Development and Validation of the Singapore Secondary Mathematics Self-Efficacy Scale

By

Siao Charn Ling

A dissertation submitted in partial satisfaction of the

requirements for the degree of

Doctor of Philosophy

in

Education

in the

Graduate Division

of the

University of California, Berkeley

Committee in charge:

Professor Frank C. Worrell, Chair

Professor Rodolfo Mendoza-Denton

Associate Adjunct Professor Erin Murphy-Graham

Spring 2016
 ProQuest Number: 10150955

All rights reserved

INFORMATION TO ALL USERS

The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

ProQuest 10150955

Published by ProQuest LLC (2016). Copyright of the Dissertation is held by the Author.

All rights reserved.

This work is protected against unauthorized copying under Title 17, United States Code
Microform Edition © ProQuest LLC.

ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346
 1

Abstract

Development and Validation of the Singapore Secondary Mathematics Self-Efficacy Scale

By

Siao Charn Ling

Doctor of Philosophy in Education

University of California, Berkeley

Professor Frank C. Worrell, Chair

To understand the influence of self-efficacy on academic achievement, it is first necessary to

assess the construct with reliability and validity. Although this has been done extensively in the
Western context, studies on the reliability and validity of self-efficacy scores in Asian contexts
have been scarce. The goal of this study is to develop a Mathematics self-efficacy scale with
sound psychometric properties that can be used in future studies to advance our knowledge on
the nature of self-efficacy in different cultural contexts. In this study, I described the
development of a mathematics self-efficacy scale and present data on the reliability and
structural validity of the scores in a sample of Singaporean adolescents (N = 1, 572). Results
indicated that scale scores had strong internal consistency. Exploratory and confirmatory factor
analyses suggested that there were two related but distinguishable self-efficacy variables:
content-specific and general. There was also evidence for convergent validity, given the
significant and positive correlations between the self-efficacy scores and related constructs. As
was found in other studies, the difference between gender and the three self-efficacy scores was
statistically but not practically significant for content-specific self-efficacy; there was not
statistical difference in the general or combined self-efficacy scores. Finally, regression analyses
suggested that self-efficacy scores made significant contributions to mathematics grade, even
after taking into account the contributions of past achievement and other attitudinal constructs.
Further research is needed to address the limitations and to ascertain the generalizability of these
findings.

Keywords: Mathematics self-efficacy; scale development; validity; Singapore

 i

Table of Contents

Acknowledgements ........................................................................................................................ iii

List of Tables ................................................................................................................................. iv

Development and Validation of the Singapore Secondary Mathematics Self-Efficacy Scale ..1
Theoretical Framework of Self-Efficacy .....................................................................................2
Social cognitive theory ............................................................................................................2
Nature and structure of self-efficacy beliefs ............................................................................2
Sources of efficacy beliefs .......................................................................................................3
Relationship Between Self-Efficacy Beliefs and Academic Achievement .................................4
Positive relationship between academic self-efficacy and achievement .................................4
Evidence for the predictive validity of academic self-efficacy ...............................................4
Degree of specificity when assessing self-efficacy .................................................................5
Existing self-efficacy scales .....................................................................................................6
Culture Influences Self-Evaluation ..............................................................................................7
Cultural context of Singapore ................................................................................................10
Research on academic self-efficacy in Singapore .................................................................10
Pilot study for development and validation of a Mathematics self-efficacy scale ................15
Relationship Between Gender and Self-Efficacy ......................................................................15
The Current Study ......................................................................................................................16
Bandura’s guide for constructing a self-efficacy scale ..........................................................16
The current scale ....................................................................................................................16
Hypotheses .............................................................................................................................17

Methods..........................................................................................................................................18
Participants.................................................................................................................................18
Procedure ...................................................................................................................................18
Measures ....................................................................................................................................19
School climate........................................................................................................................19
Mathematics attitude ..............................................................................................................19
Mathematics self-efficacy ......................................................................................................19
Demographics ........................................................................................................................20

Results ............................................................................................................................................20
Preliminary Analyses .................................................................................................................20
Major Analyses ..........................................................................................................................21
Internal consistency and structural validity ...........................................................................21
Convergent validity................................................................................................................24
Concurrent validity ................................................................................................................25
Summary of findings .............................................................................................................25

Discussion ......................................................................................................................................26
Psychometric properties .............................................................................................................26
Convergent Validity ...................................................................................................................27
Concurrent Validity ...................................................................................................................28
 ii

Limitations and Future Research ...............................................................................................29

References ......................................................................................................................................31

Tables .............................................................................................................................................37

Appendix A: School Climate Survey.............................................................................................47
Appendix B: Mathematics Attitude Survey ...................................................................................48
Appendix C: Mathematics Self-Efficacy Survey ...........................................................................49
Appendix D: Motivated Strategies for Learning Questionnaires (MSLQ) – Self-Efficacy ..........51
Appendix E: Background Information...........................................................................................52


 iii

Acknowledgements

The irony never escapes me that in this endeavor to understand the motivation of
Singaporean students, one Singaporean lost motivation countless times. I want to first thank my
parents, Chong Kiet and Gek Huang, who have inculcated in me the importance of education and
who have always told me to strive to be the best that I can be. Thank you for having vision for
me and supporting me in every way. I also want to thank my siblings, Angela and Wui Hou, for
inspiring me to go after my dreams; having your unconditional support have enabled me to
pursue this degree with boldness. I am also grateful for my husband, Travis, and my in-laws,
Anita and Warren, for cheering me on and celebrating every small success along the way.
This dissertation would not have been possible if not for the assistance from the Ministry
of Education (Singapore) and the secondary schools that participated. A special note goes out to
Esther for your encouragement and help with data collection.
Through the years of graduate school, I am deeply indebted to my advisor, Dr. Frank
Worrell. Thank you for teaching me how to write, which has been as much a lesson on
grammar, as it has been on humility and perseverance. Thank you for always being available to
answer my questions, for teaching me so much about statistics, and for reminding me to stay the
course when the challenges seemed insurmountable. I would also like to acknowledge my
qualifying exam and dissertation committee: Dr. Susan Holloway, Dr. Kaiping Peng, Dr. Erin
Murphy-Graham, and Dr. Rodolfo Mendoza-Denton. Thank you for your feedback and
comments, challenging me to deepen and broaden my understanding, refining my thought
processes, and finally, producing this work that I am proud of.
I would also like to express my gratitude to the School Psychology Faculty: Dr. Carolyn
Hartsough, Dr. Kate Perry, Dr. Kathleen Donohue, Dr. Alisa Crovetti, and Dr. Gary Yabrove.
Thank you for setting me on the path to becoming a student-centered school psychologist, for
developing my competence, as well as instilling in me the belief that I can make a difference. I
also want to thank my cohort (Kandace, Olivia, Leo, Lissa, Marie) and other fellow students,
whose feedback have transformed my work and prepared me well to meet each significant
educational milestone.
For the last two years while I worked on this dissertation, I have also been working as a
school psychologist. It was not easy managing multiple responsibilities, and I am thankful for
my co-workers for always being understanding and accommodating. A big thank you to Sarah,
Karly, Kelly, Rebecca, Lilla, Sophia, Elizabeth, Sherri, Erin, and the principals, deans, and
teachers I have worked with. You have also taught me so much about engaging in best practices
and your dedication inspires me to continue learning and growing as a school psychologist.
Throughout my time in Berkeley, my friends from Gracepoint Church have sustained me
with their prayers, encouragement, care packages, and home cooked meals. Thank you for
loving me in your thoughts and actions, for checking in to make sure I am doing fine, and for
sharing in all my burdens and my joys.
This journey began when I worked as a substitute teacher in 2006. What I thought was a
temporary job left an indelible mark on me; the struggles of the students who told me to not
waste my time on them and the students who had given up on themselves became my own
mission and my motivation. I thank God for sharing His compassion for the weak with me and
for opening my eyes to see the needs of others. I thank God for bringing me to Berkeley, where I
gained more than a degree, where through this roller coaster of a journey, I got to experience that
indeed when I am weak, He is strong; His grace is always sufficient for me.
 iv

List of Tables

Table 1. Descriptive Statistics for Independent Variables .............................................................37

Table 2. Correlation Matrix of Major Variables ............................................................................39
Table 3. Reliability Estimates for Independent Variables .............................................................40
Table 4. Measures of Factorability of Scores ................................................................................41
Table 5. Fit Indices Derived from Confirmatory Factor Analyses ................................................42
Table 6. T-tests on Gender, Mathematics Achievement, and Mathematics Self-Efficacy ............43
Table 7. Hierarchical Multiple Regression on Attitudinal Constructs, Content-Specific, and
General Self-Efficacy Variables ....................................................................................................44
Table 8. Hierarchical Multiple Regression on Attitudinal Constructs and Combined Self-Efficacy
Variable ..........................................................................................................................................45
Table 9. Hierarchical Multiple Regression on Past Achievement and Self-Efficacy ....................46


 1

Development and Validation of the Singapore Secondary Mathematics Self-Efficacy Scale

In international assessments of students’ academic achievement like the Trends in
International Mathematics and Science Study (TIMSS) and the Progress in International Reading
Literacy Study (PIRLS), students from East Asian countries are consistently the top performers
(e.g., Mullis, Martin, Foy, & Arora, 2012). One country that has performed well at international
assessments is Singapore. Of the 34 countries that participated in TIMSS and PIRLS in 2011,
students from Singapore was ranked fourth in reading literacy (Mullis, Martin, Foy, & Drucker,
2012), and top two in mathematics and science amongst fourth- and eight-grade students (Mullis,
Martin, Foy, & Arora, 2012; Martin, Mullis, Foy, & Stanco, 2012). To understand why students
from Singapore have such remarkable
Introduction
achievements, a good starting placePartisA to examine their
framework for learning.
According to the Secondary Mathematics Syllabuses in Singapore (Ministry of Education,
3 MATHEMATICS FRAMEWORK
2006), the ability to solve mathematics problems is at the core of mathematics learning. The
This framework shows the underlying principles of an effective mathematics
Ministry of Education developed
programmeathat framework
is applicable to that included
all levels, fivetointerrelated
from the primary components that
A-levels. It sets the
contribute to effective teaching, learning, and assessment of mathematics. These five
direction for the teaching, learning, and assessment of mathematics.

components are attitudes, metacognition, processes, concepts, and skills (see Figure 1).
Beliefs
Interest Monitoring of one’s own thinking
Appreciation Self-regulation of learning
Confidence
Perseverance

Numerical calculation
Algebraic manipulation
Spatial visualisation Reasoning, communication
Data analysis and connections
Measurement Thinking skills and heuristics
Use of mathematical tools Applications and modelling
Estimation

Numerical
Algebraic
Geometrical
Statistical
Probabilistic
Analytical

Figure 1. Mathematics framework

Mathematicalfor thesolving
problem teaching,
is centrallearning, and
to mathematics assessment
learning. of mathematics.
It involves the
acquisition and application of mathematics concepts and skills in a wide range of
In order to assess students’ learning, written examinations are conducted in most
situations, including non-routine, open-ended and real-world problems.
th
secondary schools twice a year.
The At theof end
development of 10problem
mathematical 12thability
andsolving grades, students
is dependent
related components, namely, Concepts, Skills, Processes, Attitudes and
sit for the General
on five inter-

Certificate of Education (GCE) O-Levels Exams and GCE A-Levels Exams respectively. These
Metacognition.

