You are on page 1of 8

Learning and Instruction 58 (2018) 193–200

Contents lists available at ScienceDirect

Learning and Instruction


journal homepage: www.elsevier.com/locate/learninstruc

Does it matter if students procrastinate more in some courses than in others? T


A multilevel perspective on procrastination and academic achievement
Kristina Kljajic∗, Patrick Gaudreau
University of Ottawa, Canada

A R T I C LE I N FO A B S T R A C T

Keywords: We proposed and tested a novel multilevel perspective on procrastination by examining the prospective relation
Procrastination between procrastination and grades across students (i.e., between-person level) and across the courses taken by
Academic achievement each student (i.e., within-person level). A sample of 208 university students completed repeated measures of
Multilevel modeling procrastination for each of their courses during the semester and the official final grades were obtained at the
Time management
end of the semester. The results of multilevel modeling revealed that students who procrastinated more than
Tertiary education
other students received lower grades than these students (i.e., between-person level). Moreover, the results
revealed that in the courses in which students procrastinated more than their own average, they received lower
grades than their own average (i.e., within-person level). These findings build upon past findings on procras-
tination across students while also moving this research field forward by offering a new understanding of
procrastination as a within-person risk factor for academic difficulties.

1. Introduction frequent stressors reported by university students (Hurst, Baranik, &


Daniel, 2013). Despite the fact that time is a limited and useful resource
Philip and Sophia are university students who believe that they have for academic success (Credé & Kuncel, 2008), many students procras-
obtained low midterm grades because they needlessly delayed studying tinate by unreasonably delaying their schoolwork despite knowing that
for their exams. An academic counsellor offers them to complete a their delay will lead them to experience undesirable outcomes
questionnaire that evaluates the frequency of their academic procras- (Klingsieck, 2013a; Steel, 2007). Meta-analytical findings have shown
tination (see Fig. 1, panel A). Their average scores reveal that Philip that procrastination relates to worse academic achievement including
frequently procrastinates, whereas Sophia rarely procrastinates (see lower assignment grades, course grades, and grade point averages (see
Fig. 1, panel B). At first glance, the academic counsellor is more worried Kim & Seo, 2015; Richardson, Abraham, & Bond, 2012; Steel, 2007; van
about Philip than Sophia. However, Sophia explains that although she Eerde, 2003). We argue that procrastination should negatively relate to
does not frequently procrastinate in general in her courses, she worries academic achievement even among many students – like Sophia – who
that her procrastination in Chemistry affects her performance and her are not typically considered as procrastinators.
likelihood of entering graduate school. The counsellor is intrigued and
asks Philip and Sophia to complete the same questionnaire of procras- 1.1. A novel multilevel perspective on procrastination
tination for each of their courses. Although Philip's scores are all the
same, Sophia's scores vary from one course to another. Despite her re- In this study, we proposed a multilevel perspective in which pro-
latively low scores of procrastination in Psychology, Philosophy, and crastination is expected to fluctuate across students (i.e., between-person
Biology, she procrastinates in Chemistry as frequently as Philip gen- level) and across the courses taken by each student during the semester
erally procrastinates in his schoolwork. The counsellor wonders whe- (i.e., within-person level). Both assertions are complementary and should
ther he should also be worried about Sophia's procrastination in be empirically investigated in order to refine our understanding of
Chemistry despite her low average score of procrastination. procrastination and its influence on academic achievement.
Poor time management is often portrayed by the media as the cul- On the one hand, procrastination is often defined as a personality
prit of many academic difficulties experienced by university students. trait (e.g., Schouwenburg, 2004; Sirois, 2014; Steel, 2007; Steel &
Time is a limited resource and the lack of time is one of the most Ferrari, 2013), partly because the construct displays substantial


Corresponding author. School of Psychology, University of Ottawa, 136 Jean-Jacques Lussier, room 5067, Ottawa, Ontario, Canada, K1N 6N5.
E-mail address: kklja060@uottawa.ca (K. Kljajic).

https://doi.org/10.1016/j.learninstruc.2018.06.005
Received 24 February 2018; Received in revised form 26 June 2018; Accepted 29 June 2018
0959-4752/ © 2018 Elsevier Ltd. All rights reserved.
K. Kljajic, P. Gaudreau Learning and Instruction 58 (2018) 193–200

academic achievement is useful to assume that Philip is more likely to


underperform than Sophia. However, research at the within-person
level is required to inform academic counsellors whether Sophia is more
likely to have significantly lower performance in Chemistry compared
to her other courses. Hence, our multilevel perspective offers a needed
roadmap to reshape our research agenda and to better understand the
influence of procrastination on the achievement outcomes of students
like Sophia and Philip.

