if one continues practicing, the positive e↵ects of the training are likely to
emerge.
Finally, our findings also suggest that the relationship between mental health
and (academic) performance is more complex than typically assumed: mental health
and performance do not necessarily move in parallel, and, in fact, an improvement
in mental health may even be associated with a worsening in performance in the
short run.
The implications of these findings are relevant for all those educational institu-
tions and organizations that o↵er – or consider o↵ering – mindfulness meditation
training to their students and employees, and in particular to all those organiza-
tions that do so with the dual objective of increasing not only well-being but also

48
performance. More generally, the findings are informative to the policy debate on
whether mindfulness training should become part of the curriculum in schools and
universities. In this respect, the message of this paper is clear: mental health and
performance should be considered separate goals of education that are possibly not
aligned in the short run.

49
References

ACHA (2018). “National college health assessment ii: Undergraduate student refer-
ence group data deport fall 2017.” Report, American College Health Association,
ACHA, Hanover, MD.
Alan, S., Boneva, T., and Ertac, S. (2019). “Ever failed, try again, succeed bet-
ter: Results from a randomized educational intervention on grit.” The Quarterly
Journal of Economics, 134 (3), 1121–1162.
Alem, Y., Behrendt, H., Belot, M., and Bı́ró, A. (2021). “Mind training, stress and
behaviour—a randomised experiment.” PloS one, 16 (11), e0258172.
Almlund, M., LeeDuckworth, A., Heckman, J., and Kautz, T. (2011). “Personality
psychology and economics.” Handbook of the Economics of Education, 4, 1–181.
Angrist, J., Lang, D., and Oreopoulos, P. (2009). “Incentives and services for college
achievement: Evidence from a randomized trial.” American Economic Journal:
Applied Economics, 1 (1), 136–163.
Baranov, V., Bhalotra, S., Biroli, P., and Maselko, J. (2020). “Maternal depression,
women’s empowerment, and parental investment: evidence from a randomized
controlled trial.” American Economic Review, 110 (3), 824–59.
Barbieri, P. N., Giuntella, O., and Saccardo, S. (2021). “Lifestyle and mental health
1 year into covid-19.” Scientific Report, 11 (1), 1–6.
Bellet, C., De Neve, J.-E., and Ward, G. (2019). “Does employee happiness have
an impact on productivity?” Working paper 2019-13, Saı̈d Business School.
Ben-Porath, Y. (1967). “The production of human capital and the life cycle of
earnings.” Journal of Political Economy, 75 (4, Part 1), 352–365.
Bernal-Morales, B., Rodrı́guez-Landa, J. F., and Pulido-Criollo, F. (2015). “Impact
of anxiety and depression symptoms on scholar performance in high school and
university students.” In F. Durbano (Ed.), A Fresh Look at Anxiety Disorders,
vol. 225, chap. 11, IntechOpen.

50
Blattman, C., Jamison, J. C., and Margaret, S. (2017). “Reducing crime and
violence: Experimental evidence from cognitive behavioral therapy in liberia.”
American Economic Review, 107 (4), 1165–1206.
Burgess, S., Metcalfe, R., and Sado↵, S. (2021). “Understanding the response to
financial and non-financial incentives in education: Field experimental evidence
using high-stakes assessments.” Economics of Education Review, 85, 102195.
Cappelen, A. W., Charness, G., Ekström, M., Gneezy, U., and Tungodden, B.
(2017). “Exercise improves academic performance.” Working Paper 1180, Re-
search Institute of Industrial Economics IFN.
Chapell, M. S., Blanding, Z. B., Silverstein, M. E., Takahashi, M., Newman, B.,
Gubi, A., and McCann, N. (2005). “Test anxiety and academic performance in
undergraduate and graduate students.” Journal of Educational Psychology, 97 (2),
268–274.
Charness, G., Le Bihan, Y., and Villeval, M. C. (2022). “Mindfulness training,
cognitive performance and stress.” Working paper.
Chiesa, A., and Serretti, A. (2009). “Mindfulness-based stress reduction for stress
management in healthy people: a review and meta-analysis.” Journal of Alter-
native and Complementary Medicine, 15 (5), 593–600.
Cohen, S., Kamarck, T., and Mermelstein, R. (1983). “A global measure of perceived
stress.” Journal of Health and Social Behavior, 24 (4), 385–396.
Coviello, D., Deserranno, E., Persico, N., and Sapienza, P. (2021). “E↵ect of mood
and worker incentives on workplace productivity.” Working paper.
Di Bartolomeo, G., and Papa, S. (2016). “Does collective meditation foster trust
and trustworthiness in an investment game?” International Review of Economics,
63 (4), 379–392.
Eberth, J., and Sedlmeier, P. (2012). “The e↵ects of mindfulness meditation: A
meta-analysis.” Mindfulness, 3 (3), 174–189.
Evans, J. S. B., and Stanovich, K. E. (2013). “Dual-process theories of higher

51
 cognition: Advancing the debate.” Perspectives on Psychological Science, 8 (3),
223–241.
Gelles, D. (2015). Mindful work: How meditation is changing business from the
inside out. Houghton Mi✏in Harcourt.
Giuntella, O., Hyde, K., Saccardo, S., and Sado↵, S. (2021). “Lifestyle and men-
tal health disruptions during covid-19.” Proceedings of the National Academy of
Sciences, 118 (9).
Goyal, M., Singh, S., Sibinga, E. M., Gould, N. F., Rowland-Seymour, A., Sharma,
R., Berger, Z., Sleicher, D., Maron, D. D., Shihab, H. M., Ranasinghe, P. D.,
Linn, S., Saha, S., Bass, E. B., and Haythornthwaite, J. A. (2014). “Meditation
programs for psychological stress and well-being: a systematic review and meta-
analysis.” JAMA internal medicine, 174 (3), 357–368.
Grobe, T. G., Steinmann, S., and Szecsenyi, J. (2018). “Barmer medical report.”
Tech. rep., Berlin: Barmer.
Hafenbrack, A. C., and Vohs, K. D. (2018). “Mindfulness meditation impairs task
motivation but not performance.” Organizational Behavior and Human Decision
Processes, 147, 1–15.
Hall, P. D. (1999). “The e↵ect of meditation on the academic performance of african
american college students.” Journal of Black Studies, 29 (3), 408–415.
Hanushek, E. A. (1979). “Conceptual and empirical issues in the estimation of
educational production functions.” Journal of Human Resources, 351–388.
Heckman, J. J., and Kautz, T. (2012). “Hard evidence on soft skills.” Labour Eco-
nomics, 19 (4), 451–464.
Heller, S. B., Shah, A. K., Guryan, J., Ludwig, J., Mullainathan, S., and Pollack,
H. A. (2017). “Thinking, fast and slow? some field experiments to reduce crime
and dropout in chicago.” The Quarterly Journal of Economics, 132 (1), 1–54.
Herbst, U., Voeth, M., Eidho↵, A. T., Müller, M., and Stief, S. (2016). “Studieren-
denstress in deutschland–eine empirische untersuchung.” Tech. rep., Berlin:

52
 AOK-Bundesverband.
Hvidman, U., and Sievertsen, H. H. (2019). “High-stakes grades and student be-
havior.” Journal of Human Resources, 56 (3), 821–849.
Hysenbegasi, A., Hass, S. L., and Rowland, C. R. (2005). “The impact of depression
on the academic productivity of university students.” Journal of Mental Health
Policy and Economics, 8 (3), 145–151.
John, A., and Orkin, K. (2022). “Can simple psychological interventions increase
preventive health investment?” Journal of the European Economic Association,
20 (3), 1001–1047.
John, O. P., Donahue, E. M., and Kentle, R. L. (1991). “The big five inventory–
versions 4a and 54.” University of California, Berkeley, Institute of Personality
and Social Research.
Jones, T. M. (2018). “The e↵ects of mindfulness meditation on emotion regulation,
cognition and social skills.” European Scientific Journal, 14 (14), 18–32.
Kabat-Zinn, J. (1994). Wherever you go, there you are: Mindfulness meditation in
everyday life. Hyperion.
Khoury, B., Sharma, M., Rush, S. E., and Fournier, C. (2015). “Mindfulness-based
stress reduction for healthy individuals: A meta-analysis.” Journal of Psychoso-
matic Research, 78 (6), 519–528.
Koedel, C., and Rocko↵, J. E. (2015). “Value-added modeling: A review.” Eco-
nomics of Education Review, 47, 180–195.
Kroenke, K., Spitzer, R. L., Williams, J. B., and Löwe, B. (2009). “An ultra-brief
screening scale for anxiety and depression: the phq–4.” Psychosomatics, 50 (6),
613–621.
Kroenke, K., Spitzer, R. L., and Williams, J. B. W. (2001). “The phq-9: Validity
of a brief depression severity measure.” Journal of General Internal Medicine,
16 (9), 606–613.
Leuven, E., Oosterbeek, H., and van der Klaauw, B. (2010). “The e↵ect of financial

53
 rewards on students’ achievement: evidence from a randomized experiment.”
Journal of the European Economic Association, 8 (6), 1243–1265.
Levitt, S., List, J., and Sado↵, S. (2016). “The e↵ect of performance-based incentives
on educational achievement: Evidence from a randomized experiment.” Working
Paper 22107, National Bureau of Economic Research.
Liang, K.-Y., and Zeger, S. L. (1986). “Longitudinal data analysis using generalized
linear models.” Biometrika, 73 (1), 13–22.
Lindo, J., Sanders, N., and Oreopoulos, P. (2010). “Ability, gender, and performance
standards: Evidence from academic probation.” American Economic Journal:
Applied Economics, 2 (2), 95–117.
List, J. A. (2020). “Non est disputandum de generalizability? a glimpse into the
external validity trial.” Working Paper 27535, National Bureau of Economic Re-
search.
Lyddy, C., Good, D. J., Bolino, M. C., Thompson, P. S., and Stephens, J. P. (2021).
“Where mindfulness falls short.” Harvard Business Review.
McKenzie, D. (2012). “Beyond baseline and follow-up: The case for more t in
experiments.” Journal of development Economics, 99 (2), 210–221.
Mrazek, M. D., Franklin, M. S., Phillips, D. T., Baird, B., and Schooler, J. W.
(2013). “Mindfulness training improves working memory capacity and gre perfor-
mance while reducing mind wandering.” Psychological Science, 24 (5), 776–781.
Muraven, M., and Baumeister, R. (2021). “Self-regulation and depletion of lim-
ited resources:does self-control resemble a muscle?” Psychological Bulletin, 126,
247–259.
Oswald, A. J., Proto, E., and Sgroi, D. (2015). “Happiness and productivity.”
Journal of Labor Economics, 33 (4), 789–822.
Owens, M., Stevenson, J., Hadwin, J. A., and Norgate, R. (2012). “Anxiety and
depression in academic performance: An exploration of the mediating factors of
worry and working memory.” School Psychology International, 33 (4), 433–449.