exams provide a single grade that reflects a student’s mathematics learning, but if a student does
poorly, it is not clear which of the five components, or combination of several components, the
student is lacking in. In order to enhance students’ mathematics learning, it is necessary to have
the means to identify the component, or components, students have not developed sufficiently.
Of these five components, I have chosen to focus on attitudes in this dissertation.
Although a myriad of beliefs have been found to be related to academic outcomes (e.g.,
attribution beliefs, self-concept, self-esteem), I have 2chosen to focus on academic self-efficacy
beliefs, because academic self-efficacy beliefs have been shown to have greater predictive
validity than other self-perception constructs such as self-concept (Pajares & Miller, 1994) and
locus of control (Smith, 1989).
Bandura (1977) defined self-efficacy as one’s perceived capabilities to organize and
execute courses of action at a specific level of difficulty, in order to attain one’s goals in a
specific context. Since the conception of self-efficacy, thousands of studies have examined self-
efficacy in relation to other variables and have found self-efficacy to be a significant predictor of
 2

various behaviors and outcomes. For example, self-efficacy beliefs were found to be significant
predictors of athletic performance (e.g., Feltz & Lirgg, 2001; Martin & Gill, 1995), intentions to
engage in healthy lifestyles (e.g., Nahas, Goldfine, & Collins, 2003; Wu, Pender, & Noureddine,
2003), and academic achievement (e.g., Chemers, Hu, & Garcia, 2001; Ferla, Valcke, & Cai,
2009).
However, many of the studies that lend support to the predictive relationship between
self-efficacy and academic achievement were conducted in Western cultures with participants of
European descent. It is not possible to determine from these studies if the predictive validity of
self-efficacy scores is universal, or whether cultural norms and values could influence the
predictive validity of self-efficacy. Cross-cultural studies on self-efficacy have reported
contradictory results regarding the relationship between self-efficacy and achievement in the
Eastern contexts (e.g., Eaton & Dembo, 1997; Salili, Chiu, & Lai, 2001). Further, many of the
current studies did not validate the self-efficacy scores used, and given the context- and task-
specificity of self-efficacy (Bandura, 2006), that should be the first step in assessing the nature of
self-efficacy in different cultures.
This dissertation study – the development and validation of a Mathematics self-efficacy
scale – was conceptualized to respond to the need to have an accurate means of assessing
mathematics self-efficacy in Singapore and the limited validation of self-efficacy scores within
the current literature. The goal of this study is to develop a Mathematics self-efficacy scale with
sound psychometric properties that can be used in future studies to advance our knowledge on
the nature of self-efficacy in different cultural contexts. I will first review the literature on the
theoretical framework of self-efficacy, the relationship between self-efficacy beliefs and
academic achievement, and the role culture plays in influencing self-evaluation. Following this
review of literature, I will present the study’s hypotheses, methods, results, and discussion.
Theoretical Framework of Self-Efficacy
Social cognitive theory. Bandura (1986) proposed a social cognitive theory of human
functioning that emphasizes the interactive nature or individuals and their environment. This
theory contains three elements: (a) interpersonal factors in the form of cognitive, affective, and
biological events; (b) human behavior; and (c) environmental factors. Bandura (1997) proposed
that human’s development of competencies and regulation of action are the results of the
dynamic interplay among these three elements. According to this theory, people are viewed as
self-regulating, proactive, and self-reflecting, and not merely reacting to and getting shaped by
the environment. The reciprocal nature of human functioning implies that to effect changes,
people can make alterations on personal factors, behavior, or environmental factors (Pajares &
Valiante, 2006). For example, for a teacher to increase the reading competence of students, the
teacher can work on improving the students’ reading self-efficacy (personal factor), engaging
students in interesting reading activities (behavior), and creating a conducive learning
environment (environmental factor).
Nature and structure of self-efficacy beliefs. Self-efficacy beliefs are an important
aspect of Bandura’s (1986) social cognitive theory of human functioning. Bandura (1997)
proposed that in order for people to function effectively in any domain, people not only need the
skills of that domain but also the efficacy beliefs that they can apply these skills well. Self-
efficacy beliefs will influence people’s choice of activities and their motivation level, which in
turn contribute to or prevent people’s acquisition of more knowledge and skills (Bandura, 1997).
Self-efficacy is a domain specific construct (Bandura, 2006). For example, a professional
baseball player may have high self-efficacy for baseball but at the same time have low self-
 3

efficacy for golf. Further, within each domain of self-efficacy, there are many facets that
contribute to individuals’ holistic sense of their own capabilities. For example, in the domain of
Mathematics self-efficacy, being able to perform computations alone may not lead to high self-
efficacy in Mathematics. For a person to have high Mathematics self-efficacy, the person may
also have to perceive himself as capable in other areas of Mathematics, for example, being able
to solve geometry questions, complete Mathematics homework without help, and score an “A” in
the next Mathematics class test. Given the domain specificity and multifaceted nature of self-
efficacy, Bandura (2006) cautioned researchers developing measures of self-efficacy to ensure
that the ways self-efficacy are being defined and measured reflect both the domain and context
characteristics.
Many scales have been developed to measure self-efficacy in the academic context (e.g.,
the Children’s Multidimensional Self-Efficacy Scale, Bandura, 1989; Motivated Strategies for
Learning Questionnaire, Pintrich & DeGroot, 1990; Pattern for Adaptive Learning Survey,
Midgley et al., 1993) but many of them were not subject- or domain-specific. Given the domain-
specific nature of self-efficacy (Bandura, 2006), it is useful to examine academic self-efficacy
scales that are subject-specific. In this paper, I will focus my review on Mathematics self-
efficacy. Adapting Bandura’s (1977) definition of self-efficacy, I define Mathematics self-
efficacy as a person’s perceived capabilities to plan and perform Mathematics tasks at specific
levels of difficulty, in order to achieve academic goals in a specific context.
Sources of efficacy beliefs. Self-efficacy may be developed through four main forms of
influence (Bandura, 1995). According to Bandura (1995), the most effective way of developing
a strong sense of self-efficacy is through mastery experiences. This refers to the positive
feedback a person experiences when a person successfully creates and executes courses of action
to achieve a goal. Alternatively when a goal is not met, this diminishes a person’s sense of self-
efficacy; however, this latter outcome does not mean failures are bad in the development of a
high sense of self-efficacy. For self-efficacy beliefs to be resilient, a person must experience
overcoming obstacles through perseverance, or success may be taken for granted and even a
person with high self-efficacy may be easily setback by failures (Bandura, 1995).
The second way self-efficacy beliefs can be developed is through vicarious experiences.
Bandura (1986) proposed that observing similar others succeed through perseverance could raise
a person’s self-efficacy beliefs about his own capabilities to achieve comparable results. On the
other hand, witnessing similar others fail despite perseverance can diminish a person’s self-
efficacy beliefs (Brown & Inouye, 1978). The impact of social modeling is dependent on two
criteria: (a) the more similar an individual is to the person being observed, the greater the
influence of modeling; and (b) the persevering attitudes displayed by the models count for more
than the actual skills that are being employed to overcome the obstacles (Bandura, 1995).
The third way self-efficacy can be developed is through social persuasion, or being
verbally told that one has the capabilities to succeed (Bandura, 1995). Schunk (1989) suggested
that when a person facing difficulties in completing a task is being encouraged verbally, the
person would expend greater effort and sustain the effort, rather than dwell on self-doubts.
Bandura (1995) argued that social persuasion on its own is perhaps the least effective way to
strengthen self-efficacy, as overly optimistic encouragement can be quickly dismissed by
negative outcomes. Another caveat is that regardless of what is being said, if a person is deeply
convinced of her own lack of abilities, the person will be less willing to make effortful attempts,
or will give up readily at the first sign of difficulty (Bandura, 1995).
 4

The fourth way self-efficacy may be developed is through a person’s physiological and
emotional states. The way individuals perceive and interpret their physical and emotional
reactions influence their self-efficacy. Ewart (1992) found that in activities where strength and
stamina are required, people interpreted their fatigue and aches as physical weakness. Kavanagh
and Bower (1985) found that people experiencing positive moods during a task have elevated
levels of self-efficacy, but people experiencing pessimistic moods have diminished levels of self-
efficacy beliefs.
Relationship Between Self-Efficacy Beliefs and Academic Achievement
Many studies have shown that academic self-efficacy is a more powerful predictor of
academic achievement compared to other self-perception constructs (e.g., self-concept vs. self-
efficacy, Pajares & Miller, 1994; self-esteem vs. self-efficacy, Mone, Baker & Jeffries, 1995;
outcome expectations vs. self-efficacy, Lent, Lopez, & Bieschke, 1993; locus of control vs. self-
efficacy, Smith, 1989). However, these studies have not explained the mechanisms through
which self-efficacy exerts greater influence on achievement.
Positive relationship between academic self-efficacy and achievement. According to
Zimmerman and Cleary (2006), academic self-efficacy influences important aspects of
motivation and achievement. Bandura (1977) defined motivation as the amount of effort a
person puts in, a person’s persistence, and a person’s choice of activities. Research relating
academic self-efficacy to motivation has repeatedly shown that students with higher self-efficacy
will expend greater effort and persevere longer in the face of difficulties than students with lower
self-efficacy (e.g., Schunk, 1995; Zimmerman & Ringle, 1981). Similarly, students with higher
self-efficacy will more readily undertake a challenging task than students with lower self-
efficacy (e.g., Bandura & Schunk, 1981; Zimmerman & Kitsantas, 1999). It can be expected that
the more effort a person puts in and the longer a person persists in the face of difficulties, the
more likely this person will succeed. Given the positive relationship between academic self-
efficacy and motivation, it is reasonable to propose that a positive relationship exists between
academic self-efficacy and academic achievement as well. To substantiate this claim, I will
review studies examining the predictive relationship between academic self-efficacy and
achievement in the following section.
Evidence for the predictive validity of academic self-efficacy. Zimmerman, Bandura,
and Martinez-Pons (1992) conducted a study to examine the causal impact of students’ self-
efficacy beliefs, academic goals, and learning strategies on academic attainment. There were
two components to self-efficacy in this study: self-efficacy for self-regulated learning (11 items,
α = .87) and self-efficacy for academic attainment (9 items, α = .70). Self-efficacy for self-
regulated learning was operationalized as students’ perceived capability to use self-regulated
learning strategies. Self-efficacy for academic attainment was operationalized as students’
perceived ability to achieve in a variety of academic domains (e.g., mathematics, biology,
reading, and writing). The self-efficacy items in this study were taken from Bandura’s (1989)
Children’s Multidimensional Self-Efficacy Scale. The researchers hypothesized that self-
efficacy for self-regulated learning would affect self-efficacy for academic attainment, which
would, in turn, affect final grades.
The self-efficacy questionnaire was administered to 102 ninth and tenth graders at the
beginning of a semester, and teachers provided the students’ social studies grades at the end of
the semester. The researchers found a significant causal path between self-efficacy for self-
regulated learning and self-efficacy for academic achievement (p = .51), and a significant causal
path between self-efficacy for academic achievement and final grades (p = .21).
 5