1.2. The present study

We conducted a study in which students reported their level of


procrastination in each of their courses during a semester. Our main
goal was to investigate the prospective relation between procrastination
during the semester and grades at the end of the semester at the be-
tween-person and within-person levels. Our complementary goal was to
propose a stringent test of the relation between procrastination and
grades by controlling for variables that could act as confounders at both
levels of analysis.
At the between-person level, some students may have experienced
difficulties in obtaining good grades before entering university (i.e.,
admission GPA) which would decrease their likelihood of obtaining
Fig. 1. An illustration of a questionnaire measuring procrastination (panel A) good grades in university. Moreover, students who have to manage the
that was hypothetically filled out by two fictional university students (panel B). requirements of multiple courses during the semester (i.e., number of
In panel B, the number at the top of each oval represents the average score of
courses) and students who have just started university and did not get
procrastination of the student. The number at the bottom of each rectangle
the chance to adjust their learning strategies in a new environment (i.e.,
represents the score of procrastination for the course. Darker shades of grey
reflect higher levels of procrastination. The filled arc with two arrows re-
year of study) may be less likely to obtain high grades at the end of the
presents the between-person level of analysis comparing students to one an- semester. Therefore, admission GPA, the number of courses taken
other. The dotted arcs with multiple arrows represent the within-person level of during the semester, and the year of study could act as confounders in
analysis comparing courses to one another for each student. PSY = Psychology. the relation between procrastination and grades at the between-person
PHI = Philosophy. BIO = Biology. CHM = Chemistry. MAT = Mathematics. level.
ECO = Economy. At the within-person level, students may attach less importance to
their performance in some courses (i.e., course importance) or they may
temporal stability through test-retest reliability (see Steel, 2007) and be overwhelmed by the amount of work they have to accomplish in
across three time points during a semester (Rice, Richardson, & Clark, some courses (i.e., course workload) which could potentially decrease
2012). Genetic factors also explain a substantial proportion of the their engagement and success in these courses. Furthermore, students
variation in procrastination (Gustavson et al., 2017; Gustavson, Miyake, may be less engaged and experience less success in some courses be-
Hewitt, & Friedman, 2014, 2015). Studies at the between-person level cause the professor is generally perceived as being unclear and dis-
have generally found that students who procrastinate more than their organised (i.e., teaching evaluation) or because of the feeling of dis-
counterparts are at greater risk of experiencing psychological and tance between the professor and the students in large classrooms (i.e.,
academic maladjustment, such as higher anxiety and depression (Steel, number of students in the course). Hence, personal (i.e., course im-
2007; van Eerde, 2003) and lower academic achievement (e.g., Kim & portance), contextual (i.e., course workload) and environmental vari-
Seo, 2015; Richardson et al., 2012). Individual differences in procras- ables (i.e., teaching evaluation and the number of students registered in
tination are thus undeniable and consequential. the course) could act as confounders in the relation between procras-
On the other hand, procrastination can also be defined as a state tination and grades at the within-person level.
(e.g., Steel & Ferrari, 2013) and measured across time during a semester In this study, our between-person hypotheses were that (1a) students
(e.g., Dewitte & Schouwenburg, 2002; Goda et al., 2015; Steel, Svartdal, who procrastinate more than others should obtain lower grades and
Thundiyil, & Brothen, 2018), across academic tasks (Clark & Hill, 1994; (1b) this finding should still be significant over and above admission
Özer, Demir, & Ferrari, 2009; Solomon & Rothblum, 1984), and across GPA, the number of courses taken during the semester, and the year of
life domains (Klingsieck, 2013b). The majority of students – even those study. Our within-person hypotheses were that (2a) in the courses in
who are not considered as procrastinators based on their typical be- which students procrastinate more than their own average, they should
havior – procrastinate from time to time depending on the circum- obtain lower grades than their own average and (2b) this finding should
stances. Therefore, researchers should not only follow the mainstream still be significant over and above personal, contextual, and environ-
research interested in who procrastinates more compared to others (i.e., mental variables likely to influence academic achievement.
between-person level), but also investigate under which circumstances
individuals procrastinate more compared to their own average (i.e., 2. Method
within-person level).
Theoretical and empirical advances in personality (e.g., Fleeson, 2.1. Design and sample
2004; Fleeson & Noftle, 2012) and human achievement (e.g., Dalal,
Bhave, & Fiset, 2014) have outlined many advantages of differentiating, We used a two-level design in which we recruited participants in
comparing, and integrating psychological constructs at distinct levels of order to evaluate the procrastination and achievement in each of their
analysis. Research has even shown that robust and replicable effects courses during a semester. Courses (i.e., level 1) are nested within each
observed at the between-person level can fail to generalize at the student (i.e., level 2), as frequently seen in intensive longitudinal stu-
within-person level (e.g., Sitzmann & Yeo, 2013). In the fictional cases dies (e.g., daily diaries). The between-person sample corresponded to
of Philip and Sophia, the extant literature on procrastination and participants who took part in the study, whereas the within-person
sample corresponded to courses taken by the participants during the