54
Pascoe, M. C., Hetrick, S. E., and Parker, A. G. (2020). “The impact of stress
on students in secondary school and higher education.” International Journal of
Adolescence and Youth, 25 (1), 104–112.
Schildberg-Hörisch, H., and Wagner, V. (2020). “Chapter 19 - Monetary and non-
monetary incentives for educational attainment: Design and e↵ectiveness.” In
S. Bradley, and C. Green (Eds.), The Economics of Education (Second Edition),
249 – 268, Academic Press.
Shreekumar, A., and Vautrey, P. L. (2022). “Managing emotions: The e↵ects of
online mindfulness meditation on mental health and economic behavior.” Working
Paper.
Spitzer, R. L., Kroenke, K., Williams, J. B., and Löwe, B. (2006). “A brief mea-
sure for assessing generalized anxiety disorder: the gad-7.” Archives of Internal
Medicine, 166 (10), 1092–1097.
Stroop, J. R. (1935). “Studies of interference in serial verbal reactions.” Journal of
Experimental Psychology, 18 (6), 643–662.
Tang, Y. Y., Hölzel, B. K., and Posner, M. I. (2015). “The neuroscience of mind-
fulness meditation.” Nature Reviews Neuroscience, 16 (4), 213–225.
Tangney, J. P., Baumeister, R. F., and Boone, A. L. (2004). “High self-control pre-
dicts good adjustment, less pathology, better grades, and interpersonal success.”
Journal of Personality, 72 (2), 271–324.
Teigen, K. H. (1994). “Yerkes-dodson: A law for all seasons.” Theory Psychology,
4 (4), 525–547.
Trautmann, S., Rehm, J., and Wittchen, H.-U. (2016). “The economic costs of
mental disorders: Do our societies react appropriately to the burden of mental
disorders?” EMBO Reports, 17 (9), 1245–1249.
Vogel, S., and Schwabe, L. (2016). “Learning and memory under stress: implications
for the classroom.” npj Science of Learning, 1 (16011).
Weidmann, B., and Deming, D. J. (2021). “Team players: How social skills improve

55
 team performance.” Econometrica, 89 (6), 2637–2657.
Wood, W., and Rünger (2016). “Psychology of habit.” Annual review of psychology,
67 (1), 289–314.
Yerkes, R. M., and Dodson, J. D. (1908). “The relation of strength of stimulus to
rapidity of habit-formation.” Journal of Comparative Neu- rology Psychology,
71, 291–301.

56
Appendix
A Balance Checks

57
 Table A.1: Balance Checks: Whole Sample
(1) (2) (3) (4)
Control Treat Overall (1) vs. (2),
p-value
Prior Grade (non-inv.), excl. failed 2.286 2.346 2.313 0.451
(0.052) (0.062) (0.040)
Prior Grade (non-inv.), incl. failed 2.665 2.757 2.706 0.457
(0.084) (0.090) (0.061)
Prior ECTS Credits 62.752 59.939 61.493 0.613
(3.669) (4.200) (2.759)
Female 0.541 0.529 0.536 0.863
(0.045) (0.050) (0.033)
Bachelor Student 0.746 0.706 0.728 0.505
(0.040) (0.045) (0.030)
BSc Business Administration 0.213 0.216 0.214 0.963
(0.037) (0.041) (0.027)
BSc Economics 0.115 0.108 0.112 0.871
(0.029) (0.031) (0.021)
BSc Economics w/ Social Sciences 0.082 0.049 0.067 0.328
(0.025) (0.021) (0.017)
BSc Social Sciences 0.180 0.196 0.188 0.765
(0.035) (0.040) (0.026)
BSc Health Economics 0.033 0.059 0.045 0.350
(0.016) (0.023) (0.014)
BSc Information Systems 0.123 0.078 0.103 0.276
(0.030) (0.027) (0.020)
MSc Business Administration 0.107 0.108 0.107 0.975
(0.028) (0.031) (0.021)
MSc Economics 0.082 0.069 0.076 0.709
(0.025) (0.025) (0.018)
MSc Political Science 0.033 0.049 0.040 0.540
(0.016) (0.021) (0.013)
MSc Other Program 0.033 0.069 0.049 0.218
(0.016) (0.025) (0.014)
1st Year Student 0.328 0.363 0.344 0.586
(0.043) (0.048) (0.032)
2nd Year Student 0.385 0.363 0.375 0.730
(0.044) (0.048) (0.032)
3rd Year Student 0.230 0.176 0.205 0.330
(0.038) (0.038) (0.027)
4th Year Student 0.057 0.098 0.076 0.254
(0.021) (0.030) (0.018)
Stress (BL) 21.328 20.578 20.987 0.145
(0.363) (0.355) (0.256)
Anxiety (BL) 9.197 8.784 9.009 0.524
(0.461) (0.441) (0.321)
Depression (BL) 9.492 9.480 9.487 0.987
(0.454) (0.510) (0.338)
Stroop Task (BL) 0.511 0.486 0.500 0.302
(0.016) (0.018) (0.012)
Self-Control (BL) 39.762 39.461 39.625 0.796
(0.781) (0.868) (0.580)
Conscientiousness (BL) 30.508 30.147 30.344 0.634
(0.524) (0.540) (0.376)
Neuroticism (BL) 21.828 21.902 21.862 0.915
(0.489) (0.480) (0.344)
Mindfulness (BL) 26.984 27.471 27.205 0.457
(0.426) (0.502) (0.325)
Study Behavior (BL) 46.893 46.951 46.920 0.957
(0.751) (0.746) (0.530)
Health Behavior (BL) 23.770 24.127 23.933 0.426
(0.301) (0.332) (0.223)
N 122 102 224
Proportion 0.545 0.455 1.000
Note: This table shows balance checks on pre-intervention outcomes for the whole sample
included in the randomization. Column (1) shows the mean value of the respective
variable in the control group. Column (2) shows the mean value of the respective variable
in the treatment group. Column (3) shows the mean value of the respective variable over
both groups. Column (4) shows the p-value from a t-test of the di↵erences of the means
of the respective variable across the treatment and control group. Standard errors are
shown in parentheses.

58
Table A.2: Balance Checks: Sample with Short-Term Grade Outcome
(1) (2) (3) (4)
Control Treat Overall (1) vs. (2),
p-value
Prior Grade (non-inv.), excl. failed 2.303 2.401 2.348 0.273
(0.059) (0.067) (0.044)
Prior Grade (non-inv.), incl. failed 2.638 2.808 2.716 0.205
(0.089) (0.101) (0.067)
Prior ECTS Credits 63.938 63.148 63.576 0.898
(3.988) (4.732) (3.052)
Female 0.526 0.548 0.536 0.770
(0.051) (0.055) (0.037)
Bachelor Student 0.794 0.762 0.779 0.608
(0.041) (0.047) (0.031)
BSc Business Administration 0.227 0.250 0.238 0.716
(0.043) (0.048) (0.032)
BSc Economics 0.124 0.131 0.127 0.885
(0.034) (0.037) (0.025)
BSc Economics w/ Social Sciences 0.082 0.060 0.072 0.553
(0.028) (0.026) (0.019)
BSc Social Sciences 0.206 0.167 0.188 0.500
(0.041) (0.041) (0.029)
BSc Health Economics 0.041 0.060 0.050 0.575
(0.020) (0.026) (0.016)
BSc Information Systems 0.113 0.095 0.105 0.693
(0.032) (0.032) (0.023)
MSc Business Administration 0.093 0.095 0.094 0.955
(0.030) (0.032) (0.022)
MSc Economics 0.093 0.060 0.077 0.406
(0.030) (0.026) (0.020)
MSc Political Science 0.010 0.036 0.022 0.249
(0.010) (0.020) (0.011)
MSc Other Program 0.010 0.048 0.028 0.128
(0.010) (0.023) (0.012)
1st Year Student 0.299 0.345 0.320 0.509
(0.047) (0.052) (0.035)
2nd Year Student 0.423 0.369 0.398 0.465
(0.050) (0.053) (0.036)
3rd Year Student 0.227 0.179 0.204 0.425
(0.043) (0.042) (0.030)
4th Year Student 0.052 0.107 0.077 0.164
(0.023) (0.034) (0.020)
Stress (BL) 21.577 20.560 21.105 0.083
(0.412) (0.408) (0.292)
Anxiety (BL) 9.165 8.500 8.856 0.347
(0.516) (0.469) (0.352)
Depression (BL) 9.155 9.631 9.376 0.522
(0.500) (0.552) (0.370)
Stroop Task (BL) 0.509 0.489 0.500 0.456
(0.018) (0.019) (0.013)
Self-Control (BL) 40.557 38.869 39.773 0.193
(0.876) (0.949) (0.645)
Conscientiousness (BL) 30.701 29.845 30.304 0.305
(0.573) (0.601) (0.415)
Neuroticism (BL) 21.887 21.726 21.812 0.836
(0.556) (0.526) (0.384)
Mindfulness (BL) 27.454 27.167 27.320 0.697
(0.482) (0.560) (0.365)
Study Behavior (BL) 47.938 46.857 47.436 0.351
(0.820) (0.805) (0.577)
Health Behavior (BL) 23.732 24.155 23.928 0.413
(0.349) (0.380) (0.257)
N 97 84 181
Proportion 0.536 0.464 1.000
Note: This table shows balance checks on pre-intervention outcomes for the sub-sample
for whom a grade outcome from the main exam period of the summer semester is ob-
served. Column (1) shows the mean value of the respective variable in the control group.
Column (2) shows the mean value of the respective variable in the treatment group. Col-
umn (3) shows the mean value of the respective variable over both groups. Column (4)
shows the p-value from a t-test of the di↵erences of the means of the respective variable
across the treatment and control group. Standard errors are shown in parentheses.