The study by Zimmerman et al. (1992) provides support for the claim that self-efficacy
for academic attainment is a significant predictor of academic achievement. Similar findings
were reported by Pajares and Kranzler (1995), where the researchers examined the influence of
mathematics self-efficacy and general mental ability on mathematics problem solving
performance. Self-efficacy was operationalized as students’ confidence in their abilities to solve
a variety of mathematics problems. The 18-item scale was adapted from Dowling’s (1978)
Mathematics Confidence Scale (MCS) and the researchers obtained an internal consistency
coefficient of .92 for scores on the instrument with their sample of 329 high school students.
Path analysis was used to examine the direct and indirect influence of the variables (e.g.,
self-efficacy, general mental ability) on mathematics performance. The researchers found that
mathematics self-efficacy (β = .35) had a direct influence on mathematics performance that
paralleled the influence of general mental ability (β = .32). Overall, the findings of this study
provide strong support for the predictive validity of mathematics self-efficacy in relation to
mathematics achievement.
Degree of specificity of when assessing self-efficacy. Given the domain- and context-
specificity nature of self-efficacy, Bandura (1997) had proposed that self-efficacy has the
greatest predictive validity for achievement when self-efficacy is measured at the same level of
specificity as the criterion variable. For example, if the criterion variable is performance in a
mathematics exam, items on a self-efficacy scale that measures an individual’s perceived ability
to solve mathematics problems would have greater predictive validity than items on a self-
efficacy scale that measures an individual’s perceived ability to do well in school. Research has
backed up Bandura’s claim.
In one study, Lent, Brown, and Gore (1997) examined the discriminant and predictive
validity of various self-efficacy and self-concept constructs at varying levels of specificity. The
researchers administered five scales to 205 college students: mathematics course self-efficacy,
mathematics problem self-efficacy, academic self-efficacy, academic self-concept, and academic
adjustment. Mathematics course self-efficacy was operationalized as students’ confidence to
complete a variety of math-intensive college courses and obtain grade of B or better.
Mathematics problem self-efficacy was operationalized as students’ perceived ability to solve 18
math problems. Academic self-efficacy was operationalized as students’ confidence to perform
12 general academic tasks (e.g., “Obtain a grade point average of at least 3.0”). Academic self-
concept was operationalized as students’ perception of their academic self (e.g., “Most courses
are easy for me,” “I am satisfied with the class assignments I turn in”). Finally, academic
adjustment was operationalized as students’ feelings about how well they were managing the
educational demands of college life. The criterion variables were overall achievement,
mathematics achievement, and choice of mathematics courses.
Through confirmatory factor analyses, the researchers found that the self-efficacy and
self-concept variables were related but empirically distinct. The researchers also found that both
mathematics self-efficacy scores (i.e., course and problem) had the highest correlations with
mathematics achievement (rcourse = .52, p < .01; rproblem = .56, p < .01), followed by academic
self-concept (racademic self-concept = .22, p < .01), and then academic self-efficacy (racademic self-efficacy
= .16, p < .05); academic adjustment was not significantly related to mathematics achievement
(radjustment = .09, p > .05). Using structural equation modeling, the researchers found that the two
mathematics self-efficacy variables were the only predictors that produced significant paths to
choice of mathematics courses (βcourse = .24 and βproblem = .21), academic self-efficacy variable
was the only variable that produced significant path to mathematics achievement (β = .30), and
 6

academic self-concept was the only variable that produced significant path to overall
achievement (β = .31). It was surprising to the researchers that academic self-efficacy score,
rather than either of the mathematics self-efficacy scores, contributed most strongly to the
prediction of mathematics achievement. However, the overall pattern is consistent with
Bandura’s (1997) claim that self-efficacy and other constructs are more useful predictors when
they are matched to the outcome variable to its domain or level of specificity.
In a similar study, Choi (2005) examined how self-efficacy and self-concept constructs of
varying levels of specificity related to college students’ academic performance. 230
undergraduate students participated in this study. The researcher measured self-efficacy at three
levels of specificity: general (e.g., “When I make plans, I m certain I can make them work”),
academic (i.e., students’ confidence to perform academic tasks like note taking), and course-
specific (i.e., students’ perceived abilities to perform academic tasks like memorization or
understanding in specific course). Self-concept was measured at two levels of specificity:
academic (e.g., “I consider myself a good student”) and course-specific (e.g., “I learn things
quickly in this type of course”). The criterion variable was course grade at the end of the term.
The researcher found that amongst the three self-efficacy variables, course grade had the
highest correlation with course-specific self-efficacy (r = .32, p < .01), followed by academic
self-efficacy (r = .22, p < .01), and general self-efficacy did not have a significant correlation
with course grade (r = .14). In a simultaneous regression of course grade on the three self-
efficacy variables, course-specific self-efficacy was the only significant predictor (β = .27, p
< .01). Between the self-concept variables, course grade had higher correlation with course-
specific self-concept (r = .42, p < .01) than academic self-concept (r = .38, p < .01). In another
simultaneous regression of course grade, both academic and course-specific self-concepts were
significant predictors, and course-specific self-concept was a relatively more significant
predictor (βcourse-specific = .31; βacademic = .22). Overall, this study also suggests that when the
predictor variable is matched to the outcome variable in level of specificity, the more predictive
power the predictor variable has.
Existing self-efficacy scales. In this section, I will review the properties of several
Mathematics self-efficacy scales, discuss where these scales have been used and validated, as
well as their limitations for use in cross-cultural studies.
Mathematics self-efficacy scales. In this section, I will be reviewing two Mathematics
self-efficacy scales, their psychometric properties, as well as their limitations.
Mathematics Self-Efficacy Scale (MSES-1, Betz & Hackett, 1983). Betz and Hackett
developed the 52-item MSES-1 with a United States college sample. The researchers
operationalized mathematics self-efficacy as students’ perceptions of their capability to solve
mathematics problems (18 items), everyday mathematics tasks (18 items), and mathematics-
related college coursework (16 items). Students were required to indicate on a 10-point rating
scale how confident they were of solving the problems posed in the items. Four scores were
obtained: one score for each subscale and a total score.
The internal consistency reliabilities for scores on the three subscales were .92
(mathematics problem), .90 (mathematics task), and .93 (mathematics course). The researchers
found that the mathematics self-efficacy total score was significantly correlated with
mathematics anxiety (r = -.56) and mathematics confidence (r = .66). In a stepwise multiple
regression with choice of college major as the dependent variable, the researchers found that
mathematics self-efficacy was a significant predictor of students choosing science majors (β
= .24).
 7

Mathematics Self-Efficacy Scale (MSES-2, Nielsen & Moore, 2003). Nielsen and Moore
developed the Mathematics Self-Efficacy Scale for use with ninth- and tenth-grade students in
Victoria (a state in Australia). The researchers used three criteria in developing the scale. First,
the items specifically referenced Victoria’s curriculum. Second, the items asked for students’
perceived abilities, rather than what they would actually do. Third, the items included a variety
of domains (e.g., algebra, geometry). Eventually, nine mathematics problems were selected to
make up the mathematics self-efficacy scale. Students were required to indicate on a 5-point
rating scale how confident they were of their abilities to perform each of the nine items. Two
hypothetical situations were presented to each student during questionnaire administration. In
the first situation, students were told to judge their confidence level when solving problems in a
classroom context. In the second situation, students had to judge their confidence in a test
context. There were three self-efficacy scores altogether: one for each of the two hypothetical
situations, and a total self-efficacy score from combining scores in the two situations.
The internal consistency reliabilities for the two situations’ scores were .86 (class)
and .90 (test), and the reliability coefficient for the total score was .93. Concurrent validity was
established with past mathematics grade (rclass = .51; rtest = .58), and desired mathematics grade
in the future (rclass = .52; rtest = .58). The researchers conducted a principal component analysis
on the combined scores. One component was extracted and this component (labeled
mathematics self-efficacy) explained 49% of the score variance. The component coefficients
ranged from .53 to .78.
Limitations of the mathematics self-efficacy scales. Although these two mathematics self-
efficacy scales had adequate psychometric properties, one of the limitations of the scales is that
both of them had questions designed for specific populations: the United States college
population (Betz & Hackett, 1983) and Victoria’s ninth- and tenth-grade students (Nielsen &
Moore, 2003). Content wise, what is relevant for a college student in the United States (e.g.,
degree of confidence in balancing a checkbook without a mistake) or a high school student in
Australia (e.g., degree of confidence in sketching a curve) may not be relevant for students in
another age group or another context.
Given the high relevance to the specific contexts, one cannot readily apply these scales to
other contexts without first validating these scales in the new contexts. On the other hand,
creating a self-efficacy scale with generalizability may come at the expense of the context-
specificity. Indeed, Bandura (1997) and Pajares (1997) had both previously cautioned that self-
efficacy beliefs have to be assessed “at the optimal level of specificity that corresponds to the
critical task being assessed” (Pajares, 1997, p. 11). However there is no consensus on what the
optimal level may be and there is also no guidance on how to develop a self-efficacy scale that is
both content-specific and generalizable to different contexts.
Culture Influences Self-Evaluation
Another question that arises from the context-specificity nature of self-efficacy is how
culture influences the conceptualization of self-efficacy. According to Bandura (2002), self-
efficacy is not a construct borne out of an individualistic culture. Bandura contend that although
self-efficacy beliefs may develop differently and carry different meanings in various cultures,
self-efficacy beliefs are essential in all cultures. For example, in a culture that values group work,
beliefs about personal capabilities to perform one’s responsibilities are important motivation for
a member to work towards the collective group success. Bandura cautioned that study of self-
efficacy cannot be reduced to polarities (e.g., individualistic vs. collectivistic) because across-
group analyses can mask-within-group difference and a lower self-efficacy belief within one
 8

group of people does not mean that the construct is less important. Here, I review some studies
that examined the intersection between culture and self-efficacy.
In a review study, Klassen (2004) examined the findings from 20 studies that looked at
self-efficacy from a cross-cultural perspective. The studies that Klassen reviewed had either an
educational or business focus, the studies included more than one cultural group (Asian or
immigrant Asians versus Western, and Eastern European versus Western European or American),
and the studies examined self-efficacy or collective efficacy. Klassen had two research questions:
how does self-efficacy vary across cultures? Do self-efficacy beliefs appear to be influenced by
cultural dimensions like individualism and collectivism? Before I summarize Klassen’s findings,
I will review two of the studies included in his analysis that specifically had an academic focus,
comparing Asians and Western students. After reviewing these two studies, I will comment on
their findings, as well as integrate Klassen’s summary.
In one of the studies reviewed by Klassen (2004), Eaton and Dembo (1997) examined
how motivational beliefs (i.e., an incremental view of ability, fear of failure, academic self-
efficacy, and effort beliefs) differentially predicted the achievement behavior for Asian
American students and a group of non-Asian students. All the participants in this study were
ninth-grade students and achievement behavior was operationalized as students’ performance on
a 20-item novel word unscrambling task. The score for this dependent variable was the total
number of correct responses. The instrument used to measure students’ beliefs in this study was
an 80-item questionnaire adapted from several different sources, and six of the items measured
self-efficacy beliefs. Self-efficacy was operationalized as students’ perceptions of how
successful they would be on the novel task. The source from which the self-efficacy items were
drawn was not specified but the total score based on the six items had an internal consistency
estimate of .78 in this sample.
The researchers found that the Asian American students outperformed their non-Asian
counterparts on the novel task, t(452.77) = 9.39, p < .001, d = .81, but the Asian American
students also had lower levels of academic self-efficacy than the non-Asian students, t(287.28) =
2.25, p < .05, d = .20. In multiple regression analyses, where performance on the novel task was
the dependent variable, the researchers found that fear of failure was a better predictor of
achievement than self-efficacy beliefs in the Asian American sample. On the other hand, self-
efficacy was a significant predictor of achievement in the non-Asian sample but fear of failure
was not.
Eaton and Dembo’s (1997) findings suggest that self-efficacy is affected by cultural
influences and may not be as good a predictor of achievement for Asian students. However, the
way self-efficacy was operationalized in this study may have deviated from Bandura’s (1977)
definition of self-efficacy. Self-efficacy as operationalized in Eaton and Dembo’s (1977) study
focused on the students’ perceived outcome expectations, rather than the student’s perceptions of
abilities to organize and perform specific tasks.
In another study reviewed by Klassen (2004), Salili, Chiu, and Lai (2001) compared the
achievement and self-efficacy of three groups of high school students: Hong Kong nationals,
Chinese Canadians, and European Canadians. A two-part questionnaire was administered to the
students. In the first part, demographic information was collected, and in the second part,
students filled in measures of learning attitudes, including an academic self-efficacy measure
adapted from the Patterns of Adaptive Learning Survey (PALS; Midgley, Maehr, & Urdan,
1993). The academic self-efficacy items assessed students’ perceived abilities to perform
general school-related tasks (e.g., If I have time, I can do a good job on my school work.), and it
 9