194
K. Kljajic, P. Gaudreau Learning and Instruction 58 (2018) 193–200

semester. scale ranging from 1 (not at all agree) to 5 (totally agree). The internal
The between-person sample consisted of 208 undergraduate uni- consistency of the scale in this study was good (within-person α = .83;
versity students (64.3% female) aged between 17 and 37 years between-person α = .95). Multilevel confirmatory factor analysis
(M = 19.53, SD = 2.46) enrolled at a four-year public university lo- (MCFA) was conducted to evaluate whether the fit of the model to the
cated in an urban area in Canada. The within-person sample consisted data was comparable at the two levels of analysis. The analyses were
of 727 courses (before dealing with missing data). Participants reported performed with the maximum likelihood estimator with robust stan-
that they were taking one (0.5%), two (1.5%) three (2.4%), four dard errors (MLR) and the factor loadings were constrained to equality
(22.4%), five (69.3%), or six (3.9%) courses during the semester. They at both levels of analysis. The results of MCFA revealed that the fit of
were in the first (70%), second (17.4%), third (7.2%), and fourth the model was acceptable: χ2 (3) = 17.38 (scaling correction
(5.3%) year of their undergraduate degree. They were enrolled at the factor = 1.55), p < .01, CFI = .955, TLI = .910, RMSEA = .077,
faculties of social sciences (34.3%), health sciences (28.9%), science within-person SRMR = .010, between-person SRMR = .032. The stan-
(23%), arts (9.8%), management (2.5%), and engineering (1.5%). dardized loadings of the items ranged between .75 and .84 at the
Participants described their ethnic background as Caucasian (64.7%), within-person level and between .88 and .99 at the between-person
Arabic (7.7%), Asian (7.2%), African-American (6.8%), Aboriginal level. The fit of this model suggests that this three-item scale has a
(4.8%), Hispanic (2.4%), and other (6.3%). They lived with their par- comparably acceptable factor structure at the within-person level and
ents (42.8%), in a residence (30.3%), or in an apartment (26.9%) and between-person level.
they reported that they received a full scholarship (10.1%), a partial
scholarship (49.5%), or no scholarship (40.4%) to cover their tuition 2.2.4. Importance (level-1 variable)
fees. A total of 97 participants (46.9%) noted that they had to work Two items were used to measure the perceived importance of each
during the semester and they were working on average 16.6 h per week. course for the students. Participants assessed the importance of per-
Most students received one point toward their introductory psychology forming well in each course on a scale ranging from 1 (not at all) to 5
course for their participation in this study. Other students received a (totally). In this study, the internal consistency of the scale was good
compensation of five Canadian dollars. All participants completed a (within-person α = .86; between-person α = .97).
consent form and the study was approved by the Research Ethics Board
of the university. 2.2.5. Workload (level-1 variable)
One item taken from the student evaluation of teaching (i.e., S-
2.2. Procedure and measures Report; see below) was used to measure perceived workload in each
course on a scale ranging from 1 (very light) to 5 (very heavy).
University students were recruited to participate in a study to de-
scribe their experience in each of their courses during the semester. 2.2.6. Student evaluation of teaching (level-1 variable)
They completed an online consent form and a separate form to au- The S-Report is a 13-item questionnaire developed by the admin-
thorize the second author to have access to their official grades of the istration of the university to enable students to evaluate their professors
winter semester through the Registrar's Office. Participants then com- and their courses. The evaluation results are published online and ac-
pleted an online survey within the last month before the final exams to cessible to students and faculty members affiliated to the university.
ensure that they could reflect on their overall experience during the The Research Ethics Board of the university authorized us to use this
semester. The survey included socio-demographic questions, general information for our research purposes. The S-Report was administered
questions, and repeated course-related questions for each course taken uniformly and mandatorily across all courses on campus during the
during the semester. For each of their courses, participants provided 10th and 11th week of the semester. Data are available for all courses
their course code (e.g., PSY1101 L) and the name of their professor which were evaluated by at least six students. In each course, at least
before they could answer a series of questions referring to that course. two students were asked to volunteer to read the instructions and dis-
Participants were then asked if they were registered in a second course. tribute the questionnaire to all the students at the start of a class while
Depending on their answer to that question, they were either redirected the professor remained outside of the classroom during the evaluation.
to a series of questions referring to that second course or to the last part Students had at least 15 min to fill out the questionnaire and write
of the survey if they were not enrolled in a second course. Students additional comments. All answers were anonymous and confidential.
went through this process for as many courses as they reported taking Once the procedure was completed, the volunteers were instructed to
during the semester. put all questionnaires in an envelope and seal it before depositing the
envelope in a designated area on campus.
2.2.1. Number of courses (level-2 variable) We selected six items that specifically evaluated the professor's
Before completing course-related questions, participants were asked teaching. Two items assessing the preparedness of the professor and
to indicate the number of courses that they were taking during the their effectiveness in conveying the subject matter were rated on a scale
semester on a scale ranging from 1 to 6. ranging from 1 (almost never) to 5 (almost always). Three items assessing
whether the teaching is stimulating, the course is well organized, and
2.2.2. Year of study (level-2 variable) the expectations of students are clear were rated on a scale ranging
Before completing course-related questions, participants were asked from 1 (strongly disagree) to 4 (strongly agree). One item assessing the
to select their level of academic training from first year to fourth year. professor as a teacher was rated on a scale ranging from 1 (very poor) to
We used contrast coding to compare students who were in their first 5 (excellent). Given that the questionnaire included 4-point and 5-point
year (coded as −1) to students who were in their second, third, and scales with varied labels, we decided to create a sum of all the scores on
fourth year (coded as 0.33). this questionnaire instead of creating an average score. Moreover, we
decided to consider the aggregated teaching evaluation of the students
2.2.3. Procrastination (level-1 and level-2 variable) in a course instead of specifically considering the teaching evaluation of
Three items were used to measure procrastination in each course our participants in our online survey. There is consensus that ag-
based on two items from the Procrastination Assessment Scale – gregating the responses of the students in a course is “the appropriate
Students (Solomon & Rothblum, 1984) and one item from the Irrational unit of analysis” (Marsh, Ginns, Morin, Nagengast, & Martin, 2011, p.
Procrastination Scale (Steel, 2010). Participants were asked to rate the 735). Furthermore, we wanted to minimize common method bias
degree to which they procrastinate, they feel that procrastination is a (Podsakoff, MacKenzie, & Podsakoff, 2012) and the risk that a student's
problem for them, and they delay tasks beyond what is reasonable on a evaluation of his/her professor would be contaminated by the grades