59
Table A.3: Balance Checks: Sample with Long-Term Grade Outcome (main exam
period winter)
(1) (2) (3) (4)
Control Treat Overall (1) vs. (2),
p-value
Prior Grade (non-inv.), excl. failed 2.386 2.426 2.403 0.707
(0.069) (0.085) (0.053)
Prior Grade (non-inv.), incl. failed 2.774 2.853 2.808 0.634
(0.107) (0.127) (0.082)
Prior ECTS Credits 58.191 56.264 57.347 0.798
(5.054) (5.498) (3.710)
Female 0.559 0.536 0.548 0.799
(0.061) (0.067) (0.045)
Bachelor Student 0.853 0.821 0.839 0.638
(0.043) (0.052) (0.033)
BSc Business Administration 0.250 0.321 0.282 0.383
(0.053) (0.063) (0.041)
BSc Economics 0.132 0.143 0.137 0.867
(0.041) (0.047) (0.031)
BSc Economics w/ Social Sciences 0.074 0.054 0.065 0.656
(0.032) (0.030) (0.022)
BSc Social Sciences 0.235 0.161 0.202 0.307
(0.052) (0.050) (0.036)
BSc Health Economics 0.029 0.071 0.048 0.282
(0.021) (0.035) (0.019)
BSc Information Systems 0.132 0.071 0.105 0.274
(0.041) (0.035) (0.028)
MSc Business Administration 0.029 0.054 0.040 0.500
(0.021) (0.030) (0.018)
MSc Economics 0.059 0.018 0.040 0.252
(0.029) (0.018) (0.018)
MSc Political Science 0.029 0.036 0.032 0.845
(0.021) (0.025) (0.016)
MSc Other Program 0.029 0.071 0.048 0.282
(0.021) (0.035) (0.019)
1st Year Student 0.382 0.393 0.387 0.906
(0.059) (0.066) (0.044)
2nd Year Student 0.324 0.357 0.339 0.697
(0.057) (0.065) (0.043)
3rd Year Student 0.235 0.196 0.218 0.605
(0.052) (0.054) (0.037)
4th Year Student 0.059 0.054 0.056 0.901
(0.029) (0.030) (0.021)
Stress (BL) 21.515 20.714 21.153 0.268
(0.487) (0.528) (0.358)
Anxiety (BL) 9.662 8.875 9.306 0.363
(0.630) (0.564) (0.429)
Depression (BL) 10.059 10.107 10.081 0.958
(0.613) (0.668) (0.450)
Stroop Task (BL) 0.504 0.486 0.496 0.562
(0.022) (0.021) (0.015)
Self-Control (BL) 39.353 39.036 39.210 0.840
(1.013) (1.206) (0.775)
Conscientiousness (BL) 30.574 30.232 30.419 0.755
(0.761) (0.772) (0.542)
Neuroticism (BL) 21.324 21.964 21.613 0.473
(0.648) (0.586) (0.442)
Mindfulness (BL) 26.529 27.339 26.895 0.369
(0.579) (0.697) (0.447)
Study Behavior (BL) 46.544 46.732 46.629 0.899
(1.031) (1.032) (0.730)
Health Behavior (BL) 23.324 23.482 23.395 0.796
(0.409) (0.459) (0.304)
N 68 56 124
Proportion 0.548 0.452 1.000
Note: This table shows balance checks on pre-intervention outcomes for the sub-sample
for whom a grade outcome from the main exam period of the winter semester is observed.
Column 1 shows the mean value of the respective variable in the control group. Column
2 shows the mean value of the respective variable in the treatment group. Column 3
shows the mean value of the respective variable over both groups. Column 4 shows the
p-value from a t-test of the di↵erences of the means of the respective variable across the
treatment and control group. Standard errors are shown in parentheses.

60
Table A.4: Balance Checks: Sample with Post-Intervention Questionnaire
(1) (2) (3) (4)
Control Treat Overall (1) vs. (2),
p-value
Prior Grade (non-inv.), excl. failed 2.275 2.292 2.283 0.846
(0.061) (0.060) (0.043)
Prior Grade (non-inv.), incl. failed 2.579 2.658 2.618 0.523
(0.090) (0.085) (0.062)
Prior ECTS Credits 66.489 62.467 64.500 0.505
(4.124) (4.400) (3.009)
Female 0.559 0.553 0.556 0.935
(0.052) (0.052) (0.036)
Bachelor Student 0.731 0.702 0.717 0.661
(0.046) (0.047) (0.033)
BSc Business Administration 0.204 0.213 0.209 0.887
(0.042) (0.042) (0.030)
BSc Economics 0.118 0.106 0.112 0.798
(0.034) (0.032) (0.023)
BSc Economics w/ Social Sciences 0.108 0.043 0.075 0.092
(0.032) (0.021) (0.019)
BSc Social Sciences 0.183 0.202 0.193 0.739
(0.040) (0.042) (0.029)
BSc Health Economics 0.032 0.064 0.048 0.316
(0.018) (0.025) (0.016)
BSc Information Systems 0.086 0.074 0.080 0.773
(0.029) (0.027) (0.020)
MSc Business Administration 0.118 0.106 0.112 0.798
(0.034) (0.032) (0.023)
MSc Economics 0.086 0.064 0.075 0.567
(0.029) (0.025) (0.019)
MSc Political Science 0.032 0.053 0.043 0.482
(0.018) (0.023) (0.015)
MSc Other Program 0.032 0.074 0.053 0.202
(0.018) (0.027) (0.016)
1st Year Student 0.333 0.351 0.342 0.800
(0.049) (0.049) (0.035)
2nd Year Student 0.376 0.383 0.380 0.926
(0.051) (0.050) (0.036)
3rd Year Student 0.258 0.170 0.214 0.145
(0.046) (0.039) (0.030)
4th Year Student 0.032 0.096 0.064 0.077
(0.018) (0.031) (0.018)
Stress (BL) 21.849 20.734 21.289 0.045
(0.405) (0.375) (0.278)
Anxiety (BL) 9.366 9.011 9.187 0.611
(0.521) (0.465) (0.348)
Depression (BL) 9.860 9.447 9.652 0.588
(0.538) (0.539) (0.380)
Stroop Task (BL) 0.526 0.489 0.508 0.154
(0.018) (0.019) (0.013)
Self-Control (BL) 39.430 39.191 39.310 0.852
(0.897) (0.904) (0.635)
Conscientiousness (BL) 30.280 30.096 30.187 0.825
(0.601) (0.571) (0.413)
Neuroticism (BL) 22.000 22.266 22.134 0.721
(0.559) (0.492) (0.371)
Mindfulness (BL) 26.742 27.255 27.000 0.479
(0.495) (0.528) (0.362)
Study Behavior (BL) 47.097 47.128 47.112 0.979
(0.886) (0.789) (0.591)
Health Behavior (BL) 23.871 24.043 23.957 0.731
(0.351) (0.353) (0.248)
N 93 94 187
Proportion 0.497 0.503 1.000
Note: This table shows balance checks on pre-intervention outcomes for the sub-sample
who answered the post-intervention questionnaire. Column 1 shows the mean value
of the respective variable in the control group. Column 2 shows the mean value of
the respective variable in the treatment group. Column 3 shows the mean value of
the respective variable over both groups. Column 4 shows the p-value from a t-test of
the di↵erences of the means of the respective variable across the treatment and control
group. Standard errors are shown in parentheses.

61
B Figures

Figure B.1: Grades by Date

Note: This figure depicts the distribution of non-aggregated grade outcomes by official
exam date. The x-axis shows the date. The y-axis shows the number of exams for which
we observe an outcome written on a particular day.

62
 Figure B.2: Attendance by Week

Note: This figure depicts the fraction of assigned students attending the meditation train-
ing in each week.

Figure B.3: Attendance by Student

.2

.15

.1

.05

0
0 2 4 6 8
#Participations

Note: This figure depicts the distribution of participations by assigned student.

C Additional Tables

63
 Table C.1: Secondary Outcomes without Controls
(1) (2) (3) (4) (5) (6) (7)
Stress (E) Anxiety (E) Depression (E) Self-Control (E) Conscientiousness (E) Neuroticism (E) Stroop Task (E)
Treat -0.656⇤⇤⇤ -0.522⇤⇤⇤ -0.661⇤⇤⇤ 0.462⇤⇤⇤ 0.437⇤⇤⇤ -0.352⇤⇤ 0.020
(0.139) (0.142) (0.139) (0.143) (0.143) (0.144) (0.147)
Constant 0.330⇤⇤⇤ 0.262⇤⇤ 0.332⇤⇤⇤ -0.232⇤⇤ -0.220⇤⇤ 0.177 -0.010
(0.113) (0.116) (0.110) (0.100) (0.105) (0.108) (0.106)
R2 0.108 0.068 0.110 0.054 0.048 0.031 0.000
N 187 187 187 187 187 187 187
Note: This table shows OLS regressions of the secondary outcomes on the treatment dummy without control variables.
All models include a constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤
p < .01

Table C.2: Grade (medium-term)

Secondary period Summer Mid-Term period Winter Both
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Grade (E) Grade (VA) Grade (FD) Grade (E) Grade (VA) Grade (FD) Grade (E) Grade (VA) Grade (FD)
Treat 0.047 0.276 0.375⇤ -0.160 0.209 0.357 -0.025 0.232 0.343⇤⇤
(0.195) (0.196) (0.194) (0.272) (0.307) (0.274) (0.186) (0.170) (0.171)
Prior Grade 0.475⇤⇤⇤ 0.571⇤⇤⇤ 0.508⇤⇤⇤
(0.111) (0.201) (0.095)
Prior ECTS Credits 0.005 0.001 0.005⇤⇤
(0.003) (0.007) (0.002)
Constant -0.020 -0.918 -0.069 0.071 -0.211 0.006 0.010 -1.205 -0.042
(0.140) (0.881) (0.150) (0.201) (0.637) (0.199) (0.122) (0.818) (0.118)
Study Program FE No Yes No No Yes No No Yes No
Study Year FE No Yes No No Yes No No Yes No
R2 0.001 0.307 0.036 0.008 0.472 0.039 0.000 0.394 0.034
N 94 91 91 43 41 41 110 107 107
Note: This table shows OLS regressions of the mean grade of the medium term on the treatment dummy. Models (1) - (3) show regressions
for the mean grade of the secondary exam period of the summer semester. Models (4) - (6) show regressions for the mean grade of the
mid-term exam period of the winter semester. Models (7) - (9) show regressions for the mean grade over both the secondary exam period of
the summer and the mid-term exam period of the winter semester. Models (1), (4), and (7) use the post-intervention outcome as dependent
variable and do not include any control variables. Models (2), (5), and (8) use the post-intervention outcome as dependent variable and
control for prior mean grade, prior total ECTS credits, study-program fixed e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies).
Models (3), (6), and (9) use the change in mean grade (medium term grade [i.e., secondary summer, mid-term winter, or both] baseline
grade) as dependent variable and do not include any control variables. All models include a constant. Heteroskedasticity robust standard
errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

Table C.3: ECTS Credits

(1) (2) (3) (4)