was not subject specific. The internal consistency for the attitude measure scores ranged
from .72 to .89.
There were three findings in this study that are related to self-efficacy. First, the
researchers found that the Hong Kong students had significantly lower academic achievement, F
(2, 565) = 408, p < .0001, d = .36, and lower levels of self-efficacy than the other two groups of
students, F (2, 564) = 40.4, p < .0001, d = .17. Second, the researchers found that although
academic self-efficacy had positive significant relationships with academic achievement for all
three groups of students, this relationship was the weakest for the Hong Kong students (Hong
Kong students: r = .19; Chinese Canadians: r = .46; European Canadians: r = .35). Third, in
multiple regression analyses where academic achievement was the dependent variable, the
researchers found that self-efficacy was a significant predictor for all three groups of students
(Hong Kong students: β = .26; Chinese Canadians: β = .45; European Canadians: β = .27).
In sum, the findings by Salili et al. (2001) support the findings of the self-efficacy studies
conducted in the Western contexts. Specifically, the researchers found that regardless of the
participants’ cultural background, academic self-efficacy is a statistically significant positive
predictor of achievement. However, there were a few points to take note about this study. First,
it was interesting that Hong Kong students in this study had the lowest grades among the three
groups, when many studies have repeatedly found that Asian students outperform their non-
Asian peers (e.g., Steinberg, Dornbusch, & Brown, 1992; Stevenson & Lee, 1996). It is unclear
from this study whether Hong Kong students with higher grades will also have higher levels of
self-efficacy. Second, it is not known whether the correlation between their grades and self-
efficacy levels would be just as strong or stronger for the Hong Kong students if they had higher
self-efficacy. Third, the self-efficacy measure used in this study has not been validated in Hong
Kong. Lastly, the self-efficacy items were about general school work and they may not have
taken into account the contextual features of the educational and cultural climates in both
countries.
The results reported by Eaton and Dembo (1997) and Salili et al. (2001) leave open the
possibility that there could be cultural differences in self-efficacy beliefs. Salili’s et al. (2001)
findings also suggest that academic self-efficacy could be an important predictive variable of
academic achievement in Asia. One of the reasons for these contrasting findings could be
because Asia is very heterogeneous in its cultural makeup; the cultural influences on self-
efficacy for Asian American students may not be the same as for Hong Kong or Chinese
Canadian students.
Like these two studies, Klassen’s (2004) finding was also not straightforward. He
concluded that although most of the studies he reviewed suggested that there are cross-cultural
differences in the levels of self-efficacy beliefs, there was also evidence that self-efficacy beliefs
are predictive of performance in all cultural contexts. Klassen suggested that a possible
explanation for the differences in levels of efficacy beliefs could be a difference in response style,
rather than a difference in beliefs. Past research (e.g., Kitayama, Markus, Matsumoto, &
Norasakkunikit, 1997) has shown that people from a more individualistic culture tend to enhance
in their self-reports, whereas people from a more collectivistic culture tend to efface in their self-
reports. However since reports of self-efficacy beliefs tend to be goal-referenced rather than
normative-referenced like other self-constructs (e.g., self-concept, self-esteem), it is not clear
how much self-enhancement or self-effacement influence self-efficacy reports.
Another possible explanation Klassen (2004) offered for the differences in levels of
efficacy beliefs was that people from a more individualistic culture may be more optimistic in
 10

predicting their ability or performance, whereas people from a more collectivistic culture may be
more realistic in predicting their ability or performance. Therefore, although Asians tend to have
lower levels of self-efficacy beliefs, their beliefs are still predictive of their subsequent
performance.
Whether it is due to self-effacement or having a more realistic prediction of one’s ability,
studies are suggesting that there are cultural differences in self-efficacy beliefs. Given that self-
efficacy is a context-specific construct and heeding Bandura’s (2002) caution to not oversimplify
cultures into polarities, it is imperative that I now present a cultural context of Singapore and
then I will review studies to provide a background on how academic self-efficacy manifest in the
cultural context of Singapore.
Cultural context of Singapore. Singapore is made up of an island and islets in the heart
of Southeast Asia. The country is approximately 697 kilometers square in area (Central
Intelligence Agency, 2016). Based on the statistics compiled by the Central Intelligence Agency
on February 2016, the country has a population of about 5.67 million. This population includes
people of Chinese (74.2%), Malay (13.3%), Indian (9.2%) and other (3.3%) descent (Central
Intelligence Agency, 2016). Although Singapore is not as ethnically homogenous as other Asian
countries (e.g., Japan, China, Korea), it has strong roots in Confucianism (e.g., Leung, 2002; Li,
2011; Tan & Chee, 2005; Tan & Yates, 2011). Singapore was not chosen to be the context for
this study to represent Asia, but as a cultural context that is distinct from Western societies like
the United States.
Although the national language in Singapore is Malay, English is the main language of
communication in commerce, and it is also the language of instruction in all schools (Ministry of
Information, Communications, and the Arts, 2009). That English is the language of instruction is
an important criterion in this study, as it allows for the development of a scale in English, which
can also be used later in cross-cultural studies (e.g., comparison with students in the United
States) without the additional concern of losing meaning in translating scales.
A school year in Singapore begins in January and ends in November (Ministry of
Education, 2012). Children in Singapore are mandated to attend six years of primary school
education. This is often followed by four or five years of secondary school education, and two to
three years of pre-college education. Prior to entering secondary schools, students have to take
the Primary School Leaving Examination (PSLE), which will place students into schools of
different rankings, as well as track students into three tracks based on their achievement: Express,
Normal Academic, and Normal Technical. Typically, a top-rank school will have students in the
Express track and the mid-rank and low-rank schools will have students in the Express, Normal
Academic, and Normal Technical tracks.
Students in the Express track complete their secondary school curriculum in four years
and they take a nation-wide secondary school leaving exam (i.e., GCE O Level) at the end of
four years. Students in the Normal Academic track complete the same curriculum as their peers
in the Express tracks but they finish their secondary school education in five years; and students
in the Normal Technical track complete a different secondary curriculum with a vocational focus
in four years.
Research on academic self-efficacy in Singapore. Although self-efficacy beliefs have
been studied extensively in Western contexts, it is a very new field of interest in Singapore. A
search on PsychINFO database (conducted February 15, 2016) with the keywords self-efficacy
and Singapore showed that 136 peer-reviewed journal articles have been published to date, and
84 of them were published in the last five years. Of the 136 peer-reviewed journal articles,
 11

authors of 29 articles reported on the self-efficacy of students and the authors of the other articles
reported on adults’ self-efficacy in various settings. Of those 29 articles on students’ self-
efficacy, only nine of them were on academic self-efficacy. Four of these nine articles were
excluded from this review because two of them looked how self-efficacy was developed or
looked at self-efficacy as a moderator between perceive task-importance and anxiety (Chen &
Looi, 2011; Nie & Lau, 2011), and two of the articles also did not examine academic self-
efficacy in relation to academic outcomes (Nie, Lau, & Liau, 2011; Yeung, Lau, & Nie, 2011).
Given that self-efficacy is a domain- and context-specific construct (Bandura, 1997), in this
section, I will review the remaining five studies on academic self-efficacy in relation to academic
outcomes in the context of Singapore.
Study 1. Chong (2007) conducted a study to investigate the usefulness and relevance of
Western conceived ideas of self-efficacy and several domains of self-concept (e.g., moral,
intellectual, family, social, and general) in understanding academic self-regulation in the
Singapore context. Specifically, the researcher wanted to examine (a) the relationships among
academic self-efficacy, self-concept, fear or failure, and self-regulation; and (b) how well self-
efficacy, self-concept, and fear of failure will predict self-regulation. The participants in this
study were 1,304 seventh-grade students from seven middle schools in Singapore. Three self-
report questionnaires (academic self-regulation and academic efficacy, self-concept, and fear of
failure) were administered to the students. Academic self-regulation was operationalized as the
degree to which students apply various cognitive, metacognitive, and self-regulatory learning
strategies in doing their schoolwork. Academic efficacy was operationalized as “students’
judgments of their capability in organizing and exercising control over their performance, that is,
efficacy in schoolwork (Chong, 2007, p. 68).” The self-efficacy items in this questionnaire were
adapted from the Motivated Strategies for Learning Questionnaires (MSLQ – Secondary,
Pintrich & DeGroot, 1990; MSLQ – Post-secondary, Pintrich, Smith, Garcia, & McKeachie,
1991). According to Chong, the self-efficacy items were “not subject-specified and more
generalized than most typical scales (Chong, 2007, p. 68).” The internal consistency estimate for
the self-efficacy scores was .73, but the authors did not examine the structural validity of the
scores in the sample.
Consistent with previous research, Chong (2007) found that academic efficacy was
positively related to academic self-regulation (r = .56, p < .01). Chong also found that academic
efficacy was a stronger predictor of academic self-regulation (β = .31, p < .01) than previous
grade (β = .01, p > .05) and various domains of self-concept (βmoral = .20, p < .01; βfamily = .13, p
< .01; βsocial = .10, p < .01; βintellectual = .09, p < .01; βgeneral = .24, p > .05). The findings of this
study provide support for the relevance and usefulness of self-efficacy in predicting Singapore
students’ self-regulated behavior. Since research has shown that self-regulated learners tend to
do better in school than learners who do not self-regulate (e.g., Schunk & Zimmerman, 1994),
one can infer from the findings that the positive relationship between academic self-efficacy and
academic achievement may also be present in the Singapore context.
Study 2. In another study of academic self-efficacy in Singapore, Lau, Liem, and Nie
(2008) examined how self-efficacy beliefs, task value, and achievement goals predict deep
learning in Mathematics. The researchers hypothesized that students with higher levels of self-
efficacy would be more likely to pursue performance-approach goals (i.e., demonstrate that they
are more competent than their classmates) and less likely to pursue performance-avoidance goals
(i.e., avoid situations for fear of appearing less competent). In addition, the researchers
hypothesized that achievement goals (i.e., task-approach, task-avoidance, performance-approach,
 12

and performance-avoidance) would mediate the relationship between self-efficacy and class
attentiveness and participation, which would in turn predict deeper learning in Mathematics.
The participants in this study were 1,476 ninth-grade students from 39 middle schools in
Singapore. An online self-report questionnaire was administered to the students and the
questionnaire included items in the following domains: task value, self-efficacy, achievement
goals, classroom attentiveness, group participation, and deep learning (i.e., critical thinking,
elaboration, and organization). Self-efficacy was operationalized as students’ confidence in
mastering the skills taught in their Mathematics class, and the five items were taken from the
MSLQ – Post-secondary (Pintrich et al., 1991). The internal consistency estimate for the self-
efficacy score was .87. The structural validity of the self-efficacy scale was examined in Lau’s
et al. (2008) study by confirmatory factor analysis. Results suggest that the five self-efficacy
items loaded onto one factor, and the factor loadings ranged from .67 to .80.
Further, results of this study showed that self-efficacy was a significant predictor of task-
approach goals (β = .19, p < .001), performance-approach goals (β = .30, p < .001), and
performance-avoidance goals (β = - .09, p < .01). Self-efficacy also had an indirect effect on
deep learning in mathematics (β = .16, p < .001), as mediated by the achievement goals, class
attentiveness, and group participation. These findings suggest that by increasing students’ self-
efficacy beliefs, their achievement goals would also be influenced, which could in turn lead to
deep learning in mathematics.
Study 3. In another article published in the same year by the preceding three authors,
Liem, Lau, and Nie (2008), reported on their study which looked at how self-efficacy beliefs,
task value, and achievement goals predict students’ learning strategies, task disengagement, peer
relationship, and English achievement outcome. The researchers hypothesized that students’
self-efficacy beliefs would positively predict the adoption of mastery and performance-approach
achievement goals and negatively predict the adoption of performance-avoidance achievement
goal. In addition, the researchers hypothesized that students’ self-efficacy beliefs would be
positively related to the use of deep learning strategies (e.g., elaborating ideas, thinking critically,
integrating one concept with another), which would in turn predict English achievement
outcomes.
The participants in this study were 1,475 ninth-grade students from 39 middle schools in
Singapore. A two-part survey was administered to the students; part one was made up of items
that asked for participants’ demographics information and part two was made up of sets of items
in the following domains: task value, self-efficacy, achievement goals, surface and deep learning,
task disengagement, and perception of relationships with peers. Self-efficacy was
operationalized as students’ confidence in mastering the skills taught in their English class, and
the five items were taken from the MSLQ – Post-secondary (Pintrich et al., 1993). The internal
consistency estimate for the self-efficacy score was .84. The structural validity of the self-
efficacy scale was also examined by confirmatory factor analysis. Results indicated that the five
self-efficacy items loaded onto one factor, and the factor loadings ranged from .56 to .76.
Further, results of this study showed that self-efficacy was a significant predictor of
mastery goals (β = .15, p < .01), performance-approach goals (β = .31, p < .01), and
performance-avoidance goals (β = - .17, p < .01). Self-efficacy also had a direct effect on deep
learning in English (β = .22, p < .01), which in turn was a significant predictor of English
achievement outcome (β = .04, p < .01). Self-efficacy did not have a significant, direct effect on
achievement. These findings suggest that students’ self-efficacy beliefs play an important role in
 13