195
K. Kljajic, P. Gaudreau Learning and Instruction 58 (2018) 193–200

obtained in that course (e.g., Grimes, Millea, & Woodruff, 2004). In this models converged at between 10 000 and 45 000 iterations and the PSR
study, the internal consistency of the 6-item questionnaire with the values were between 1.000 and 1.004. The significance testing was
aggregated responses was excellent (within-person α = .97). based on the posterior probability intervals (also known as credibility
intervals). The 95% credibility interval can be interpreted as a “95%
2.2.7. Number of students registered in the course (level-1 variable) probability that the population value is within the limits of the interval”
In the cases of participants who authorized us to get access to their (van de Schoot et al., 2013, p. 844). If the credibility interval did not
grades, the Registrar's Office of the university also provided the number cover zero, the parameter was interpreted as being statistically sig-
of students who were registered in each course during the semester. The nificant. Three sets of models were created to test the hypotheses of this
numbers ranged between 5 and 415 students. study: (a) null model, (b) main model, and (c) complementary model.
The results of all the models can be found in Table 2.
2.2.8. Grade (level-1 and level-2 variable) and admission GPA (level-2
variable)
3.2. Null model
The Registrar's Office provided us the official grades of our parti-
cipants in each of their courses during the winter 2012 semester. The
The null models – also known as unconditional models – were used
following grading scale is used in the university where the study took
to determine the percentage of variance that is attributable to each level
place: 0 = F (0–39%), 1 = E (40–49%), 2 = D (50–54%), 3 = D+
of analysis. The null models can only be estimated with the main
(55–59%), 4 = C (60–64%), 5 = C+ (65–69%), 6 = B (70–74%),
variables (i.e., procrastination and grades) and control variables that
7 = B+ (75–79%), 8 = A- (80–84%), 9 = A (85–89%), and 10 = A+
were measured across courses (i.e., course importance, workload,
(90–100%). The Registrar's Office also provided the admission GPA of
teaching evaluation, and the number of students registered in the
the participants which represents their achievement level before they
course). In these models, each variable was entered separately as an
were admitted to university (e.g., GPA of the last year of high school).
outcome variable at level 1 without including any predictors. The intra-
The admission GPA was converted from a percentage notation to the
class correlation (ICC) was then calculated to estimate the proportion of
grading scale used by the university.
the total variance that could be accounted by between-person variance.
The results revealed that considerable variance in procrastination
3. Results
(38.93%; σ2 = 0.547; τ = 0.858), grades (38.65%; σ2 = 2.468;
τ = 3.918), course importance (54.17%; σ2 = 0.468; τ = 0.396), per-
3.1. Preliminary analyses
ceived workload (97.20%; σ2 = 0.765; τ = 0.022), teaching evaluation
(99.32%; σ2 = 6.387; τ = 0.044) and the number of students registered
The initial sample of 209 undergraduate students was screened for
in the courses (83.80%; σ2 = 8819.314; τ = 1705.429) was attributable
outliers on the main variables of this study. Whenever possible, the
to within-person variability. Taken together, these results demonstrate
correlation between course procrastination and course grade was
the importance of conceptualizing procrastination and grades with
computed for each participant. The correlations were then transformed
multilevel modeling, while also integrating control variables that sub-
into z-scores to facilitate the detection of outliers. A visual inspection
stantially fluctuate across the courses taken by the students during the
did not reveal z-scores that were considerably extreme and different
semester.
from the rest of the sample (i.e., z-scores varied between −1.30 and
2.11). Therefore, we did not exclude participants due to outliers.
However, one participant did not complete any of the questionnaires 3.3. Main model
used in this study and was thus excluded from the analyses. The final
level-2 sample consisted of 208 undergraduate students. Based on a The main model was used to examine the relation between pro-
simulation study, a level-2 sample size of 50 is sufficient to provide an crastination and grades at the between-person and the within-person
accurate estimation of the various parameters, although a sample size levels. In this model, grades were entered as the outcome variable,
of at least 100 is preferable (Maas & Hox, 2005). Thus, our final sample procrastination across courses was entered as a group-centered pre-
size is sufficiently large to conduct the analyses and obtain accurate dictor at level 1, and the average score of procrastination for each
estimates of our parameters. student was entered as a grand-mean centered predictor at level 2. The
Descriptive statistics and correlations for within-person variables level-1 residual variance decreased by 13.45% after adding procrasti-
(i.e., level 1) and between-person variables (i.e., level 2) are presented nation at level 1. The level-2 intercept variance decreased by 2.81%
in Table 1. Multilevel modeling was performed using Mplus (version 8; after adding procrastination at level 2.1 Procrastination was sig-
Muthén & Muthén, 1998–2017) with the Bayes estimator to handle nificantly and negatively associated with grades at both levels of ana-
missing data. We decided to keep the default non-informative priors to lysis, thus supporting Hypotheses 1a and 2a.
ensure that the estimates were computed based only on our data. One of At the between-person level, the finding revealed that students who
the many advantages of using the Bayes estimator compared to the procrastinated more than their colleagues also received lower grades
traditional full information maximum likelihood estimator (ML) is that than their colleagues. Specifically, each 1-point increase in average
it can provide more accurate results when parameters are non-normally procrastination (e.g., going from 2 to 3 in the 5-point procrastination
distributed and skewed (Muthén, 2010; van de Schoot et al., 2013; scale) was related to a drop of 0.433 point in average grade. In other
Zyphur & Oswald, 2013). For many models, the Bayes estimator is also words, the expected average grade at the 15th percentile of average
a more flexible and less burdensome computational method compared procrastination was 6.45 (i.e., between B and B+), whereas the ex-
to ML (Muthén, 2010; Zyphur & Oswald, 2013). In this study, we pected average grade at the 85th percentile of average procrastination
decided to use the Bayes estimator because the traditional full in- was 5.56 (i.e., between C+ and B).
formation ML implemented in Mplus was not able to handle the type of
missing data (i.e., some participants with no data available on the de- 1
The proportional reduction in variance was calculated at each level of
pendent variable) in the context of multilevel modeling with random
analysis with the following equation: (varianceNoPredictor – variancePredictor)/
intercepts and slopes. varianceNoPredictor. The variance refers to the level-1 residual variance (σ2) or
When using a Bayes estimator, it is important to monitor the con- the level-2 intercept variance (r0). The subscript “NoPredictor” refers to the
vergence of the iteration process and increase the number of iterations variance of the null model predicting grades, whereas the subscript “Predictor”
when needed. The Proportional Scale Reduction (PSR) factor is a con- refers to the variance of the model with predictors (see Peugh, 2010, p. 98;
vergence criterion that should be close to 1 (Muthén, 2010). All our Raudenbush & Bryk, 2002).