Credits (short, E) Credits (short, VA) Credits (long, E) Credits (long, VA)
Treat -1.181 -1.053 0.235 0.075
(1.129) (1.098) (1.268) (1.103)
Prior Grade 1.640⇤⇤⇤ 1.326⇤
(0.592) (0.682)
Prior ECTS Credits 0.017 0.010
(0.021) (0.017)
Constant 12.030⇤⇤⇤ 8.285⇤⇤ 10.849⇤⇤⇤ 2.427
(0.787) (3.394) (0.882) (3.037)
Study Program FE No Yes No Yes
Study Year FE No Yes No Yes
R2 0.006 0.265 0.000 0.364
N 186 182 132 129
Note: This table shows OLS regressions of the total ECTS credits acquired during a given exam period on the treatment
dummy. Models (1) and (2) show regressions for the total ECTS credits acquired during the main summer semester exam
period. Models (3) and (4) show regression for the total ECTS credits acquired during the main winter semester exam period.
Models (1) and (3) use the ECTS credits acquired during the given exam period as dependent variable and do not contain any
control variables. Models (2) and (4) use the ECTS credits acquired during the given exam period as dependent variable and
control for prior mean grade, prior total ECTS credits, study program fixed e↵ects (9 dummies), and study year fixed e↵ects
(3 dummies). All models contain a constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05,
⇤⇤⇤
p < .01

64
 Table C.4: Grade (overall)
all exams summer all exams winter all exams summer+winter
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Grade (E) Grade (VA) Grade (FD) Grade (E) Grade (VA) Grade (FD) Grade (E) Grade (VA) Grade (FD)
Treat -0.249⇤ -0.169 -0.044 0.227 0.341⇤⇤ 0.390⇤⇤ -0.089 -0.014 0.076
(0.140) (0.126) (0.137) (0.165) (0.151) (0.161) (0.136) (0.111) (0.124)
Prior Grade 0.468⇤⇤⇤ 0.532⇤⇤⇤ 0.524⇤⇤⇤
(0.074) (0.092) (0.077)
Prior ECTS Credits 0.006⇤⇤⇤ 0.005⇤⇤ 0.007⇤⇤⇤
(0.002) (0.002) (0.002)
Constant 0.114 -0.334 -0.012 -0.099 -0.587 -0.116 0.039 -0.445 -0.054
(0.091) (0.595) (0.085) (0.118) (0.513) (0.110) (0.094) (0.361) (0.078)
Study Program FE No Yes No No Yes No No Yes No
Study Year FE No Yes No No Yes No No Yes No
R2 0.016 0.334 0.001 0.014 0.349 0.043 0.002 0.424 0.002
N 197 193 193 133 130 130 206 202 202
Note: This table shows OLS regressions of the mean grade by semester. Models (1) - (3) show regressions for the mean grade over the
main and secondary summer-semester exam period. Models (4) - (6) show regressions for the mean grade over the mid-term and main
winter-semester exam period. Models (7) - (9) show regressions for the mean grade over the whole time period. Models (1), (4), and (7) use
the post-intervention outcome as dependent variable and do not include control variables. Models (2), (5), and (8) use the post-intervention
outcome as dependent variable and control for prior mean grade, prior total ECTS credits, study-program fixed e↵ects (9 dummies), and
study-year fixed e↵ects (3 dummies). Models (3), (6), and (9) use the change in mean grade (medium term grade [i.e., all summer exams,
all winter exams, or both] baseline grade) as dependent variable and do not include control variables. All models contain a constant.
Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

Table C.5: Robustness checks with matching samples for short- and long-term grade
outcomes
(1) (2) (3) (4) (5) (6)
Grade (short, E) Grade (short, VA) Grade (short, FD) Grade (long, E) Grade (long, VA) Grade (long, FD)
Treat -0.378⇤⇤ -0.339⇤ -0.228 0.167 0.304⇤⇤ 0.332⇤⇤
(0.181) (0.177) (0.188) (0.180) (0.153) (0.165)
Prior Grade 0.527⇤⇤⇤ 0.593⇤⇤⇤
(0.095) (0.097)
Prior ECTS Credits 0.001 0.005
(0.003) (0.003)
Constant 0.096 0.259 0.101 -0.035 -0.776 -0.030
(0.121) (0.439) (0.122) (0.127) (0.783) (0.110)
Study Program FE No Yes No No Yes No
Study Year FE No Yes No No Yes No
R2 0.039 0.329 0.014 0.008 0.428 0.037
N 111 108 108 111 108 108
Note: This table shows OLS regressions of the mean grade on the treatment dummy. Models (1) - (3) show regressions for the mean grade
over the main summer-semester exam period, given that a grade for the main winter-semester exam period is observed for the student. Models
(4) - (6) show regressions for the mean grade over the main winter-semester exam period, given that a grade for the main summer-semester
exam period is observed for the student. Models (1) and (4) use the post-intervention outcome as dependent variable and do not include
control variables. Models (2) and (5) use the post-intervention outcome as dependent variable and control for prior mean grade, prior total
ECTS credits, study-program fixed e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies). Models (3) and (6) use the change in mean
grade (medium term grade [i.e., main exam period summer or winter] baseline grade) as dependent variable and do not include control
variables. All models include a constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

65
 Table C.6: Short-Term Grades by Baseline Grade and Scales
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Treat -0.244 -0.214 -0.243 -0.244 -0.497⇤⇤ -0.283 -0.088 -0.205 -0.312
(0.174) (0.189) (0.183) (0.205) (0.196) (0.176) (0.193) (0.203) (0.192)
TreatXHigh Stress -0.054
(0.282)
High Stress -0.029
(0.172)
TreatXHigh Anxiety -0.093
(0.288)
High Anxiety 0.063
(0.173)
TreatXHigh Depression -0.015
(0.273)
High Depression -0.164
(0.169)
TreatXHigh Self-control -0.032
(0.274)
High Self-control 0.025
(0.172)
TreatXHigh Conscientiousness 0.495⇤
(0.260)
High Conscientiousness -0.287⇤
(0.162)
TreatXHigh Neuroticism 0.054
(0.267)
High Neuroticism -0.201
(0.161)
TreatXHigh Stroop Task -0.338
(0.262)
High Stroop task 0.308⇤
(0.185)
TreatXHigh Prior Grade -0.115
(0.295)
High Prior Grade -0.005
(0.260)
TreatXHigh Heath Index 0.095
(0.264)
High Health Index 0.026
(0.175)
Prior Grade 0.455⇤⇤⇤ 0.457⇤⇤⇤ 0.445⇤⇤⇤ 0.456⇤⇤⇤ 0.463⇤⇤⇤ 0.468⇤⇤⇤ 0.440⇤⇤⇤ 0.481⇤⇤⇤ 0.458⇤⇤⇤
(0.082) (0.081) (0.081) (0.080) (0.080) (0.077) (0.080) (0.133) (0.080)
Prior ECTS Credits 0.005⇤ 0.005⇤ 0.004⇤ 0.005⇤ 0.005⇤⇤ 0.004⇤ 0.005⇤⇤ 0.005⇤ 0.005⇤
(0.002) (0.003) (0.003) (0.003) (0.002) (0.002) (0.002) (0.003) (0.002)
Constant 0.018 -0.052 0.125 -0.013 0.107 0.125 -0.219 -0.020 -0.046
(0.681) (0.426) (0.706) (0.711) (0.417) (0.699) (0.427) (0.448) (0.406)
Study Program FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Study Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.295 0.295 0.301 0.294 0.310 0.302 0.306 0.295 0.296
N 177 177 177 177 177 177 177 177 177
Note: This table shows OLS regressions of the mean grade of the main exam period of the summer semester on the treatment
dummy that control for prior mean grade, prior total ECTS credits, study-program fixed e↵ects (9 dummies), and study-year
fixed e↵ects (3 dummies). Each model additionally contains a dummy equal to 1 for those students who scored above the
sample median on the respective baseline variable (mental health outcomes, non-cognitive skills, Stroop task, prior grade,
health behavior index) and an interaction term of the respective dummy and the treatment dummy. All models include a
constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

66
 Table C.7: Long-term grades by baseline and scales
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Treat 0.370⇤ 0.367 0.213 0.402⇤ 0.475⇤⇤ 0.210 0.391⇤ 0.289 0.233
(0.196) (0.237) (0.238) (0.228) (0.220) (0.195) (0.225) (0.234) (0.226)
TreatXHigh Stress 0.040
(0.323)
High Stress 0.038
(0.238)
TreatXHigh Anxiety 0.023
(0.317)
High Anxiety -0.074
(0.223)
TreatXHigh Depression 0.302
(0.352)
High Depression -0.152
(0.257)
TreatXHigh Self-control -0.052
(0.338)
High Self-control 0.350
(0.279)
TreatXHigh Conscientiousness -0.243
(0.314)
High Conscientiousness 0.559⇤⇤
(0.250)
TreatXHigh Neuroticism 0.425
(0.330)
High Neuroticism -0.306
(0.251)
TreatXHigh Stroop Task -0.011
(0.311)
High Stroop task 0.039
(0.233)
TreatXHigh Prior Grade 0.221
(0.337)
High Prior Grade -0.128
(0.299)
TreatXHigh Heath Index 0.288
(0.320)
High Health Index 0.139
(0.237)
Prior Grade 0.522⇤⇤⇤ 0.516⇤⇤⇤ 0.528⇤⇤⇤ 0.486⇤⇤⇤ 0.452⇤⇤⇤ 0.516⇤⇤⇤ 0.513⇤⇤⇤ 0.528⇤⇤⇤ 0.520⇤⇤⇤
(0.109) (0.106) (0.111) (0.109) (0.109) (0.106) (0.110) (0.170) (0.104)
Prior ECTS Credits 0.003 0.003 0.003 0.001 0.001 0.003 0.003 0.003 0.002
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
Constant -0.136 -0.084 0.029 -0.156 -0.122 0.036 -0.173 -0.092 0.002
(0.656) (0.691) (0.708) (0.605) (0.722) (0.673) (0.669) (0.653) (0.634)
Study Program FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Study Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.319 0.319 0.323 0.338 0.361 0.332 0.318 0.321 0.339
N 121 121 121 121 121 121 121 121 121
Note: This table shows OLS regressions of the mean grade of the main exam period of the winter semester on the treatment
dummy that control for prior mean grade, prior total ECTS credits, study-program fixed e↵ects (9 dummies), and study-year
fixed e↵ects (3 dummies). Each model additionally contains a dummy equal to 1 for those students who scored above the
sample median on the respective baseline variable (mental health outcomes, non-cognitive skills, Stroop task, prior grade,
health behavior index) and an interaction term of the respective dummy and the treatment dummy. All models include a
constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