influencing the students’ adoption of achievement goals and learning strategies, which in turn
predict their achievement outcomes.
Study 4. In a study on students’ achievement goal, Luo, Paris, Hogan and Luo (2011) had
two consecutive research agendas: first, by adopting the multiple goals perspective (e.g., mastery,
performance approach, and performance avoidance), they wanted to examine if these goals
combine in different ways to produce different clusters amongst ninth-grades students in
Singapore. Second, after obtaining the clusters, the researchers analyzed how each of these
clusters related to 11 cognitive, behavioral, and affective learning outcomes.
The participants in this study were 1,697 ninth-grade students from 39 schools in
Singapore. First, a survey was administered to the students and it contained items on
achievement goals, self-efficacy, subjective task values, learning activities (five scales:
classroom engagement, homework engagement, time management, metacognitive self-regulation,
and effort regulation), and affective outcomes (three scales: anxiety, positive affect, and negative
affect). Self-efficacy was operationalized as students’ beliefs about their confidence in
mastering the skills taught in their Mathematics class. The five items on self-efficacy were taken
from the MSLQ – Post-secondary (Pintrich et al., 1993). The internal consistency estimate for
the self-efficacy scores was .84, but the researchers did not examine the structural validity of the
scores in the sample. A month after the survey was administered, a Mathematics achievement
test was administered.
Luo et al. (2011) found that self-efficacy had the highest correlation with current
Mathematics score (r = .19, p < .001). For reference, the three highest correlations with current
Mathematics score were self-efficacy and performance avoidance goals (r = -.19, p < .001), test
anxiety (r = -.15, p < .001), and class engagement (r = .12, p < .001). The researchers also
found that cluster (η2 = .08, p < .001) had a larger effect size than past achievement (η2 = .02, p
< .001) on students self-efficacy. Of the four clusters, students in the Approach group (i.e., high
mastery and performance approach, low performance avoidance) reported the highest level of
self-efficacy, followed by students in the Success Oriented (i.e., moderate mastery, high
performance approach and performance avoidance) group, followed by students in the Diffuse
and Moderate Mastery groups (no significant difference between these two groups). The
interesting finding though is that students with the highest self-efficacy scores (i.e., Approach
group) had the highest current Mathematics score of the four groups but students with the second
highest self-efficacy scores (i.e., Success Oriented group) had the lowest current Mathematics
score of the four groups. Although this study seems to suggest that as a single variable, self-
efficacy has the highest correlation with current Mathematics score, the strength of the
relationship between self-efficacy and current score was moderated by the students’ achievement
orientation.
Study 5. In the most recently published study on academic self-efficacy in Singapore,
Stankov, Lee, Luo, and Hogan (2012) looked at how four constructs – confidence, self-efficacy,
self-concept, and anxiety – compared in their prediction of achievement outcomes in
Mathematics and English. Confidence was operationalized as how certain (i.e., 0% to 100%) a
person thinks that his or her answer was correct. This differs from self-efficacy, which was
operationalized in this study as a person’s beliefs about his or her capability to produce outcomes.
The authors pointed out that the main conceptual difference between confidence and self-
efficacy is that self-efficacy is domain specific but confidence, although closely tied to individual
items, is much broader in scope. The researchers attempted to address three issues in their paper:
(a) the strength and broadness of confidence trait; (b) relationship between confidence and
 14

accuracy scores; and (c) prediction of accuracy from confidence and self-belief measures. Of
particular relevance to this study is the third issue.
There were two groups of participants in this study: one that took the Mathematics survey
(N = 1,940) and one that took the English survey (N = 1,786). These ninth-grade students were
drawn from 31 middle schools in Singapore. A 200-question survey was administered online to
the students and this survey comprised of questions in the following areas: self-efficacy, anxiety,
subject-specific self-concept, academic self-concept, memory self-concept, and reasoning self-
concept. There were five items measuring Mathematics self-efficacy (α = .91) and five items
measuring English self-efficacy (α = .90). It was not clear if the researchers came up with these
items themselves or if they were adapted from another scale. The researchers also did not
examine the structural validity of the scores in the sample. After completing the survey, students
also had to take either a Mathematics or an English achievement test in a separate session and
after each item on the test, they had to rate their level of confidence that they had answered
correctly, on a 11-point scale of 0% to 100%.
The researchers found that confidence had high correlations with achievement and
measures of self-beliefs and confidence was a better predictor of achievement (βMathematics = .65, p
< .001; βEnglish = .56, p < .001), than self-efficacy (βMathematics = .05, p > .05; βEnglish = -.00,
p > .05), self-concept (βMathematics = .03, p > .05; βEnglish = -.10, p > .05), or anxiety (βMathematics = -
.11, p < .001; βEnglish = -.19, p < .001). The researchers acknowledged that the standardized
regression weight between confidence and achievement, especially that for Mathematics, was
higher than usual but based on these results, they concluded that confidence is an important
construct that is a better predictor of achievement than domain-specific construct. This was the
first study that compared the predictive ability of confidence and self-beliefs, so more follow-up
studies have to be done.
Summary of the studies conducted in Singapore. These five studies examined a myriad
of variables alongside self-efficacy and they yielded mixed results. For example, Chong (2007)
and Lau, Liem, and Nie (2008) chose self-regulation strategies and deep learning as their
dependent variables but the other three studies examined self-efficacy’s predictive ability for
achievement outcomes. Findings of the first four studies reviewed here are consistent with
findings from research conducted in a Western context, that is, self-efficacy is a significant
predictor of academic and achievement outcomes, but the last study by Stankov et al. (2012)
showed that self-efficacy was not a significant predictor of achievement. These different
research variables and mixed outcomes lead to an inconclusive outcome and more studies on
academic self-efficacy in Singapore are needed.
One area of future research, which is also a limitation of the studies reviewed here, is
how self-efficacy has been measured. Interestingly, four of the five studies used the MSLQ and
the fifth study did not specify how or where the self-efficacy questions came from. The MSLQ
has been in use since 1986 (Pintrich, Smith, Garcia, & McKeachie, 1993) and researchers have
consistently found high internal consistency estimates for the self-efficacy score of the MSLQ (α
= .89, Pintrich & DeGroot, 1990; α = .93, Printrich et al., 1993). Researchers have also
consistently found significant correlations between the self-efficacy score and final school grades
(r = .36, Pintrich & DeGroot, 1990; r = .41, Printrich et al., 1993). However, most of the
validation studies of MSLQ scores were conducted with a predominantly White sample in
Michigan (e.g., Pintrich & DeGroot, 1990, Pintrich et al., 1993). The internal consistency
estimates of the four studies in Singapore were lower than in the validation studies done by
Pintrich and his colleagues, ranging from .73 to .87. Also, only two of the four studies validated
 15

the MSLQ self-efficacy scores in a Singapore sample. Thus, the psychometric properties of
MSLQ scores have not yet been sufficiently established for use in Singapore.
A third limitation of these studies is how self-efficacy was operationalized. To recap,
Bandura (1977) defined self-efficacy as one’s perceived capabilities to organize and execute
courses of action at a specific level of difficulty, in order to attain one’s goals in a specific
context. Upon closer examination of the MSLQ, the self-efficacy items are not all about a
person’s perceived abilities to organize and execute actions; rather they contain elements of
social comparison (e.g., “Compared with other students in this class I expect to do well;” “My
study skills are excellent compared with other students in this class;” and “Compared with others
in this class, I think I am a good student” [Pintrich & DeGroot, 1990]). These make up three of
the five self-efficacy questions used in four of the studies. The deviation from Bandura’s (1977)
definition of self-efficacy makes it unclear if the MSLQ is an appropriate measure of academic
self-efficacy.
Pilot study for development and validation of a Mathematics self-efficacy scale. In
2008, I examined the reliability and validity of scores on a newly developed English and
Mathematics self-efficacy scale among 279 secondary school students in Singapore. The 28-
item survey included 10 items on self-efficacy (five on each subject), eight items on attitude
towards school and each of the subjects, and eight items on demographics. The self-efficacy
items were all general (e.g., I can do 100% of my Mathematics homework on my own) and not
specific to the curriculum. In addition, the items on attitude towards school, attitude toward
English, and attitude toward Mathematics included questions that asked students to rate
themselves compared to their peers.
Results indicated that scale scores had moderate to strong internal consistency (αMath self-
efficacy = .83; αEnglish self-efficacy = .79). Exploratory factor analysis suggested that two factors should
be extracted and Factor 1 consisted of the five items on Mathematics self-efficacy and Factor 2
consisted of the five items on English self-efficacy. The English and Mathematics items loaded
cleanly on their home factors with no cross-loading greater than .21. Finally, two hierarchical
multiple regressions were performed, one for each subject. In both regressions, self-efficacy
scores accounted for significant amount of variance in the respective subject grades.
These results suggest that the nature and structure of self-efficacy beliefs in the Singapore
context is similar to that in the Western contexts. However, this was a pilot study and more
research had to be done to clarify the nature and structure of self-efficacy in a non-Western
context. Further, the self-efficacy questions in this study, being general questions about students’
perceive abilities (e.g., I can get an “A” in the next English class test), did not reflect the context-
specific nature of self-efficacy. Further studies are needed to examine the relationships of degree
of specificity and its predictive ability of final grades. Lastly, another limitation of this pilot
study was that all participants came from one grade level in an all-girls secondary school. The
sample was not representative and research has been inconclusive on the relationship between
gender and self-efficacy.
Relationship Between Gender and Self-Efficacy
Yeung, Lau, and Nie (2011) investigated the relationships between self-efficacy and five
other motivation constructs in a sample of 4,214 students in 5th and 9th grade in Singapore. The
researchers examined how these constructs related with grade and gender and they found that the
difference between genders was greater at 5th grade than at 9th grade. Overall, boys tended to
have lower scores on the motivation constructs, although the difference between genders was
small. In a U.S. sample, Hampton and Mason (2003) looked at the relationship among several
 16

variables: gender, learning disability status, sources of efficacy, self-efficacy beliefs, and
academic achievement. The researchers found that gender was the only variable that did not
have an influence of self-efficacy beliefs. Busch (1995) examined gender differences in
perceived self-efficacy and academic achievement in six business-related courses among a
sample of 154 Norwegian college students. The researcher found that female students had
significantly higher self-efficacy in statistics but lower self-efficacy in computing and marketing.
A meta-analysis of 187 studies on gender differences in academic self-efficacy found that
males had higher self-efficacy but the effect size was small at .08 (Huang, 2013). Similar to the
findings of Busch (1995), Huang also found that the relationship between gender and academic-
self-efficacy is subject-specific. Males had higher self-efficacy scores in the areas of
mathematics, computer science, and social sciences, and females had higher language arts self-
efficacy scores. It was also noted that differences in mathematics self-efficacy only occurred in
late adolescence. Overall, the findings on relationship between gender and self-efficacy beliefs
have been mixed. There are findings that favor males and also findings that favor females, but
most of these findings though significant have small effect sizes.
The Current Study
This study aims to address the limitations of academic self-efficacy studies done in
Singapore, by first developing and validating a mathematics self-efficacy scale. In order to stay
true to Bandura’s original conceptualization of self-efficacy, this mathematics self-efficacy scale
was designed according to Bandura’s (2006) guide for constructing a self-efficacy scale.
Bandura’s guide for constructing a self-efficacy scale. Bandura (2006) listed out
several considerations in constructing a self-efficacy scale. First, to ensure the items stay true to
the original definitions of self-efficacy, items should measure a person’s perceived capabilities to
perform certain actions. Therefore, items should be phrased in terms of what a person thinks he
can do rather than how a person feels, will do, or has done in the past. Second, self-efficacy is a
domain-specific concept. There are many facets to a given domain and a comprehensive self-
efficacy scale should cover the range of behavioral factors related to the domain. A scale to
measure self-efficacy for mathematics should include many topics and related skills. Third,
Bandura pointed out that self-efficacy exists on a continuum. As such, items should represent
gradations of challenges and should be pegged to different levels of task difficulties. Fourth,
Bandura cautioned against response bias. He suggested that items should reduce social
evaluative concerns and should focus on how a person perceives his or her own capabilities.
Bandura (2006) also suggested constructing a scale with more response steps in order to
yield more reliable scores. He cited the study by Pajares, Hartley, and Valiante (2001), which
compared the reliability measures of two writing self-efficacy scales: one of them had a response
scale from a 0-100 and the other had a response scale of 1-6. Pajares et al. (2001) found that the
efficacy scale with the 0-100 response scale (β = .449, p < .001) was a stronger predictor than the
scale with five intervals (β = -.024, p = .726). To examine the psychometric properties of self-
efficacy scales, Bandura (2006) has suggested using Cronbach’s alpha and factor analysis to
examine the reliability and validity of the scores on self-efficacy scores.
The current scale. The current scale (see Appendix C) was developed in close adherence
to Bandura’s (2006) guidelines, and the cultural and educational context of Singapore. One of
the features of this scale is the emphasis on participants’ effort in academic pursuits (see
Appendix C; e.g., Q4, “I can put in greater efforts to get better results in Math;” Q23, “When I
do badly in a class test, I can learn from my mistakes and try harder in the next test;”) rather than
innate academic abilities (e.g., “I am good at Math”). A second feature of the self-efficacy scale
 17