196
K. Kljajic, P. Gaudreau Learning and Instruction 58 (2018) 193–200

Table 1
Descriptive statistics and correlations between variables at the within-person and between-person levels of analysis.
Variable M SD 1 2 3 4 5 6 7 8 9 10 11

a
Within-person level
1. Course procrastination 2.66 1.17 –
2. Course importance 3.96 0.92 −.10∗ –
3. Course perceived workload 3.21 0.88 .12∗ .09∗ –
4. Course teaching evaluation 24.84 2.53 −.09∗ .10∗ −.04 –
5. Course number of students 182.37 102.34 .01 .02 .08∗ −.04 –
6. Course grade 5.97 2.51 −.24∗ .19∗ −.16∗ .06∗ −.10∗ –
Between-person levelb
7. Average procrastination 2.69 1.02 –
8. Admission GPA 8.04 1.50 −.04 –
9. Number of courses 4.70 0.69 .02 .11 –
10. Year of study – – −.07 −.23∗ −.16∗ –
11. Average grade 6.01 2.11 −.21∗ .40∗ −.08 .20∗ –


Note. p < .05.
a
N = 1248 courses (after dealing with missing data).
b
N = 208 students. Year of study was coded with contrast coding (first year = −1; second, third, and fourth year = 0.33).

Table 2
Results of multilevel models predicting grades.
Null Model Main Model Complementary Model

Estimate [95% PPI] SD Estimate [95% PPI] SD Estimate [95% PPI] SD

Level 1 fixed effect


β00 = Intercept 6.027∗ [5.698, 6.355] 0.168 6.006∗ [5.686, 6.328] 0.164 6.613∗ [6.203, 7.024] 0.210
β10 = Course procrastination −0.580∗ [−0.818, −0.342] 0.120 −0.423∗ [−0.673, −0.182] 0.126
β20 = Course importance 0.260 [−0.003, 0.507] 0.131
β30 = Course workload −0.425∗ [−0.633, −0.202] 0.110
β40 = Course teaching evaluation 0.035 [−0.022, 0.095] 0.030
β50 = Course number of students −0.002 [−0.006, 0.002] 0.002
Level 2 fixed effect
Intercept as outcome
β01 = Procrastination −0.433∗ [−0.750, −0.113] 0.162 −0.357∗ [−0.643, −0.074] 0.145
β02 = Admission GPA 0.639∗ [0.442, 0.840] 0.102
β03 = Number of courses −0.213 [−0.647, 0.214] 0.220
β04 = Year of study 0.986∗ [0.504, 1.477] 0.248

Random effects Variance [95% PPI] SD Variance [95% PPI] SD Variance [95% PPI] SD

σ2 2.468∗ [2.202, 2.781] 0.148 2.136∗ [1.878, 2.441] 0.144 1.733∗ [1.450, 2.061] 0.156
r0 3.918∗ [3.044, 5.080] 0.522 3.808∗ [2.966, 4.908] 0.495 2.940∗ [2.275, 3.813] 0.393
r1 0.393∗ [0.097, 0.870] 0.199 0.196∗ [0.018, 0.600] 0.155
r2 0.142∗ [0.006, 0.621] 0.164
r3 0.326∗ [0.058, 0.747] 0.179
r4 0.005∗ [0.001, 0.021] 0.005
r5 < 0.001∗ [ < 0.001, 0.001] < 0.001

Note. ∗p < .05. PPI = posterior probability interval (also known as credibility interval). SD = standard deviation of the posterior distribution. The parameters are
unstandardized. Random effects were included in all the models. Year of study was coded with contrast coding (first year = −1; second, third, and fourth
year = 0.33). The significance testing was based on the credibility intervals. If the credibility interval did not cover zero, the parameter was interpreted as being
statistically significant.

At the within-person level, the finding revealed that in the courses 3.4. Complementary model
in which students procrastinated more than their own average, they
also received lower grades than their own average. Specifically, each 1- The complementary model was used to test whether procrastination
point increase in course procrastination (e.g., going from 2 to 3 in the 5- could predict grades at the between-person and within-person levels
point procrastination scale) was related to a drop of 0.580 point in over and above control variables. In the complementary model, grades
course grade. In other words, the expected course grade at the 15th were entered as the outcome variable. Course procrastination, course
percentile of course procrastination was 6.68 (i.e., between B and B+), importance, course workload, course teaching evaluation, and the
whereas the expected course grade at the 85th percentile of course number of students registered in the course were entered as group-
procrastination was 5.33 (i.e., between C+ and B). centered predictors at level 1. The average score of procrastination, the
We compared the between-person and within-person relation be- admission GPA, the number of courses taken during the semester, and
tween procrastination and grades to determine whether one is sig- the year of study were entered as grand-mean centered predictors at
nificantly stronger than the other. The result revealed that the relation level 2. The level-1 residual variance decreased by 29.78% after adding
did not significantly differ when comparing the two levels of analysis procrastination and four control variables at level 1. The level-2 in-
(coefficient = 0.148, SD = 0.202, 95% PPI = −0.252, 0.549, p > .05). tercept variance decreased by 24.96% after adding procrastination and
three control variables at level 2.
At the within-person level, higher perceived course workload was