67
 Table C.8: Heterogeneity Analysis of Secondary Outcomes
(1) (2) (3) (4) (5) (6) (7) (8)
Stress Anxiety Depression Self-Control Conscientiousness Neuroticism Stroop Task Health Index
Treat -0.292⇤ -0.245 -0.417⇤⇤⇤ 0.459⇤⇤⇤ 0.512⇤⇤⇤ -0.212 0.298 2.045⇤⇤⇤
(0.168) (0.176) (0.133) (0.164) (0.149) (0.188) (0.227) (0.628)
TreatXHigh Stress -0.842⇤⇤⇤
(0.278)
High Stress 1.050⇤⇤⇤
(0.203)
TreatXHigh Anxiety -0.407
(0.263)
High Anxiety 1.239⇤⇤⇤
(0.200)
TreatXHigh Depression -0.425⇤
(0.255)
High Depression 1.184⇤⇤⇤
(0.202)
TreatXHigh Self-control 0.082
(0.220)
High Self-control 1.250⇤⇤⇤
(0.158)
TreatXHigh Conscientiousness -0.143
(0.216)
High Conscientiousness 1.245⇤⇤⇤
(0.163)
TreatXHigh Neuroticism -0.300
(0.253)
High Neuroticism 1.317⇤⇤⇤
(0.177)
TreatXHigh Stroop Task -0.292
(0.286)
High Stroop task 0.899⇤⇤⇤
(0.213)
TreatXHigh Heath Index -1.065
(0.913)
High Health Index 3.920⇤⇤⇤
(0.668)
Prior Grade -0.093 0.011 -0.051 0.014 0.166⇤⇤ -0.128⇤ 0.087 0.298
(0.074) (0.076) (0.070) (0.064) (0.065) (0.074) (0.091) (0.237)
Prior ECTS Credits -0.004 -0.002 -0.001 0.002 0.002 0.001 -0.001 -0.003
(0.003) (0.002) (0.003) (0.002) (0.002) (0.003) (0.002) (0.007)
Constant -0.097 -0.162 -0.213 -1.410⇤⇤⇤ -1.223⇤⇤⇤ -0.473 -1.165⇤⇤⇤ -3.284⇤⇤⇤
(0.370) (0.360) (0.452) (0.314) (0.358) (0.337) (0.382) (1.127)
Study Program FE Yes Yes Yes Yes Yes Yes Yes Yes
Study Year FE Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.294 0.378 0.401 0.559 0.524 0.362 0.276 0.365
N 182 182 182 182 182 182 182 182
Note: This table shows OLS regressions of the secondary outcomes elicited in the post-intervention questionnaire on the treatment dummy
that control for prior mean grade, prior total ECTS credits, study-program fixed e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies).
Each model additionally contains a dummy equal to 1 for those students who scored above the sample median on the respective baseline
variable (mental health outcomes, non-cognitive skills, Stroop task, prior grade, health behavior index) and an interaction term of the
respective dummy and the treatment dummy. All models include a constant. Heteroskedasticity robust standard errors are in parentheses. ⇤
p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

68
 Table C.9: Outcomes by Motivation
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Grade (s, VA) Grade(s, D) Grade (l, VA) Grade (l, D) Stress Anxiety Depression Self-Control Conscient. Neurot. Stroop Health B.
⇤⇤ ⇤⇤ ⇤⇤⇤ ⇤⇤⇤ ⇤⇤⇤ ⇤⇤⇤ ⇤⇤⇤ ⇤⇤
Treat -0.262 -0.171 0.491 0.390 -0.681 -0.416 -0.562 0.485 0.587 -0.328 0.192 0.556⇤⇤⇤
(0.165) (0.174) (0.208) (0.194) (0.165) (0.147) (0.144) (0.117) (0.124) (0.136) (0.157) (0.151)
TreatXPerf-motivated 0.001 0.151 -0.222 0.106 0.083 -0.009 -0.109 0.088 -0.144 -0.025 -0.005 -0.341
(0.282) (0.232) (0.316) (0.234) (0.272) (0.264) (0.244) (0.171) (0.193) (0.249) (0.252) (0.225)
Perf-motivated 0.030 0.070 -0.140 -0.148 0.060 -0.049 -0.008 -0.100 -0.107 0.138
(0.182) (0.224) (0.234) (0.210) (0.210) (0.137) (0.157) (0.195) (0.199) (0.178)
Prior Grade 0.459⇤⇤⇤ 0.507⇤⇤⇤ -0.065 0.040 0.011 -0.034 0.070 -0.077 0.047 0.064
(0.081) (0.107) (0.074) (0.069) (0.066) (0.049) (0.054) (0.063) (0.077) (0.060)
Prior ECTS Credits 0.005⇤ 0.003 -0.002 -0.002 -0.002 0.003⇤⇤ 0.002⇤ -0.001 -0.001 -0.001
(0.002) (0.002) (0.002) (0.002) (0.002) (0.001) (0.001) (0.002) (0.002) (0.002)
Stress (BL) 0.476⇤⇤⇤
(0.072)
Anxiety (BL) 0.594⇤⇤⇤
(0.066)
Depression (BL) 0.577⇤⇤⇤
(0.065)
Self-Control (BL) 0.757⇤⇤⇤
(0.046)
Conscient. (BL) 0.705⇤⇤⇤
(0.050)
Neurot. (BL) 0.724⇤⇤⇤
(0.060)
Stroop Task (BL) 0.553⇤⇤⇤
(0.077)
Health B. (BL) 0.649⇤⇤⇤
(0.051)
Constant -0.022 0.044 -0.656 -0.159 0.269 0.456 0.501 -0.329 -0.682⇤⇤ 0.208 -0.312 -0.478⇤
(0.692) (0.095) (0.647) (0.123) (0.348) (0.319) (0.341) (0.262) (0.268) (0.246) (0.238) (0.256)
Study Program FE Yes No Yes No Yes Yes Yes Yes Yes Yes Yes Yes
Study Year FE Yes No Yes No Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.295 0.006 0.321 0.054 0.373 0.457 0.505 0.702 0.657 0.540 0.407 0.537
N 177 177 121 121 182 182 182 182 182 182 182 182
Note: This table shows OLS regressions of the grades and secondary outcomes on the treatment dummy. Models (1) and (3) use the mean grade of the main summer- and
winter-semester exam periods, respectively, as dependent variables. Models (2) and (4) use the change in grade (summer grade - baseline grade and winter grade - baseline grade,
respectively) as dependent variables. Models (5) - (7) use the mental health scales, models (8) - (10) use the non-cognitive skills outcomes, model (11) uses the Stroop task and
model (12) uses the health behavior index as dependent variables. All models except (2) and (4) control for prior mean grade, prior total ECTS credits, study-program fixed
e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies). All models additionally contain a dummy “Perf-motivated” equal to 1 for those students who indicated that they
applied for the meditation course because they wanted to improve their performance and an interaction term of this dummy and the treatment dummy. All models include a
constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

Table C.10: Heterogeneous E↵ects by Gender

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Grade (s, VA) Grade(s, D) Grade (l, VA) Grade (l, D) Stress Anxiety Depression Self-Control Conscient. Neurot. Stroop Health B.
⇤ ⇤⇤ ⇤⇤ ⇤⇤⇤ ⇤⇤⇤ ⇤⇤⇤ ⇤⇤⇤ ⇤⇤⇤ ⇤
Treat -0.344 -0.068 0.473 0.451 -0.585 -0.455 -0.548 0.512 0.557 -0.320 0.181 1.044
(0.205) (0.179) (0.235) (0.210) (0.183) (0.174) (0.164) (0.132) (0.137) (0.176) (0.182) (0.662)
TreatXFemale 0.155 -0.062 -0.142 -0.009 -0.135 0.027 -0.089 0.003 -0.068 -0.054 -0.013 0.830
(0.275) (0.226) (0.306) (0.237) (0.255) (0.236) (0.214) (0.180) (0.178) (0.230) (0.250) (0.802)
Female -0.235 0.264 0.023 -0.019 0.166 -0.124 -0.140 0.095 -0.012 -0.676
(0.171) (0.234) (0.203) (0.189) (0.167) (0.120) (0.126) (0.160) (0.186) (0.547)
Stress (BL) 0.488⇤⇤⇤
(0.079)
Anxiety (BL) 0.606⇤⇤⇤
(0.065)
Depression (BL) 0.565⇤⇤⇤
(0.069)
Self-Control (BL) 0.758⇤⇤⇤
(0.043)
Conscient. (BL) 0.707⇤⇤⇤
(0.047)
Neurot. (BL) 0.722⇤⇤⇤
(0.060)
Stroop Task (BL) 0.546⇤⇤⇤
(0.078)
Health B. 0.745⇤⇤⇤
(0.056)
Prior Grade 0.458⇤⇤⇤ 0.516⇤⇤⇤ -0.058 0.054 0.007 -0.035 0.070 -0.067 0.058 0.143
(0.079) (0.105) (0.073) (0.070) (0.064) (0.049) (0.053) (0.062) (0.077) (0.198)
Prior ECTS Credits 0.005⇤⇤ 0.003 -0.002 -0.002 -0.002 0.003⇤⇤ 0.003⇤ -0.001 -0.001 -0.001
(0.002) (0.002) (0.002) (0.002) (0.002) (0.001) (0.001) (0.002) (0.002) (0.005)
Constant 0.143 0.044 -0.272 -0.159 0.224 0.569 0.391 -0.636⇤⇤ -0.719⇤⇤ 0.047 -0.704⇤⇤ -0.864
(0.442) (0.095) (0.666) (0.123) (0.373) (0.367) (0.351) (0.298) (0.330) (0.337) (0.326) (0.798)
Study Program FE Yes No Yes No Yes Yes Yes Yes Yes Yes Yes No
Study Year FE Yes No Yes No Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.302 0.003 0.328 0.053 0.373 0.453 0.508 0.705 0.661 0.538 0.405 0.483
N 177 177 121 121 182 182 182 182 182 182 182 182
Note: This table shows OLS regressions of the grades and secondary outcomes on the treatment dummy. Models (1) and (3) use the mean grade of the main summer- and
winter-semester exam periods, respectively, as dependent variables. Models (2) and (4) use the change in grade (summer grade - baseline grade and winter grade - baseline grade,
respectively) as dependent variables. Models (5) - (7) use the mental health scales, models (8) - (10) use the non-cognitive skills outcomes, model (11) uses the Stroop task and
model (12) uses the health behavior index as dependent variables. All models except (2) and (4) control for prior mean grade, prior total ECTS credits, study-program fixed
e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies). All models additionally contain a dummy equal to 1 for female students and an interaction term of this dummy
and the treatment dummy. All models include a constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