is the inclusion of explicit goals in the questions (see Appendix C; e.g., Q5, “I can calculate the
gradient of a linear graph;” Q17, “I can get an ‘A’ in the next Mathematics class test;” Q22, “I
can list two properties of similar polygons,”) as well as generic statements (see Appendix C; e.g.,
Q1, “I can understand new Mathematics topics when taught;” Q19, “I can pass the next
Mathematics class test”). Having both questions with explicit goals and questions with general
goals allow for comparison, to see which level of specificity in self-efficacy best accounts for
variances in Mathematics grades. Students’ responses will give us an insight into the domain
specificity nature of self-efficacy.
A third feature addresses gradations of challenges in self-efficacy. The questions were
designed in consultation with the secondary education curriculum in Singapore, and the scope of
the questions will draw from Secondary 2 to 4 (equivalent to Grades 8 to 10) curricula. Lastly,
elements of social comparison was avoided so as to minimize the likelihood that participants
would mark down their responses in order to downplay their own abilities to elevate that of
others. There are therefore no items that required students to evaluate their abilities in relation to
other students. All items are phrased in terms of what a student things he or she can do.
Hypotheses. To validate this new self-efficacy scale, I had three research goals. First I
examined the psychometric properties of the scale, including the internal consistency and
structural validity of scores. Second, I examined the convergent validity of the mathematics self-
efficacy scores by correlating the scores with other measures (e.g., school climate, Mathematics
attitude). Third, I examined the relative contribution of mathematics self-efficacy to student
achievement by simultaneous and hierarchical regression methods.
Internal consistency and structural validity. I had two hypotheses regarding the internal
consistency and structural validity of the scores. First, I hypothesized that scores on the
mathematics self-efficacy scale would have acceptable reliability estimates at or above .70, as I
was collecting data from adolescents, who generally have more stable self-perceptions. Second,
I performed exploratory and confirmatory factor analyses to determine the factor structure of the
mathematics self-efficacy scale. I hypothesized that the scores would yield a one-factor (i.e.,
mathematics self-efficacy) structure and a two-factor (i.e., general mathematics self-efficacy and
content-specific mathematics self-efficacy) correlated structure, both with acceptable fit indices.
However, I hypothesized that the two-factor correlated structure would have better fit indices
than the one-factor structure because there were distinct differences in level of specificity
between general mathematics self-efficacy and content-specific mathematics self-efficacy.
Relationship of self-efficacy scores with other measures. I had six hypotheses regarding
the relationship of self-efficacy scores with other measures. First, I hypothesized that school
climate and mathematics attitude would both be positively correlated with mathematics self-
efficacy. Second, I further hypothesized that mathematics self-efficacy would have a stronger
correlation with mathematics attitude than with school climate. Third, I hypothesized that
Secondary 3 (equivalent to Grade 9, around 15 years old) overall percentage and Secondary 3
mathematics grade would both be positively correlated with mathematics self-efficacy. Fourth, I
further hypothesized that the correlation between Secondary 3 mathematics grade and
mathematics self-efficacy would be stronger than the correlation between Secondary 3 overall
percentage and mathematics self-efficacy.
Fifth, I also included the MSLQ – Post-secondary self-efficacy items (Appendix D) in
this study since it has been used in many studies of self-efficacy in Singapore but its
psychometric properties have not been sufficiently established in Singapore. If the MSLQ scores
have acceptable reliability, I will examine its correlation to the mathematics self-efficacy scale I
 18

am developing. I hypothesized that there will be strong correlation between the MSLQ – Post-
secondary self-efficacy items and the mathematics self-efficacy scale I am developing since the
two scales are were designed to measure the same construct. Lastly, I hypothesized that males
will have a higher level of self-efficacy than females. I predict that though significant, the effect
size for the gender difference will be small.
Relative contribution of mathematics self-efficacy to student achievement. Depending
on the outcome of the factor analysis, mathematics self-efficacy may be operationalized as a
single independent variable – the sum score of all the mathematics self-efficacy items (if one
factor structure has best fit) – or mathematics self-efficacy may be operationalized as two
independent variables – a general mathematics self-efficacy score and content-specific
mathematics self-efficacy score (if two factor structure has best fit). Following these two
possible outcomes, I had two sets of hypotheses regarding the relative contribution of
mathematics self-efficacy to student achievement.
One-factor structure. If mathematics self-efficacy was found to be a single independent
variables, I hypothesized that mathematics self-efficacy will account for a significant amount of
variance in students Secondary 3 mathematics grade, and this contribution will be greater than
the contributions of other attitudinal constructs (school climate and mathematics attitude).
Second, I hypothesized that after accounting for students’ past achievement (i.e., PSLE
mathematics grade and T-score), mathematics self-efficacy will still account for a significant
amount of variance in students Secondary 3 mathematics grade.
Two-factor structure. If mathematics self-efficacy was found to be two independent
variables, I first hypothesized that content-specific mathematics self-efficacy score would
account for a significant and greater amount of variance in students’ Secondary 3 mathematics
grade, compared to general mathematics self-efficacy. Next, I hypothesized that either
Mathematics self-efficacy variable would account for a greater amount of variance in students’
mathematics grade, compared to other attitudinal constructs (school climate and mathematics
attitude). Lastly, I hypothesized that either Mathematics self-efficacy variable would account for
a greater amount of variance in students’ mathematics grade, compared to past achievement
(PSLE mathematics grade, PSLE overall score).
Method
Participants
There were a total of 1,587 participants in this study, ranging in age from 12 to 18 years
(M = 14.9, SD = .54). 52.4% (n = 823) of the participants are female. The participants were
drawn from seven secondary schools in Singapore and they were all in Secondary 3 (equivalent
to 9th grade). The students represent three academic tracks: 62.7% were from the Express track
(n = 995), 28.8% were from the Normal Academic track (n = 457), and 8.1% were from the
Normal Technical track (n = 128). This distribution is representative of the overall distribution
of students in Singapore.
Procedure
Approval to conduct this study was granted by the Committee for Protection of Human
Subjects in April 2015 (CHPS Protocol Number: 2014-10-6763). Following this, approval to
collect data in Singapore was sought from the Ministry of Education, Singapore and it was
granted in June 2015 (Request Number: RQ72-15(06)). Finally, consent was sought from
principals of multiple schools to participate in this study and seven principals agreed to
participate.
 19

In October 2015, the surveys were brought to Singapore and administered to the
Secondary 3 students in these seven schools. In some of the schools, students were gathered in a
large hall where the primary investigator gave the instructions prior to administering the surveys.
In other schools, written instructions were given to the teachers who read them to the students
prior to administering the survey. The students took approximately 25 minutes to complete the
survey. At the end of the survey, students were given opportunities to share their feedback or
ask questions.
Measures
Five surveys were included in this study (see Appendixes A, B, C, D, and E): school
climate (Appendix A); mathematics attitude (Appendix B); mathematics self-efficacy (Appendix
C); self-efficacy items from the MSLQ (Appendix D); and demographics (Appendix E). The
order of the attitudinal surveys (Appendixes A to D) was randomized for each school; the
demographics section was always the last section.
School climate. School climate was assessed using 10 Likert-type items measuring
students’ attitude towards school, including perceptions of teachers and friends (Items 1, 2, 7,
and 8), sense of belonging (Items 3, 5, and 9), and importance of school (Items 4, 6, and 10).
These items were drawn from the Trends in International Mathematics and Science Study 2011 –
Grade 8 Student Questionnaire (TIMSS 2011; International Association for the Evaluation of
Educational Achievement, 2011) and the Program for International Student Assessment 2012 –
Student Questionnaire Form B (PISA 2012; Organisation for Economic Co-operation and
Development, 2012). Six of the 10 school climate items were positively worded and four were
negatively worded. Two examples of school climate items are, “Most of the teachers in this
school are interested in students’ well being,” and “School has been a waste of time.” Each item
had six possible response options, ranging from 0% Agree to 100% Agree.
Mathematics attitude. The Mathematics attitude scale included 10 Likert items
measuring students’ attitude towards mathematics, including emotions associated with
mathematics (Items 3, 5, and 8), importance of mathematics (Items 2 and 7), effort input to
learning mathematics, (Items 4, 6, and 9), and enjoyment of mathematics (Items 1 and 10).
These items were drawn the TIMSS 2011 (International Association for the Evaluation of
Educational Achievement, 2011) and PISA 2012 (Organisation for Economic Co-operation and
Development, 2012). Among the 10 mathematics attitude items, six were positively worded and
four were negatively worded. Two examples of mathematics attitude items are, “Mathematics is
one of my favorite subjects,” and “I do not understand why I have to study mathematics.” Each
item had six possible response options, ranging from 0% Agree to 100% Agree.
Mathematics self-efficacy. For comparison purposes, Mathematics self-efficacy was
assessed using two scales: the MSLQ – Post Secondary self-efficacy items (Pintrich et al., 1991)
that has been used in many Singapore studies but has not been validated, and the 30-item
Mathematics self-efficacy scale developed for this study. There were eight self-efficacy items on
the MSLQ and the items assess students’ attitude towards a subject (e.g., “I’m confident I can
understand the most complex materials presented by the instructor in this course.”). The new
self-efficacy scale that was developed for this study included 10 items on general mathematics
self-efficacy (Appendix C; Items 1, 4, 7, 10, 11, 12, 17, 19, 23, and 29) and 20 content-specific
mathematics self-efficacy items (Appendix C; Items 2, 3, 5, 6, 8, 9, 13, 14, 15, 16, 18, 20, 21, 22,
24, 25, 26, 27, 28, and 30). Some of these items were from the pilot study and the other items
were newly formulated for this scale. The items are all worded in neutral terms in the form of “I
can …” An example of a general mathematics self-efficacy item is, “I can pass the next
 20

mathematics class test.” An example of a content-specific mathematics self-efficacy item is, “I