197
K. Kljajic, P. Gaudreau Learning and Instruction 58 (2018) 193–200

significantly associated with lower grades. Course importance, teaching 4.2. Limitations and future directions
evaluation, and the number of students registered in the course were
not significantly associated with grades. At the between-person level, The within-person unit of analysis proposed in this study (i.e.,
lower admission GPA was significantly associated with lower grades. courses) represents only one of the many ways in which procrastination
Moreover, students who were in their first year had significantly lower can be conceptualized within a multilevel perspective. Preliminary re-
grades than students who were in their second, third, and fourth year of sults have shown that university students seem to generally procrasti-
study. The number of courses taken during the semester was not sig- nate more in some life domains (Klingsieck, 2013b) and in some types
nificantly associated with grades. Over and above these control vari- of academic tasks than in others (e.g., Clark & Hill, 1994; Özer et al.,
ables, higher procrastination was still associated with lower grades at 2009). As such, future studies could focus on within-person fluctuations
the between-person and within-person levels. Therefore, both in procrastination across life domains and academic tasks. Alter-
Hypotheses 1b and 2b were supported. natively, future research could enrich the multilevel framework of
academic procrastination by adding levels of analysis. For instance,
academic tasks (level 1) could be nested within courses (level 2),
4. Discussion courses could be nested within semesters (level 3), and semesters could
be nested within students (level 4). This novel multilevel framework
The purpose of this study was to propose a multilevel framework of would offer a richer understanding of the experience of procrastination
procrastination and to investigate the association between procrasti- in post-secondary education.
nation and grades at the between-person and within-person levels of Future studies would be needed to explain why students procrasti-
analysis. Our findings showed a significant and negative association nate more in some courses than in others. Past research conducted at
between procrastination and grades at both levels of analysis. Although the between-person level can inform future research at the within-
the association did not significantly differ in magnitude when com- person level. For example, Steel (2007) found in his meta-analysis that
paring the two levels, the key feature of this study is that the results self-efficacy and task aversiveness were moderately negatively and
hold a distinct conceptual meaning depending on the level of interest. positively associated with procrastination, respectively. At the within-
At the between-person level, the finding revealed that students who person level, we could suggest that students may be more likely to
procrastinated more than other students also achieved lower grades procrastinate in the courses in which they do not believe in their ca-
than other students. At the within-person level, the finding indicated pacity to successfully perform in evaluations (i.e., self-efficacy) or in
that in the courses in which students procrastinated more than their which they perceive the course material and evaluations as being un-
own average, they achieved lower grades than their own average. The pleasant (i.e., task aversiveness). Additionally, some in-class behaviors
relation between procrastination and grades remained significant at known to influence academic achievement – namely laptop usage (e.g.,
both levels over and above various control variables. These results Gaudreau, Miranda, & Gareau, 2014; Kraushaar & Novak, 2010) and
highlight the value and richness of a multilevel framework of procras- cell phone usage (e.g., Clayson & Haley, 2012; Rosen, Lim, Carrier, &
tination by showing the importance of differentiating, comparing, and Cheever, 2011) – could also influence procrastination in some courses.
integrating the construct at two levels of analysis. Thus, future studies are needed to test whether these variables could be
potential within-person antecedents of course procrastination.
The prospective and correlational design used in this study limits
4.1. A multilevel perspective on procrastination the possibility of suggesting that procrastination causally influences
grades. Students are often evaluated throughout the semester in each of
In our study, higher scores of procrastination in some courses came their courses. As such, their final grades generally reflect a weighted
at the price of obtaining lower grades in these courses. University stu- average of all their grades in the courses rather than a single mea-
dents often have to handle the requirements of multiple courses while surement of their achievement at the end of the semester. A cross-
also managing several competing activities and responsibilities (e.g., lagged design – in which procrastination and grades would be measured
work, exercise, and social relationships). Given that time is a limited around the same time and across multiple time points during the se-
resource, even students who are not typically considered as procrasti- mester – could offer some insights on the unidirectional or bidirectional
nators will sometimes irrationally delay the completion of their nature of the relation between procrastination and academic achieve-
schoolwork. This result is consistent with qualitative findings with 67 ment.
students (Schraw, Wadkins, & Olafson, 2007) and quantitative findings The grades obtained in a course could also have an influence on
with more than 2000 students (Schouwenburg, 2004) indicating that procrastination in another course. According to research conducted on
almost all university students procrastinate to a certain extent. More the Dimensional Comparison Theory, higher achievement in a course
studies are needed to fully explore the extent to which procrastination (e.g., math) is related to lower self-concept in another course (e.g.,
across courses can also influence subjective indicators of academic English; see Möller & Marsh, 2013 for a review). Therefore, we could
adjustment and success (e.g., course goal progress and course satisfac- suggest that if students obtain higher grades in math courses, they
tion). might not feel confident in their abilities and even undervalue the
Our findings also revealed that a higher average level of procrasti- importance of language art courses which would lead them to pro-
nation came at the price of obtaining a lower average grade at the end crastinate more in these courses. The Dimensional Comparison Theory
of the semester. Although momentary increases of procrastination from would be difficult to test in a university setting because of the large
time to time seem almost inevitable for most students, some students number of varied courses. However, it would be interesting to in-
display a general pattern of procrastination across all of their courses corporate the hypotheses of this theory within our multilevel model of
which negatively influences their overall academic achievement over procrastination in high schools where students take the same manda-
the long haul. This result is consistent with numerous meta-analytical tory courses that can be compared more easily.
findings showing a negative association between procrastination and In addition to the possibility that grades may influence future pro-
academic achievement at the between-person level of analysis (Kim & crastination, grades obtained during the semester could also influence
Seo, 2015; Richardson et al., 2012; Steel, 2007; van Eerde, 2003). the perception of past procrastination in at least two ways. On the one
Overall, our multilevel study strengthens past findings comparing the hand, students who obtain relatively good grades after having pro-
average score of procrastination of students to one another while also crastinated in their academic tasks might believe retrospectively that
raising attention toward fluctuations of procrastination within the same they started or completed their academic tasks earlier than they actu-
person. ally did. On the other hand, students who obtain relatively good grades