69
 Table C.11: Outcomes by Attendance
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Grade (s, VA) Grade(s, D) Grade (l, VA) Grade (l, D) Stress Anxiety Depression Self-Control Conscient. Neurot. Stroop Health B.
TreatXAttended often -0.353⇤ -0.328⇤ 0.464⇤⇤ 0.370⇤ -0.729⇤⇤⇤ -0.515⇤⇤⇤ -0.672⇤⇤⇤ 0.566⇤⇤⇤ 0.561⇤⇤⇤ -0.433⇤⇤⇤ 0.305⇤⇤ 0.569⇤⇤⇤
(0.183) (0.195) (0.185) (0.191) (0.146) (0.131) (0.138) (0.104) (0.103) (0.128) (0.139) (0.106)
TreatXAttended rarely -0.161 0.129 0.276 0.545⇤⇤ -0.561⇤⇤⇤ -0.319⇤ -0.491⇤⇤⇤ 0.440⇤⇤⇤ 0.461⇤⇤⇤ -0.230 -0.027 0.197
(0.164) (0.172) (0.221) (0.229) (0.162) (0.164) (0.140) (0.109) (0.141) (0.154) (0.151) (0.187)
Prior Grade 0.470⇤⇤⇤ 0.500⇤⇤⇤ -0.044 0.066 0.023 -0.040 0.069 -0.055 0.034 0.040
(0.083) (0.108) (0.074) (0.073) (0.066) (0.050) (0.055) (0.064) (0.078) (0.061)
Prior ECTS Credits 0.005⇤ 0.003 -0.002 -0.002 -0.002 0.003⇤⇤ 0.002⇤ -0.001 -0.001 -0.000
(0.002) (0.002) (0.002) (0.002) (0.002) (0.001) (0.001) (0.002) (0.002) (0.002)
Stress (BL) 0.480⇤⇤⇤
(0.070)
Anxiety (BL) 0.608⇤⇤⇤
(0.065)
Depression (BL) 0.583⇤⇤⇤
(0.065)
Self-Control (BL) 0.753⇤⇤⇤
(0.044)
Conscient. (BL) 0.706⇤⇤⇤
(0.049)
Neurot. (BL) 0.735⇤⇤⇤
(0.058)
Stroop Task (BL) 0.561⇤⇤⇤
(0.078)
Health B. (BL) 0.648⇤⇤⇤
(0.050)
⇤⇤⇤
Constant 0.071 0.044 -0.166 -0.159 0.254 0.443 0.162 -0.369 -0.697 0.214 -0.397 -0.493⇤
(0.438) (0.095) (0.666) (0.123) (0.365) (0.338) (0.338) (0.255) (0.265) (0.270) (0.246) (0.251)
Study Program FE Yes No Yes No Yes Yes Yes Yes Yes Yes Yes Yes
Study Year FE Yes No Yes No Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.298 0.028 0.321 0.056 0.374 0.457 0.508 0.703 0.656 0.542 0.416 0.545
N 177 177 121 121 182 182 182 182 182 182 182 182
Note: This table shows OLS regressions of the grades and secondary outcomes on course attendance. Models (1) and (3) use the mean grade of the main summer- and winter-
semester exam periods, respectively, as dependent variables. Models (2) and (4) use the change in grade (summer grade - baseline grade and winter grade - baseline grade,
respectively) as dependent variables. Models (5) - (7) use the mental health scales, models (8) - (10) use the non-cognitive skills outcomes, model (11) uses the Stroop task and
model (12) uses the health behavior index as dependent variables. All models except (2) and (4) control for prior mean grade, prior total ECTS credits, study-program fixed
e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies). All models additionally contain two interaction terms. ”Treat*Attendend often” identifies the treatment e↵ect –
relative to the control group – for students whose course attendance was above the treatment-group mean. ”Treat*Attendend rarely” identifies the treatment e↵ect – relative to
the control group – for students whose course attendance was below the treatment-group mean. All models include a constant. Heteroskedasticity robust standard errors are in
parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

Table C.12: Correlations of Grades (non-inv.), Non-Cognitive and Cognitive Skills

at Baseline

Prior Grade Stress (BL) Anxiety (BL) Depr. (BL) Self-C. (BL) Consc. (BL) Neurot. (BL) Stroop T. (BL)
Prior Grade 1
Stress (BL) 0.0952 1
Anxiety (BL) 0.0308 0.694⇤⇤⇤ 1
Depr. (BL) 0.247⇤⇤⇤ 0.583⇤⇤⇤ 0.608⇤⇤⇤ 1
Self-C. (BL) -0.268⇤⇤⇤ -0.234⇤⇤⇤ -0.119⇤ -0.347⇤⇤⇤ 1
Consc. (BL) -0.238⇤⇤⇤ -0.215⇤⇤⇤ -0.149⇤⇤ -0.336⇤⇤⇤ 0.703⇤⇤⇤ 1
Neurot. (BL) -0.0914 0.646⇤⇤⇤ 0.710⇤⇤⇤ 0.450⇤⇤⇤ -0.154⇤⇤ -0.196⇤⇤⇤ 1
Stroop T. (BL) -0.214⇤⇤⇤ -0.105 -0.102 -0.0556 -0.0988 -0.122⇤ -0.0239 1
⇤
Prior grade is the non-inverted, unstandardized average grade before the intervention including marks from failed exams, i.e. larger grades indicate worse grades.
p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

Table C.13: Long-Term E↵ects by Short-Term E↵ects

If short-term absolute change is positive If short-term absolute change is negative
(1) (2) (3) (4) (5) (6)
Grade (long, VA) Grade (long, VA) Grade (long, D) Grade (long, VA) Grade (long, VA) Grade (long, D)
Treat 0.542⇤⇤ 0.566⇤⇤⇤ 0.761⇤⇤⇤ 0.022 0.246 0.206
(0.199) (0.197) (0.214) (0.296) (0.278) (0.226)
Prior Grade 0.249 0.777⇤⇤⇤
(0.180) (0.152)
Prior ECTS Credits 0.004 0.004
(0.003) (0.005)
Constant -0.132 -1.487⇤ -0.124 -0.121 -2.366⇤⇤ -0.231
(0.144) (0.785) (0.167) (0.230) (1.027) (0.148)
Study Program FE No Yes No No Yes No
Study Year FE No Yes No No Yes No
R2 0.082 0.382 0.121 0.000 0.562 0.016
N 74 71 71 50 50 50
Note: This table shows OLS regressions of the mean grade of the main winter semester exam period on the treatment dummy. Models (1)
- (3) focus on students whose short-run average grade was higher than at baseline while models (4) - (6) focus on students whose short-run
average grade was lower than at baseline. Models (1) and (4) use the post-intervention outcome as dependent variable and do not include
control variables. Models (2) and (5) use the post-intervention outcome as dependent variable and control for prior mean grade, prior total
ECTS credits, study-program fixed e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies). Models (3) and (6) use the change in mean
grade (winter-semester grade baseline grade) as dependent variable and do not include control variables. All models include a constant.
Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

70
 Table C.14: Outcomes by Own Practice
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Grade (s, VA) Grade(s, D) Grade (l, VA) Grade (l, D) Stress Anxiety Depression Self-Control Conscient. Neurot. Stroop Health B.
Treat -0.401 -0.390 -0.269 -0.310 -0.307 -0.180 -0.314⇤ 0.238 0.353⇤⇤ -0.124 0.438⇤⇤ 0.383
(0.306) (0.327) (0.413) (0.385) (0.242) (0.197) (0.177) (0.177) (0.178) (0.208) (0.174) (0.238)
TreatXOwn practice 0.175 0.347 0.893⇤⇤ 0.969⇤⇤ -0.458 -0.415 -0.364 0.265 0.135 -0.324 -0.399 0.093
(0.355) (0.371) (0.445) (0.420) (0.296) (0.269) (0.240) (0.203) (0.204) (0.257) (0.263) (0.295)
Own practice -0.039 -0.062 -0.307 -0.252 0.060 0.221 0.035 0.158 0.149 0.101 0.172 -0.097
(0.166) (0.194) (0.239) (0.253) (0.186) (0.190) (0.160) (0.116) (0.127) (0.155) (0.203) (0.155)
Prior Grade 0.461⇤⇤⇤ 0.543⇤⇤⇤ -0.066 0.049 0.002 -0.018 0.086 -0.074 0.052 0.063
(0.079) (0.108) (0.072) (0.071) (0.064) (0.047) (0.053) (0.063) (0.076) (0.058)
Prior ECTS Credits 0.005⇤ 0.003 -0.002 -0.003 -0.002 0.002⇤ 0.002 -0.001 -0.001 -0.000
(0.002) (0.002) (0.002) (0.002) (0.002) (0.001) (0.001) (0.002) (0.002) (0.002)
Stress (BL) 0.485⇤⇤⇤
(0.070)
Anxiety (BL) 0.602⇤⇤⇤
(0.064)
Depression (BL) 0.585⇤⇤⇤
(0.065)
Self-Control (BL) 0.751⇤⇤⇤
(0.043)
Conscient. (BL) 0.699⇤⇤⇤
(0.048)
Neurot. (BL) 0.732⇤⇤⇤
(0.058)
Stroop Task (BL) 0.546⇤⇤⇤
(0.077)
Health B. (BL) 0.655⇤⇤⇤
(0.051)
Constant 0.006 0.081 -0.122 -0.025 0.247 0.518 0.546 -0.466⇤ -0.775⇤⇤⇤ 0.165 -0.385 -0.386
(0.691) (0.150) (0.641) (0.225) (0.362) (0.381) (0.341) (0.249) (0.256) (0.262) (0.237) (0.281)
Study Program FE Yes No Yes No Yes Yes Yes Yes Yes Yes Yes Yes
Study Year FE Yes No Yes No Yes Yes Yes Yes Yes Yes Yes Yes
R2 0.296 0.008 0.346 0.090 0.381 0.461 0.511 0.716 0.662 0.542 0.411 0.531
N 177 177 121 121 182 182 182 182 182 182 182 182
Note: This table shows OLS regressions of the grades and secondary outcomes on the treatment dummy. Models (1) and (3) use the mean grade of the main summer- and
winter-semester exam periods, respectively, as dependent variables. Models (2) and (4) use the change in grade (summer grade - baseline grade and winter grade - baseline grade,
respectively) as dependent variables. Models (5) - (7) use the mental health scales, models (8) - (10) use the non-cognitive skills outcomes, model (11) uses the Stroop task and
model (12) uses the health behavior index as dependent variables. All models except (2) and (4) control for prior mean grade, prior total ECTS credits, study-program fixed
e↵ects (9 dummies), and study-year fixed e↵ects (3 dummies). All models additionally contain a dummy “Own practice” equal to 1 for those students who indicated in the endline
questionnaire that they practiced meditation on their own during the period of the meditation course and an interaction term of this dummy and the treatment dummy. All
models include a constant. Heteroskedasticity robust standard errors are in parentheses. ⇤ p < .1, ⇤⇤ p < .05, ⇤⇤⇤ p < .01

71
D Invitation, Registration, Questionnaires, and
Session Content

Figure D.1: Invitation Text (translated from German)

Subject: Course “Fundamentals of Mindfulness Meditation” now open for bachelor’s and master’s students

Dear Students,

we are pleased to offer an 8-week course during the summer semester in which you can learn the basics of mindfulness
meditation.

Please note: The course is now open to WiSo faculty master’s students as well! The application deadline has been
extended to April 24!