can calculate the gradient of a linear graph.” Each item had six possible response options,
ranging from 0% Agree to 100% Agree.
Although I had earlier reviewed that Bandura had suggested creating scales with more
response steps (e.g., 0-100), I had deviated from his guidelines by only including six response
steps. One of the reasons for this was that most scales in the literature continue to use less than
10 response steps. Pajares, one of the authors who published the 2001 (Pajares, Hartley, &
Valiente, 2001) article supporting the guidelines of including more steps, later created and used
self-efficacy scales that only included six response options (Usher & Pajares, 2008). Another
reason for deviating from Bandura’s guidelines was that research have shown that the
distribution of responses on five- to seven-point scales mirror the distribution of responses on
100-point scales; there is currently no strong evidence to support Bandura’s claim for including
more response steps (Norman, 2010).
Demographics. Participants were asked to answer questions on their gender, age,
academic track they were in, PSLE T-Score (theoretical range of 0 to 300), PSLE mathematics
grade (A*, A, B, C, or D), and mathematics subjects they were taking in Secondary 3
(Elementary Mathematics only or Elementary Mathematics and Additional Mathematics).
Participants were also be asked to report on their overall percentage in their Secondary 3 year-
end examinations, which is the mean result across all subjects. The overall percentage was used
to rank students within each class as well as within each level. Participants were also asked to
report their mathematics grade in the same year-end exam (possible grades are A1, A2, B3, B4,
C5, C6, D7, E8, or F9). This method of self-reporting grades was chosen to preserve anonymity
of students’ responses, thus reducing the influence of social desirability on responses. Research
(Crockett, Schulenberg, & Petersen, 1987) has also shown that there is a 0.76 correlation
between self-reported grades and actual grades and this high correlation makes self-reported
grades a reliable form of data collection.
Results
Preliminary Analyses
IBM SPSS Statistics (Version 23.0) was used for all analyses, except confirmatory factor
analyses (CFAs); CFAs were run using MPlus 7 (Muthén & Muthén, 1998–2012). There were
fewer than 5% missing values in all the variables. Little’s Missing Completely At Random
(MCAR) test was conducted to determine whether missing values were missing at random. All
of the independent variables (i.e., school climate, mathematics attitude, mathematics self-
efficacy, MSLQ) had missing values that were random, so expectation maximum likelihood was
used to replace the missing values. Means and standard deviations were then calculated for the
independent variables and are presented in Table 1.
The same method was also used to replace missing values for Elementary Mathematics
grade, overall percentage, PSLE Math grade, and PSLE T-score. Participants reported their
Elementary Mathematics grades on a nine-point scale ranging from A1 (highest possible score)
to F9 (lowest possible score). The mean was 5.73 (SD = 2.54). Overall percentage was reported
in seven possible ranges: less than 41%, 41-49%, 50-59%, 60-69%, 70-79%, 80-89%, and 90-
100%. Most students reported their overall percentage in the 60-69% range (n = 464), followed
by the 50-59% range (n = 417), followed by 70-79% range (n = 260); this mirrors a normal
distribution. Participants reported that PSLE Mathematics grades on a five-point scale ranging
from A-star (highest possible) to D. The mean was 3.11 (SD = 1.08). PSLE T-score was a free
response item. Participants’ responses ranged from 90 to 260 (M = 202, SD = 28.4).
 21

Correlations were calculated for the following 10 variables: school climate mean,
mathematics attitude mean, mathematics self-efficacy combined mean, mathematics self-efficacy
general mean, mathematics self-efficacy content-specific mean, MSLQ mean, Elementary
mathematics grade, overall percentage, PSLE mathematics grade, and PSLE T-score (Table 2).
Of the 45 correlations, 43 were significant at the .01 level. Of these 43 correlations, 30 attained
values of at least .30, accounting for at least 9% of shared variance. The directions and relative
magnitudes of these correlations also made theoretical sense. For example, Mathematics self-
efficacy combined score had a larger correlation with Elementary mathematics grade (r = .57)
than with overall percentage (r = .47).
Major Analyses
Internal consistency and structural validity. Internal consistency estimates based on
Cronbach’s alpha were calculated for school climate, mathematics attitude, mathematics self-
efficacy, and MSLQ scores. The reliability estimates were all acceptable, ranging from .76
to .96 (Table 3). Next, to examine the structural validity of the scales, approximately 25% (n =
414) of the sample was selected at random for exploratory factor analyses. The remaining 75%
of the cases (n = 1,173) were reserved for confirmatory factor analyses.
Exploratory factor analyses. Exploratory factor analyses (principal axis extraction) were
used to examine the factor structure of the school climate, mathematics attitude, mathematics
self-efficacy, and MSLQ scales. Factorability of the scores was based on the Kaiser-Meyer-
Olkin (KMO) Measure of Sampling Adequacy and Barlett’s Test of Sphericity. All scores
yielded acceptable statistics on both measures (Table 4), which suggested that the correlation
matrix of scores were factorable. Next, multiple criteria – theory and parallel analysis (Watkins,
2000) – were used to determine the number of factors to extract. Item salience was set at .40,
which indicates at least 16% shared variance (Tabachnick & Fidell, 2001).
School climate. Two-factor (based on parallel analysis) and one-factor (based on theory)
structures were explored. In the two-factor structure, seven items had coefficients greater
than .40 (except items 7, 8, 10); five items loaded on Factor 1 and two items loaded on Factor 2.
The first factor accounted for 37% of variance in school climate scores, and the second factor
accounted for an additional 12% of variance in school climate scores. The correlation between
the two factors was .51. In the one-factor structure, eight of the ten school climate items had
coefficients greater than .40 (except items 6 and 7).
Three items with coefficients of less than .40 were excluded (i.e., items 7, 8, 10) and
further exploratory factor analyses – two-factor (based on parallel analysis) and one-factor
(based on theory) – were performed with seven items. In the two-factor structure, all seven items
had coefficients greater than .40; five items loaded on Factor 1 and two items loaded on Factor 2.
The first factor accounted for 42.2% of variance in school climate scores, and the second factor
accounted for an additional 17% of variance in school climate scores. The correlation between
the two factors was .42. In the one factor-structure, five of the seven items had coefficients
greater than .40 (except items 2 and 6).
Overall, the seven-item school climate scale accounted for a greater amount of variance
than the ten-item school climate scale. This suggests that subsequent confirmatory factor
analysis should begin with the seven-item scale. Further, although the two-factor structure
accounted for more variance than the one-factor structure, the medium to large correlation (r
= .42 to .51) between the two factors suggests that the two factors could be manifestation of the
same underlying variable, which supports theory’s recommendation that a one-factor structure is
most viable. An examination of the items that loaded onto each of the two factors also did not
 22

reveal any similarities within the factor or differences between the factors. High correlations
between factors, lack of discernible pattern of factor loadings, and theory all support that a one-
factor structure is most viable. As such, subsequent confirmatory factor analyses were
performed on a one-factor structure.
Mathematics attitude. Two-factor (based on parallel analysis) and one-factor (based on
theory) structures were explored. In the two-factor structure, all ten items had coefficients
greater than .40. Six items loaded on Factor 1 and four items loaded on Factor 2; one of the
items (i.e., Item 1) had coefficients of greater than .40 on both factors. The first factor accounted
for 40.8% of variance in mathematics attitude scores, and the second factor accounted for an
additional 16.4% of variance in mathematics attitude scores. The correlation between the two
factors was -.27. In the one-factor structure, eight of the 10 mathematics attitude items had
coefficients greater than .40 (except items 2 and 7).
Two items with coefficients of less than .40 were excluded (i.e., items 2 and 7) and
further exploratory factor analyses – two-factor (based on parallel analysis) and one-factor
(based on theory) – were performed with eight items. In the two-factor structure, seven of the
eight items had coefficients greater than .40; four items loaded on Factor 1 and three items
loaded on Factor 2. The first factor accounted for 48.6% of variance in mathematics attitude
scores, and the second factor accounted for an additional 17.2% of variance in mathematics
attitude scores. The correlation between the two factors was .41. In the one factor-structure, all
eight items had coefficients greater than .40.
Overall, the eight-item mathematics attitude scale accounted for a greater amount of
variance than the ten-item mathematics attitude scale. This suggests that subsequent
confirmatory factor analysis should begin with the eight-item scale. Further, although a two-
factor structure accounted for more variance than a one-factor structure, an examination of the
items that loaded onto each of the two factors did not reveal any similarities within the factor or
differences between the factors. Therefore, pattern of the factor loadings and theory both support
that a one-factor structure is most viable. As such, subsequent confirmatory factor analyses were
performed on a one-factor structure.
Mathematics self-efficacy. Three-factor (based on parallel analysis) and two-factor (based
on theory) structures were explored. In the three-factor structure, 28 of the 30 mathematics self-
efficacy items had coefficients greater than .40 (except items 22 and 29). Ten items loaded on
Factor 1, six items loaded on Factor 2, and 12 items loaded on Factor 3. Factor 1 accounted for
42.4% of variance in mathematics self-efficacy scores, Factor 2 accounted for 8.5% of variance
in mathematics self-efficacy scores, and Factor 3 accounted for 5.2% of variance in mathematics
self-efficacy scores. The correlation between Factor 1 and Factor 2 was .39, correlation between
Factor 1 and Factor 3 was -.68, and correlation between Factor 2 and Factor 3 was -.37. An
examination of how the items distributed across factors revealed that nine of the 10 items on
Factor 1 were general mathematics self-efficacy items. Factor 2 contained content-specific
mathematics self-efficacy items on the topics of probability (i.e., items 6, 13, and 28), statistics
(i.e., items 9 and 21), and scalar vector (i.e., item 27). Factor 3 contained the other content-
specific mathematics self-efficacy items.
In the two-factor structure, 27 of the 30 mathematics self-efficacy items had coefficients
greater than .40 (except items 10, 16, and 22). Twenty-one items loaded on Factor 1 and six
items loaded on Factor 2. Factor 1 housed a mix of general and content-specific mathematics
self-efficacy items, whereas Factor 2 had the same set of items from Factor 2 of the three-factor
structure (i.e., items 6, 9, 13, 21, 27, and 28). The correlation between the two factors was .53.
 23

During survey administration, students from all the schools shared with the principal
investigator that probability was a topic they had not learned yet. Given how items 21 27, and 28
consistently grouped with the three probability, the entire set of six items were excluded from
further factor analyses. Three-factor (based on parallel analysis) and two-factor (based on theory)
structures were explored with the remaining 24 items.
In the three-factor structure, 22 of the 24 mathematics self-efficacy items had coefficients
greater than .40 (except items 22 and 29). Eleven items loaded onto Factor 1, nine items loaded
onto Factor 2, and two items loaded onto Factor 3. The first factor accounted for 47.2% of total
mathematics self-efficacy scores, the second factor accounted for an additional 6.3%, and the
third factor accounted for 4.6%. The correlation between Factor 1 and Factor 2 was .73,
correlation between Factor 1 and Factor 3 was .30, and correlation between Factor 2 and Factor 3
was .21. An examination of how the items distributed across factors revealed that Factor 2
housed all the general mathematics self-efficacy items; Factor 3 housed two items on the topic of
statistics; and Factor 1 housed all the other content-specific mathematics self-efficacy items.
In the two-factor structure, 22 of the 24 mathematics self-efficacy items had coefficients
greater than .40 (except items 2 and 22). Twelve items loaded on Factor 1 and 10 items loaded
on Factor 2. Factor 1 housed content-specific mathematics self-efficacy items and Factor 2
housed general mathematics self-efficacy items. The correlation between the two factors was .76.
Overall, the twenty-four-item mathematics self-efficacy scale accounted for a greater
amount of variance than the thirty-item mathematics self-efficacy scale. This suggests that
subsequent confirmatory factor analysis should begin with the twenty-four-item scale. Pattern of
the factor loadings in the two-factor structure matched what theory would suggest, that two
distinct factors (i.e., general mathematics self-efficacy and content-specific mathematics self-
efficacy) exist. However given the high correlation between the two factors and the relatedness
between general mathematics self-efficacy and content-specific mathematics self-efficacy, it is
possible that a one-factor structure would also be a good fit. As such, subsequent confirmatory
factor analyses were performed on a one-factor and a two-factor structure.
Confirmatory factor analyses. Following the exploratory factor analyses, confirmatory
factor analyses (CFA) were performed on the remaining sample (n = 1, 173). Multiple criteria
were used to assess goodness of fit, and these include the Satorra-Bentler scaled chi-square, the
comparative fit index (CFI), the Tucker Lewis index (TLI) or non-normed fit index (NNFI), the
root mean square error of approximation (RMSEA), the 90% confidence interval for RMSEA,
and the standardized root mean square residual (SRMR). According to Hu and Bentler (1999),
an acceptable fit will be indicated by CFI and NNFI values at or greater than .95, a RMSEA
value at or less than .06, and a SRMR value at or less than .08. Other researchers (e.g., Byrne,
2006; Marsh, Hau, & Wen, 2004) have suggested a wider range of values for acceptable fit: CFI
and NNFI values at or greater than .92 and RMSEA and SRMR values smaller than .08.
School climate. A one-factor model comprised of seven school climate items (i.e., items 1,
2, 3, 4, 5, 6, 9) was examined but Item 2 did not have a substantial coefficient and was dropped.
A second one-factor model comprised of the six remaining school climate items was then
examined. CFA results are presented in Table 5. This model had CFI and TLI values greater
than .95, and RMSEA and SRMR values smaller than .06. The factor loadings were in the range
of .37 to .79. These indices suggest that the one-factor model was a good fit to the data.
Mathematics attitude. A two-factor structure comprised of eight mathematics attitude
items (i.e., items 1, 3, 4, 5, 6, 8, 9, 10) was examined but two of the four indices were not within
the acceptable range. A one-factor structure comprised of the same eight items was then
 24