198
K. Kljajic, P. Gaudreau Learning and Instruction 58 (2018) 193–200

after having procrastinated in their academic tasks might forget some Psychological Science, 13, 83–87. https://doi.org/10.1111/j.0963-7214.2004.
undesirable outcomes that were more salient while they were pro- 00280.x.
Fleeson, W., & Noftle, E. E. (2012). Personality research. In M. R. Mehl, & T. S. Conner
crastinating (e.g., anxiety, lack of sleep). Given that procrastination is (Eds.). Handbook of research methods for studying daily life (pp. 525–538). New York,
considered as an irrational delay (e.g., Steel, 2007, 2010), these stu- NY: The Guilford Press.
dents might believe retrospectively that they were not procrastinating Gaudreau, P., Miranda, D., & Gareau, A. (2014). Canadian university students in wireless
classrooms: What do they do on their laptops and does it really matter? Computers &
and that their delay was the best course of action at that time. Future Education, 70, 245–255. https://doi.org/10.1016/j.compedu.2013.08.019.
studies are needed to test whether the perception of past procrastina- Goda, Y., Yamada, M., Kato, H., Matsuda, T., Saito, Y., & Miyagawa, H. (2015).
tion can change after receiving a grade for an academic task (e.g., exam, Procrastination and other learning behavioral types in e-learning and their re-
lationship with learning outcomes. Learning and Individual Differences, 37, 72–80.
term paper, assignment). https://doi.org/10.1016/j.lindif.2014.11.001.
This study was conducted with a sample of university students Grimes, P. W., Millea, M. J., & Woodruff, T. W. (2004). Grades—who’s to blame? Student
which limits our ability to generalize these results to other educational evaluation of teaching and locus of control. The Journal of Economic Education, 35,
129–147. https://doi.org/10.3200/JECE.35.2.129-147.
contexts. University is a complex learning environment where students
Gustavson, D. E., du Pont, A., Hatoum, A. S., Rhee, S. H., Kremen, W. S., Hewitt, J. K.,
are enrolled in various programs with different norms, expectations, et al. (2017). Genetic and environmental associations between procrastination and
and learning agendas. Therefore, the results found in this study might internalizing/externalizing psychopathology. Clinical Psychological Science, 5,
not be comparable to studies conducted in more homogeneous and 798–815. https://doi.org/10.1177/2167702617706084.
Gustavson, D. E., Miyake, A., Hewitt, J. K., & Friedman, N. P. (2014). Genetic relations
structured learning environments (i.e., elementary and high schools) in among procrastination, impulsivity, and goal-management ability: Implications for
which students usually have the same courses with the same professor the evolutionary origin of procrastination. Psychological Science, 25, 1178–1188.
and classmates. Given the substantial contextual variations, future re- https://doi.org/10.1177/0956797614526260.
Gustavson, D. E., Miyake, A., Hewitt, J. K., & Friedman, N. P. (2015). Understanding the
search should examine if the results of this study can be replicated cognitive and genetic underpinnings of procrastination: Evidence for shared genetic
across different educational levels. influences with goal management and executive function abilities. Journal of
Experimental Psychology: General, 144, 1063–1079. https://doi.org/10.1037/
xge0000110.
5. Conclusion Hurst, C. S., Baranik, L. E., & Daniel, F. (2013). College student stressors: A review of the
qualitative research. Stress and Health, 29, 275–285. https://doi.org/10.1002/smi.
A large number of studies have found that students who procrasti- 2465.
Kim, K. R., & Seo, E. H. (2015). The relationship between procrastination and academic
nate more than their counterparts tend to obtain lower grades than performance: A meta-analysis. Personality and Individual Differences, 82, 26–33.
their counterparts. Based on these findings, we would expect Philip to https://doi.org/10.1016/j.paid.2015.02.038.
obtain a lower average grade at the end of the semester compared to Klingsieck, K. B. (2013a). Procrastination: When good things don't come to those who
wait. European Psychologist, 18, 24–34. https://doi.org/10.1027/1016-9040/
Sophia. However, we often overlook students like Sophia who do not
a000138.
have the typical profile of procrastination, but remain at risk of ex- Klingsieck, K. B. (2013b). Procrastination in different life-domains: Is procrastination
periencing negative outcomes as a result of their procrastination in domain specific? Current Psychology, 32, 175–185. https://doi.org/10.1007/s12144-
some courses. This study highlighted the value of a multilevel per- 013-9171-8.
Kraushaar, J. M., & Novak, D. C. (2010). Examining the affects of student multitasking
spective on procrastination while also shedding light on the importance with laptops during the lecture. Journal of Information Systems Education, 21(2),
of studying procrastination as a within-person risk factor for academic 241–251.
difficulties. Maas, C. J. M., & Hox, J. J. (2005). Sufficient sample sizes for multilevel modeling.
Methodology, 1, 86–92. https://doi.org/10.1027/1614-2241.1.3.86.
Marsh, H. W., Ginns, P., Morin, A. J. S., Nagengast, B., & Martin, A. J. (2011). Use of
Author note student ratings to benchmark universities: Multilevel modeling of responses to the
Australian Course Experience Questionnaire (CEQ). Journal of Educational Psychology,
103, 733–748. https://doi.org/10.1037/a0024221.
This research was supported by a scholarship from the Social Möller, J., & Marsh, H. W. (2013). Dimensional comparison theory. Psychological Review,
Sciences and Humanities Research Council of Canada (SSHRC) awarded 120, 544–560. https://doi.org/10.1037/a0032459.
to the first author and a research grant from SSHRC (410-2011-1774) Muthén, B. (2010). Bayesian analysis in Mplus: A brief introductionTechnical Report (ver-
sion 3) . Retrieved from http://www.statmodel2.com/download/IntroBayesVersion
and a teaching release from the Faculty of Social Sciences awarded to
%203.pdf.
the second author. The granting agency had no role in the design, data Muthén, L. K., & Muthén, B. O. (1998-2017). Mplus User's Guide (8th ed.). [Software and
collection, analyses, and writing/reporting of this manuscript. We user guide]. Los Angeles, CA: Muthén & Muthén.
Özer, B. U., Demir, A., & Ferrari, J. R. (2009). Exploring academic procrastination among
thank Gabrielle Martinelli for her help with data screening and the
Turkish students: Possible gender differences in prevalence and reasons. The Journal
members of the Laboratory for research on Achievement, Motivation, of Social Psychology, 149, 241–257. https://doi.org/10.3200/SOCP.149.2.241-257.
and the Regulation of Action (LAMRA) for their feedback on an earlier Peugh, J. L. (2010). A practical guide to multilevel modeling. Journal of School Psychology,
version of this manuscript. 48, 85–112. https://doi.org/10.1016/j.jsp.2009.09.002.
Podsakoff, P. M., MacKenzie, S. B., & Podsakoff, N. P. (2012). Sources of method bias in
social science research and recommendations on how to control it. Annual Review of
References Psychology, 63, 539–569. https://doi.org/10.1146/annurev-psych-120710-100452.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods (2nd ed.). Thousand Oaks, CA: Sage Publications.
Clark, J. L., & Hill, O. W. J. (1994). Academic procrastination among African-American Rice, K. G., Richardson, C. M. E., & Clark, D. (2012). Perfectionism, procrastination, and
college students. Psychological Reports, 75, 931–936. https://doi.org/10.2466/pr0. psychological distress. Journal of Counseling Psychology, 59, 288–302. https://doi.
1994.75.2.931. org/10.1037/a0026643.
Clayson, D. E., & Haley, D. A. (2012). An introduction to multitasking and texting: Richardson, M., Abraham, C., & Bond, R. (2012). Psychological correlates of university
Prevalence and impact on grades and GPA in marketing classes. Journal of Marketing students' academic performance: A systematic review and meta-analysis.
Education, 35, 26–40. https://doi.org/10.1177/0273475312467339. Psychological Bulletin, 138, 353–387. https://doi.org/10.1037/a0026838.
Credé, M., & Kuncel, N. R. (2008). Study habits, skills, and attitudes: The third pillar Rosen, L. D., Lim, A. F., Carrier, L. M., & Cheever, N. A. (2011). An empirical examination
supporting collegiate academic performance. Perspectives on Psychological Science, 3, of the educational impact of text message-induced task switching in the classroom:
425–453. https://doi.org/10.1111/j.1745-6924.2008.00089.x. Educational implications and strategies to enhance learning. Psicologia Educativa,
Dalal, R. S., Bhave, D. P., & Fiset, J. (2014). Within-person variability in job performance: 17(2), 163–177.
A theoretical review and research agenda. Journal of Management, 40, 1396–1436. van de Schoot, R., Kaplan, D., Denissen, J., Asendorpf, J. B., Neyer, F. J., & van Aken, M.
https://doi.org/10.1177/0149206314532691. A. G. (2013). A gentle introduction to Bayesian analysis: Applications to develop-
Dewitte, S., & Schouwenburg, H. C. (2002). Procrastination, temptations, and incentives: mental research. Child Development, 85, 842–860. https://doi.org/10.1111/cdev.
The struggle between the present and the future in procrastinators and the punctual. 12169.
European Journal of Personality, 16, 469–489. https://doi.org/10.1002/per.461. Schouwenburg, H. C. (2004). Procrastination in academic settings: General introduction.
van Eerde, W. (2003). A meta-analytically derived nomological network of procrastina- In H. C. Schouwenburg, C. H. Lay, T. A. Pychyl, & J. R. Ferrari (Eds.). Counseling the
tion. Personality and Individual Differences, 35, 1401–1418. https://doi.org/10.1016/ procrastinator in academic settings (pp. 3–17). Washington, DC: American
S0191-8869(02)00358-6. Psychological Association.
Fleeson, W. (2004). Moving personality beyond the person-situation debate: The chal- Schraw, G., Wadkins, T., & Olafson, L. (2007). Doing the things we do: A grounded theory
lenge and the opportunity of within-person variability. Current Directions in of academic procrastination. Journal of Educational Psychology, 99, 12–25. https://