The course is based on the Mindfulness Based Stress Reduction (MBSR) program. This program has been successfully
used worldwide in corporate, university, and healthcare settings, among others, to effectively reduce stress, promote
mental and physical health, and enhance performance.

Participation in the course is free of charge, as the entire course is financed by Techniker Krankenkasse (regardless of
which health insurance company someone is insured with).

The course will be given by two experienced meditation teachers and will take place once a week starting May 15, 2019.
There will be 6 course groups that will meet for one hour each Wednesday or Friday at the following times in a seminar
room in the SSC building:

- 2:00 pm-3:00 pm

- 3:15 pm-4:15 pm

- 4:30 pm-5:30 pm

The course consists of the weekly meeting, where meditation and relaxation techniques are learned, and daily home
exercises, where the techniques are practiced and deepened independently.

Since the course is scientifically evaluated and very expensive, regular attendance and high motivation are essential for
participation.

The number of places in the course is limited and only students of the WiSo faculty can apply for it. Places are allocated
by lottery. More details on the application page.

The course will be scientifically evaluated by a team of researchers led by Jun.-Prof. Dr. Lea Cassar. For this reason, we
rely on all students to participate in an online survey when applying for a course place, as well as a survey at the end of
the semester. Please see the registration page for more information.

If you are interested in taking a course, we would appreciate it if you apply to participate by April 24 at the following
website:

https://unikoelnwiso.eu.qualtrics.com/jfe/form/SV_cGB9wExtZWMbl1b

Some browser plugins prevent the page from displaying properly. If this happens, please disable them temporarily.

Best regards,

Jun.-Prof. Dr. Lea Cassar

72
 Figure D.2: Registration Page (translated from German)

Course description “Basics of Mindfulness Meditation”

Mindfulness is a state of mind that involves being aware of the present moment as best we can. That means looking at things as they
are right now, without judging them and without actively intervening.

This sounds relatively unspectacular, but in reality it is not that easy. Because most of the time our mind is very busy - we think about
the past, we make plans, we worry and think about everything that still has to be done.

With mindfulness we get back in touch with ourselves and draw strength from the present moment. We can look at problems with
more distance and gain clarity for new decision-making possibilities and actions. We develop a better feeling for physical and
psychological signals, for our stress reactions and stress limits.

Mindfulness is a key competence for healthy living and one's own personal development, especially when it comes to coping with the
diverse demands of everyday life at university and at work, while remaining productive and at the same time satisfied and healthy in
the long term.

In this 8-week course, you will learn meditation and relaxation techniques that allow you to deal with stress and stressful situations in
a more mindful way.

What will participants take away from the course?

 The basics of mindfulness meditation

 Learning formal mindfulness exercises and how to apply them in everyday life
 Improving body awareness and learning to consciously relax
 Improving understanding and regulation of emotions
 Better insight into stress-reinforcing thought patterns and the ability to gradually dissolve them

What are the positive effects of practicing mindfulness?

 Better concentration
 More clarity and objectivity even in difficult situations
 More calmness and composure in dealing with stress
 Increased effectiveness with less effort at the same time
 Higher well-being

The course is based on the concept of Mindfulness Based Stress Reduction (MBSR) according to Jon Kabat-Zinn. This program was
developed in the 1970s in the USA and is now successfully used worldwide in organizations, educational institutions, in health care
and in psychotherapy for many people suffering from stress. Meanwhile, numerous international studies prove the positive effects of
MBSR on mental and physical health as well as effectiveness.

The course is scientifically evaluated by a research team led by Jun.-Prof. Dr. Lea Cassar. For this purpose, all course applicants fill out
an online questionnaire once directly at the time of application as well as at the end of the summer semester. They should plan about
20 minutes for each of these.

Course Schedule

The course will take place from May 15 to July 12, 2019 in the SSC building and will consist of one hour per week. In addition, there
will be daily exercises to do at home.

To help us plan, please click on all times that are compatible with your schedule and when you could attend the course:

Wednesdays 2:00 pm-3:00 pm Fridays 2:00 pm-3:00 pm

3:15 pm-4:15 pm 3:15 pm-4:15 pm
4:30 pm-5:30 pm 4:30 pm-5:30 pm

73
 Figure D.2: Registration Page (translated from German, continued)

Personal Information

First Name:
Last Name:
Email Address:
Gender:
Year of birth:
Matriculation number:
Programm of study:
Semester of study:
In which courses do you plan to take an exam this summer semester?:

Conditions of Participation

1. Participation in the course is free of charge, but there is only a limited number of places available. These will be allocated
among the applicants by lottery and taking into account the lectures taken and time availability in the summer semester
2019.
2. How course applicants answer the questionnaire has no influence on the allocation of places. However, only applicants who
have completely filled out the questionnaire and the application can participate in the allocation of places.
3. All course applicants, regardless of whether or not they have been awarded a place, agree to take part in an online survey
lasting approximately 20 minutes at the end of the summer semester. They will receive appropriate financial compensation
for doing so.
4. Participation in the course is voluntary and at the participant's own responsibility. Withdrawal is possible at any time without
giving reasons by sending an e-mail to lcassar@uni-koeln.de.

Privacy Policy

1. All course applicants, regardless of whether you have received a place or not, agree that their examination results from their
current course of study from previous semesters as well as from the academic year 2019/2020 will be stored by Jun.-Prof. Dr.
Lea Cassar in the IT network of the University of Cologne for research purposes until the end of the study. After that, the data
will be deleted.
2. All course applicants, regardless of whether they have received a place or not, agree that Jun.-Prof. Dr. Lea Cassar will link the
data from the registration and the two surveys with the applicants’ examination data in the IT network of the University of
Cologne and then process it in anonymized form with the other two researchers involved in the study (Dr. Mira Fischer, WZB
Berlin, and Dr. Vanessa Valero, University of Zurich) for the purpose of scientific research.
3. Consent for data processing can be revoked at any time by emailing lcassar@uni-koeln.de.

Last but not least: For the scientific evaluation it is important that participants attend the course until the end. All course applicants
declare to intend to attend the complete course if they get a place.

YES. I have read and agree to the course description, conditions of participation, and privacy policy. I hereby apply for a place
on the course. (Continue to questionnaire.)

NO. I do not agree and do not wish to apply for a place.

If you have any questions, please contact Jun.-Prof. Dr. Lea Cassar (lcassar@uni-koeln.de).

Submit

74
 Table D.3: Questionnaire

Variable Name Text Categories Pre Post

First name First name [open field] x x
Last name Last name [open field] x x
Matriculation no. Matriculation number [open field] x x
Gender Gender 1- Male; 2- Female; x
3- Diverse
Year of birth In which year were you born? [open field] x
Bachelor Are you currently enrolled as a Bachelor’s stu- 1- Yes; 2- No x
dent at the WiSo faculty?
Bachelor program In which Bachelor’s program are you enrolled? 1- Business Admin- x
istration; 2- Eco-
nomics; 3- Eco-
nomics Social Sci-
ence; 4- Social Sci-
ences; 5- Health
Economics; 6- In-
formation Systems;
7- other
Master Are you currently enrolled as a Master’s stu- 1- Yes; 2- No x
dent at the WiSo faculty?
Master program In which Master’s program are you enrolled? 1- M.Sc. Interna- x
tional Management
/ CEMS MIM; 2-
M.Sc. Business
Administration; 3-
M.Sc. Economics;
4- M.Sc. Health
economics; 5-
M.Sc. Information
Systems; 6- M.A.
Political Science;
7- M.Sc. Soci-
ology and Social
Research; 8- M.Ed.
Business education

Semester In which semester are you? (in your current 1st; 2nd; 3rd; x
degree program) 4th; 5th; 6th; 7th;
8th (being 8th or
higher)
Mindfulness Below are a number of statements about your 1- Almost never; 2- x x
everyday experiences. Please indicate on the Very rarely, 3 -
scale below how often or rarely each experi- Rarely 4- Often; 5-
ence has happened to you in the last 2 weeks. Very often, 6 - Al-
Your answers should reflect your true experi- most always
ences, and not depend on your expectations of
yourself. 1- I could experience an emotion and
only realize it later. 2- I find it hard to focus on
what is going on. 3- I tend not to notice feel-
ings of physical tension or discomfort until they
really grab my attention. 4- It seems like I’m
functioning ”automatically” without really be-
ing aware of what I am doing. 5- I rush through
activities without paying attention to them. 6-
I catch myself listening to others with one ear
while doing something else at the same time.
7- I find myself absorbed in thoughts of the fu-
ture or the past. 8- I nibble, not realizing that
I am eating. 75
 Table D.3: Questionnaire (continued)

Variable Name Text Categories Pre Post

Stress The following questions are about how often 1- Never; 2- Almost x x
you feel stressed during the last 2 weeks. 1- In never; 3- Some-
the last two weeks, how often have you been times; 4- Fairly of-
upset because of something that happened un- ten; 5- Very often
expectedly? 2- In the last two weeks, how often
have you felt that you were unable to control
the important things in your life? 3- In the
last two weeks, how often have you felt ner-
vous and stressed? 4- In the last two weeks,
how often have you felt confident about your
ability to handle your personal problems? 5-
In the last two weeks, how often have you felt
that things were going your way? 6- In the
last two weeks, how often have you found that
you could not cope with all the things that you
had to do? 7- In the last two weeks, how of-
ten have you been able to control irritations in
your life? 8- In the last two weeks, how often
have you felt that you were on top of things?
9- In the last two weeks, how often have you
been angered because of things that happened
that were outside of your control? 10- In the
last two weeks, how often have you felt difficul-
ties were piling up so high that you could not
overcome them?
Anxiety Over the last 2 weeks, how often have you been 1- Not at all; 2- Sev- x x
bothered by any of the following problems? 1- eral days; 3- More
Feeling nervous, anxious or on edge. 2- Not be- than half the days;
ing able to stop or control worrying. 3- Worry- 4- Nearly everyday
ing too much about di↵erent things. 4- Trouble
relaxing. 5- Being so restless that it is hard to
sit still. 6- Becoming easily annoyed or irri-
table. 7- Feeling afraid as if something awful
might happen.
Depression Over the last 2 weeks, how often have you been 0- Not at all; 1- Sev- x x
bothered by any of the following problems? 1- eral days; 2- More
Little interest or pleasure in doing things. 2- than half the days;
Feeling down, depressed, or hopeless. 3- Trou- 3- Nearly everyday
ble falling or staying asleep, or sleeping too
much. 4- Feeling tired or having little energy.
5- Poor appetite or overeating. 6- Feeling bad
about yourself - or that you are a failure or have
let yourself or your family down. 7- Trouble
concentrating on things, such as reading the
newspaper or watching television. 8- Moving
or speaking so slowly that other people could
have noticed. Or the opposite - being so fidgety
or restless that you have been moving around
a lot more than usual. 9- Thoughts that you
would be better o↵ dead or of hurting yourself
in some way.

76
 Table D.3: Questionnaire (continued)

Variable Name Text Categories Pre Post

Self-control Please indicate how much each of the following 1- Not at all; 2- x x
statements reflects how you typically are. 1- Slightly; 3- Moder-
I am good at resisting temptation. 2- I have ately; 4-?; 5- Very
a hard time breaking bad habit. 3- I am lazy. much
4- I say inappropriate things. 5- I do certain
things that are bad for me, if they are fun.
6- I refuse things that are bad for me. 7- I
wish I had more self-discipline. 8- People would
say that I have iron self-discipline. 9- Pleasure
and fun sometimes keep me from getting work
done. 10- I have trouble concentrating. 11- I
am able to work e↵ectively toward long-term
goals. 12- Sometimes I can’t stop myself from
doing something, even if I know it is wrong.
13- I often act without thinking through all
the alternatives.
Conscientiousness How true the following are about you. 1- Does 1-Disagree, 2- x x
a thorough job. 2- Can be somewhat careless. Rather Disagree; 3-
3- Is a reliable. 4- Tends to be disorganized. Neutral; 4- Rather
5- Tends to be lazy. 6- Perseveres until the Agree 5- Agree
task is finished. 7- Does things efficiently. 8-
Make plans and follows through with them. 9-
Is easily distracted.
Neuroticism How true the following are about you. 1- Is de- 1-Disagree, 2- x x
pressed, blue 2- Is relaxed, handles stress well. Rather Disagree; 3-
3- Can be tense. 4- Worries a lot. 5- Is emo- Neutral; 4- Rather
tionally stable, not easily upset. 6- Remains Agree 5- Agree
calm in tense situations. 7- Gets nervous eas-
ily.
Study behavior How exactly do these statements currently ap- 1-Disagree, 2- x x
ply to you? 1- I usually rightfully estimate how Rather Disagree; 3-
much time I need to complete a task. 2- Ev- Neutral; 4- Rather
ery day, I know what things I have to do and Agree 5- Agree
how far I can handle them. 3- If I cannot keep
up with my work, I often give up. 4- I always
make the same mistakes. 5- I distribute my
work and my learning evenly throughout the
semester. 6- I often dig with thoughts while
learning. 7- I consciously gather my concen-
tration before I start learning. 8- I regularly
check my messages on my smartphone while
I’m learning. 9- I set up my learning place in a
quiet place without distractions. 10- I usually
start learning only when the pressure is very
high. 11- It is easy for me to concentrate on
learning for a long time. 12- I am sure that I
can learn all the skills to be successful in my
studies. 13- I like to study. 14- I am always
attentive in lectures. 15- I am not a good stu-
dent. 16- I am nervous before exams. 17- I
find it easy to manage time well while writing
an exam. 18- I panic easily when I can’t solve
an exam problem. (Statements 16-18 only con-
tained in post-intervention questionnaire.)

77
 Table D.3: Questionnaire (continued)

Variable Name Text Categories Pre Post

Health behavior Please rate a few more statements about your 1-Never; 2- Hardly x x
current habits. 1- I get up at the same time ev- ever; 3- Occasion-
ery morning., 2- I consciously relax., 3- I drink ally; 4- Rather Reg-
alcohol. 4- I drink co↵ee or tea to stay awake. ularly; 5- Very Reg-
5- I take medication to be more efficient. 6- I ularly
go to bed late in the evening and then get tired
the next day. 7- I smoke.
Stroop task (20 items of the type: ”Click on the answer (Correct answer x x
that matches the color of the following word: involved clicking
Blue” on separate screens. Students were in- on the color in
structed that among the people with the great- which the word
est number of correct answers the three fastest (e.g. ”Blue”) was
would be paid 20 euros.) written, which
could be any of
black, blue, yellow,
green, or red. )
Experience Do you have experience with meditation? 1- No, no expe- x
rience.; 2- Yes, I
have meditated,
but not regularly.
I do not meditate
at the moment.; 3-
Yes, I meditated
regularly, but I
do not meditate
at the moment.;
4- Yes, I meditate
sometimes.; 5- Yes,
I meditate (almost)
every week.; 6- Yes,
I meditate (almost)
every day.
Motivation What motivation is most important to your de- 1- I am curious.; 2- x
sire to learn mindfulness meditation? Please I want to improve
choose an option. my concentration.;
3- I want to learn
to better relax.; 4-
I want to learn to
deal better with my
emotions.; 5- I want
to loose weight.; 6-
I want to be more
productive
Father’s education Does your father have a university degree? 1- Yes; 2- No x
Mother’s education Does your mother have a university degree? 1- Yes; 2- No x
Future interest Would you like to be informed by e-mail if an- 1- Yes; 2- No x
other mindfulness course is o↵ered at the Uni-
versity of Cologne?
Liked course How did you like the course ”Fundamentals of 1- Very much; 2- x (T)
Mindfulness Meditation”? Much; 3- Rather
less; 4- Not at all;
5- I don’t know
Learned course How much did you learn on the course? 1- Very much; 2- x (T)
Much; 3- Rather
less; 4- Nothing at
all; 5- I don’t know

78
 Table D.3: Questionnaire (continued)

Variable Name Text Categories Pre Post

Recommend course Would you recommend participation in the 1- Definitely; 2- x (T)
course to other students? Probably; 3- Prob-
ably not; 4- Def-
initely not; 5- I
don’t know
Exercises first half During the FIRST HALF of the course, how 1- (Almost) daily, x (T)
often did you do the exercises on your own? about three times
per day; 2- (Al-
most) daily, about
once a day; 3- On
at least half of the
days, about three
times per day; 4-
On at least half
of the days, about
once a day; 5- Irreg-
ularly, about three
times per day; 6-
Irregularly, about
once a day; 7-
Rarely; 8- Never
Exercises second During the SECOND HALF of the course, how 1- (Almost) daily, x (T)
half often did you do the exercises on your own? about three times
per day; 2- (Al-
most) daily, about
once a day; 3- On
at least half of the
days, about three
times per day; 4-
On at least half
of the days, about
once a day; 5- Irreg-
ularly, about three
times per day; 6-
Irregularly, about
once a day; 7-
Rarely; 8- Never
Excercises now Are you currently continuing to do the exer- 1- Yes; 2- No x (T)
cises?
Own practice In the last two months, how often have you 1- (Almost) every x
meditated on your own (i.e., beyond or inde- day; 2- (Almost)
pendent of the course exercises)? [treatment every week; 3-
group]; In the last two months, how often have Sometimes; 4-
you meditated on your own? [control group] Never
App Are you currently using a meditation app? 1- Yes; 2- No x
Friends Are you friends with students who have taken 1- Yes; 2- No x (C)
the ”Fundamentals of Mindfulness Medita-
tion” course?
Course materials Have any participants in the ”Fundamentals of 1- Yes; 2- No x (C)
Mindfulness Meditation” course shared course
materials with you?
Comments Anything else you would like to tell us? - Oth- [open field] x
erwise, simply leave the field blank.
Note: This table lists the items contained in the pre- and/or post-intervention questionnaire. The column “Vari-
able name” contains the name used in the analysis. The column “Text” contains the item text shown to partic-
ipants. The column ”Categories” contains the answer categories available to participants. Columns “Pre” and
“Post” indicate whether the variable was contained in the pre- and/or post-intervention questionnaire, respec-
tively. T = only contained in treatment group questionnaire; C = only contained in control group questionnaire.

79
 Table D.4: Sessions of the Meditation Course

Group session Individual exercises

Week 1 - time out - time out (3 times per day)
- introduction to the course, motivation to - body scan (once a day)
participate in the course, introduction of
participants
- topics of teacher’s talk: mindfulness, focus
on the body, present moment awareness
- body scan
- exchange of thoughts
Week 2 - time out - time out (3 times per day)
- exchange of experiences in the last week - body scan (once a day)
- topics of teacher’s talk: somatic markers, - observing the body in everyday
the body as a resource, listening to the body, life – stopping to pay attention
dealing with unpleasant emotions and pain to sensations
- body scan
- exchange of thoughts
Week 3 - time out - time out (3 times per day)
- exchange of experiences in the last week - body scan or sitting meditation
- topics of teacher’s talk: attitudes (once a day, alternating)
while practicing meditation (beginner’s - observing judgments in stress-
mind, non-intentionality, not judging, let- ful situations
ting go, not grasping, trust, benevo-
lence/compassion), sitting posture
- sitting meditation, observing the breath
- exchange of thoughts
Week 4 - time out - time out (3 times per day)
- exchange of experiences in the last week - body scan or sitting meditation
- topics of teacher’s talk: stress, triggers and (once a day, alternating)
responses, judgement, expectations towards - observing the arising of stress
ourselves, autopilot, creating a gap between and stress-related thoughts in
triggers and responses, introduction to yoga everyday life; stress diary
- yoga (standing)
- exchange of thoughts
Week 5 - time out - time out (3 times per day)
- exchange of experiences in the last week - sitting meditation, yoga or
- topics of teacher’s talk: judgements are body scan (once a day, alternat-
thoughts, dealing with thoughts (not iden- ing)
tifying with and observing thoughts) - observing thoughts that trigger
- sitting meditation difficult emotions in every day
- exchange of thoughts life

80
 Table D.4: Sessions of the Meditation Course (continued)

Group session Individual exercises

Week 6 - time out - time out (3 times per day)
- exchange of experiences in the last week - sitting meditation, yoga or
- topics of teacher’s talk: dealing with emo- body scan (once a day, alternat-
tions (observing and not ignoring emotions, ing)
not identifying with emotions) - observing (difficult) emotions
- guided self-reflection using RAIN method in everyday life with kindness
(recognize, allow, investigate, nurture) by
Tara Brach
- sitting meditation, observing with compas-
sion and kindness
- exchange of thoughts
Week 7 - time out - time out (3 times per day)
- exchange of experiences in the last week - sitting meditation, yoga or
- topics of teacher’s talk: mindful communi- body scan (once a day, alternat-
cation ing)
- practice of mindful communication in pairs - integrating new skills into ev-
- exchange of thoughts eryday life
Week 8 - time out - time out (3 times per day)
- exchange of experiences in the last week - sitting meditation, yoga or
- topics of teacher’s talk: tips on integrat- body scan (once a day, alternat-
ing new skills into everyday life, cultivating ing)
gratitude and self-esteem - integrating new skills into ev-
- sitting meditation eryday life
- exchange of experiences and thoughts with
respect to the course
Note: This table lists the contents of the sessions of the meditation course that took place weekly and lasted
60 minutes, and the individual exercises that participants were asked to do daily and that were accompanied by
written handouts and audio recordings to follow along.

81