199
K. Kljajic, P. Gaudreau Learning and Instruction 58 (2018) 193–200

doi.org/10.1037/0022-0663.99.1.12. org/10.1037/0033-2909.133.1.65.
Sirois, F. M. (2014). Out of sight, out of time? A meta-analytic investigation of procras- Steel, P. (2010). Arousal, avoidant and decisional procrastinators: Do they exist?
tination and time perspective. European Journal of Personality, 28, 511–520. https:// Personality and Individual Differences, 48, 926–934. https://doi.org/10.1016/j.paid.
doi.org/10.1002/per.1947. 2010.02.025.
Sitzmann, T., & Yeo, G. (2013). A meta-analytic investigation of the within-person self- Steel, P., & Ferrari, J. (2013). Sex, education and procrastination: An epidemiological
efficacy domain: Is self-efficacy a product of past performance or a driver of future study of procrastinators' characteristics from a global sample. European Journal of
performance? Personnel Psychology, 66, 531–568. https://doi.org/10.1111/peps. Personality, 27, 51–58. https://doi.org/10.1002/per.1851.
12035. Steel, P., Svartdal, F., Thundiyil, T., & Brothen, T. (2018). Examining procrastination
Solomon, L. J., & Rothblum, E. D. (1984). Academic procrastination: Frequency and across multiple goal stages: A longitudinal study of temporal motivation theory.
cognitive-behavioral correlates. Journal of Counseling Psychology, 31, 503–509. Frontiers in Psychology, 9, 1–16. https://doi.org/10.3389/fpsyg.2018.00327.
https://doi.org/10.1037/0022-0167.31.4.503. Zyphur, M. J., & Oswald, F. L. (2013). Bayesian estimation and inference: A user's guide.
Steel, P. (2007). The nature of procrastination: A meta-analytic and theoretical review of Journal of Management, 41, 390–420. https://doi.org/10.1177/0149206313501200.
quintessential self-regulatory failure. Psychological Bulletin, 133, 65–94. https://doi.

200

You might also like