examined but Items 3 and 6 did not have substantial coefficients and were dropped. Finally, a
one-factor model comprised of six mathematics attitude items (i.e., items 1, 4, 5, 8, 9, and 10)
was examined. CFA results are presented in Table 5. This model had CFI and TLI values
greater than .94 and the SRMR value was less than .04; the RMSEA value at .098 was the only
index outside of the acceptable range. The factor loadings were in the range of .56 to .88. These
indices suggest that the one-factor model was a fair fit to the data.
Mathematics self-efficacy. Of the 24 mathematics self-efficacy items examined, four
items (i.e., items 2, 3, 16, and 17) did not have substantial loadings were thus excluded from
further analyses. Next, a one-factor and a two-factor model comprised of 20 items (11 content-
specific mathematics self-efficacy items and nine general mathematics self-efficacy items) were
examined. CFA results are presented in Table 5. Two of the four indices were not within the
acceptable range: the CFI and TLI values less than .90. These indices suggest that the one-factor
model was not a good fit to the data.
A two-factor model comprised of 20 mathematics self-efficacy items was then examined.
The two factors were content-specific mathematics self-efficacy (i.e., items 5, 8, 14, 15, 18, 20,
22, 24, 25, 26, and 30) and general mathematics self-efficacy (i.e., items 1, 4, 7, 10, 11, 12, 19,
23, 29). CFA results are presented in Table 5. This model had CFI and TLI values greater
than .93, and the RMSEA and SRMR values were less than .06. The loadings on Factor 1
(content-specific mathematics self-efficacy) were in the range of .59 to .84 and the loadings of
Factor 2 (general mathematics self-efficacy) were in the range of .51 to .81. These indices
suggest that the two-factor model was a good fit to the data.
Convergent validity. A series of t-tests, correlations, and effect sizes were used to
examine the relationships among mathematics self-efficacy and various variables in this sample.
School climate, mathematics attitude, and mathematics self-efficacy. All three measures
of mathematics self-efficacy (combined, general, and content-specific) had significant positive
correlations with both school climate and mathematics attitude (Table 2). The effect sizes of the
correlations between school climate and mathematics self-efficacy ranged from small to medium
(.25 ≤ r ≤ .37) and the effect sizes of the correlations between mathematics attitude and
mathematics self-efficacy were all large (r > .48).
Elementary mathematics grade, overall percentage, and mathematics self-efficacy. All
three measures of mathematics self-efficacy (combined, general, and content-specific) had
significant positive correlations with both Elementary mathematics grade and overall percentage
(Table 2). All three measures of mathematics self-efficacy also had stronger correlations with
Elementary mathematics grade than overall percentage. The effect sizes of the correlations
between Elementary mathematics grade and mathematics self-efficacy were all large (r > .52)
and the effect sizes of the correlations between overall percentage and mathematics self-efficacy
ranged from medium to large (.43 ≤ r ≤ .49).
MSLQ and mathematics self-efficacy. The correlations between the self-efficacy items
from MSLQ and the three measures of mathematics self-efficacy (combined, general, and
content-specific) were all significant and positive (Table 2). The effect sizes were all large as
well (r > .59).
Gender, mathematics achievement, and mathematics self-efficacy. A first set of t-tests
was performed on gender and mathematics achievement. There was no significant difference
between the males and females in either past mathematics achievement or current mathematics
grade. A second set of t-tests was performed on gender and three mathematics self-efficacy
scores (combined, general, and content-specific). There was no significant effect for gender in
 25

combined mathematics self-efficacy and general mathematics self-efficacy. There was a

significant effect for gender in content-specific mathematics self-efficacy, favoring females (d
= .22). Results are presented in Table 6.
Concurrent validity. Elementary mathematics grade was entered as the dependent
variable and the two mathematics self-efficacy variables were entered as the independent
variables. Tests to see if the data met the assumption of collinearity indicated that
multicollinearity was not a concern (Content-specific mathematics self-efficacy, Tolerance = .43,
VIF = 2.33; General mathematics self-efficacy, Tolerance = .43, VIF = 2.33). Results indicated
that the two self-efficacy variables accounted for a significant amount of variance in Elementary
mathematics grade, Adjusted R2 = 33.1%, F(2, 1584) = 393.8, p < .001. Further, general
mathematics self-efficacy items accounted for more variance than content-specific mathematics
self-efficacy items (βContent = .24, p < .001; βGeneral = .38, p < .001).
Next, to examine the concurrent validity of the self-efficacy scores, three hierarchical
multiple regressions were performed to examine the relative contributions of mathematics self-
efficacy, other attitudinal constructs (i.e., school climate and mathematics attitude), and past
achievement (i.e., PSLE mathematics grade and PSLE overall score). In the first hierarchical
multiple regression, Elementary mathematics grade was entered as the dependent variable,
attitudinal constructs (i.e., school climate and mathematics attitude) were entered in Block 1, and
the two mathematics self-efficacy variables (i.e., general and content-specific) were entered in
Block 2. The attitudinal constructs accounted for a significant amount of variance in Elementary
mathematics grade. When mathematics self-efficacy was added, the amount of variance
accounted for in Elementary mathematics grade increased. The contributions of both attitudinal
variables decreased with the inclusion of the self-efficacy variables (Table 7).
A follow-up hierarchical multiple regression was performed, with Elementary
mathematics grade entered as the dependent variables, the attitudinal constructs entered in Block
1, and combined mathematics self-efficacy entered in Block 2. Once again, the attitudinal
constructs accounted for a significant amount of variance in Elementary mathematics grade and
this amount of variance increased with the inclusion of the combined self-efficacy. The
contributions of both attitudinal constructs decreased with the inclusion of the combined self-
efficacy variable (Table 8).
In the third hierarchical multiple regression, Elementary mathematics grade was entered
as the dependent variable, past achievement (i.e., PSLE mathematics grade and PSLE overall
score) were entered in Block 1, and the two new mathematics self-efficacy variables (i.e., general
and content-specific) were entered in Block 2. Past achievement accounted for a significant
amount of variance in Elementary mathematics grade. When mathematics self-efficacy was
added, the amount of variance accounted for in Elementary mathematics grade increased. PSLE
overall score was not a significant contributor to Elementary mathematics grade in either Block 1
or 2 and the contribution of PSLE mathematics grade decreased with the inclusion of the self-
efficacy variables (Table 9).
Summary of findings. To validate this new self-efficacy scale, I had three research goals.
First I examined the psychometric properties of the scale, including the internal consistency and
structural validity of scores. There is evidence to suggest that the scale scores had strong internal
consistency and that there were two related but distinguishable self-efficacy variables: content-
specific and general. Second, I examined the convergent validity of the mathematics self-
efficacy scores by correlating the scores with other measures (e.g., school climate, Mathematics
attitude) and found significant and positive correlations between self-efficacy scores and related
 26

constructs (i.e., school climate, mathematics attitude, overall achievement, and mathematics
grade). There was however small or no statistical difference between gender and the self-
efficacy scores. Finally, I examined the relative contribution of mathematics self-efficacy to
student achievement by simultaneous and hierarchical regression methods. Results suggested
that self-efficacy scores made significant contributions to mathematics grade, even after taking
into account the contributions of past achievement and other attitudinal constructs.
Discussion
Many studies on academic self-efficacy have been conducted in Western contexts and it
is a growing field of interest in Singapore. Many of the studies in Singapore have used scales
developed for the Western contexts, and in some cases, these scores were not validated prior to
further interpretation. Given that self-efficacy is a context- and domain-specific construct, it is
imperative to use measures of self-efficacy that closely capture aspects of the context. The
purpose of this study was to establish the psychometric properties of scores on a newly
developed mathematics self-efficacy scale for use in Singapore. The goal was that by
developing a mathematics self-efficacy scale that closely adhered to Bandura’s (2006) scale
development suggestions, this scale would yield reliable and valid academic self-efficacy scores
and that this scale can be used in future studies.
Results from this study supported the hypotheses about the reliability of the mathematics
self-efficacy scores and that a two-factor model provided a better fit for the data. Results also
supported the hypotheses that mathematics self-efficacy would be positively correlated with
other attitudinal factors, current achievement variables, and the MSLQ. I also found evidence
that the greater the correspondence in degree of specificity between the variables (e.g.,
mathematics self-efficacy and mathematics attitude vs. mathematics self-efficacy and school
climate), the greater the correlation. I did not find support for my hypotheses on gender and self-
efficacy. Of the three mathematics self-efficacy variables (i.e., content-specific, general, and
combined), the only statistical difference was found between males and females in content-
specific mathematics self-efficacy but in the opposite direction as hypothesized. This difference,
favoring females, was statistically but not practically significant.
Next, general mathematics self-efficacy accounted more variance than content-specific
mathematic self-efficacy, which was the opposite of my hypothesis. However, results did
support my other two hypotheses, that is, mathematics self-efficacy would account for more
variance than other attitudinal factors and past achievement. In the following sections, I will
explain how these findings inform current theory, explore possible reasons for why certain
hypotheses were not supported, and discuss the implications of these findings.
Psychometric Properties
Internal consistency estimates (i.e., Cronbach’s alpha) were above .80 for all major
variables. These reliability estimates suggest that the scores obtained in this study can be
interpreted. The structural validity analyses indicated that the school climate and mathematics
attitude scales had acceptable fits to the proposed theoretical model, that is, a one-factor model
for school climate and a one-factor model for mathematics attitude. The CFAs also indicated
that the mathematics self-efficacy scale had acceptable fit to the proposed theoretical model,
consisting of two factors – general mathematics self-efficacy and content-specific mathematics
self-efficacy.
These findings are important for several reasons. Firstly, Bandura (2006) had suggested
that items on a self-efficacy scale should represent gradations of challenges and should be
pegged to different levels of task difficulties. In this study, participants from all schools reported
 27

that they had not yet learned probability and statistics; about 17% of the self-efficacy items were
on these topics. These items could be conceived to represent the most difficult tasks. However
these items became problematic in the factor analysis: these items had the most number of
missing responses and these items also got parsed out as a separate factor in the EFAs. Instead
of representing a gradation of challenge, the probability and statistics items combined into a
method factor. A contribution of this study is that gradations of challenges have to be examined
further in future studies.
Another important finding of this study is that general mathematics self-efficacy and
content-specific self-efficacy are two distinct factors. This is an important finding because
although Bandura (1997) and Pajares (1997) had both previously cautioned that self-efficacy
beliefs have to be assessed “at the optimal level of specificity that corresponds to the critical task
being assessed” (Pajares, 1997, p. 11), many researchers have operationalized and assessed self-
efficacy in general terms and lacked domain- or context-specificity. However, there has been
little research to propose what this optimal level of specificity may be. Lent, Brown, ad Gore
(1997) have suggested that self-efficacy beliefs may be organized hierarchically, that is,
academic self-efficacy beliefs may encompass self-efficacy beliefs for various subjects (e.g.,
mathematics self-efficacy, science self-efficacy), which in turn encompass self-efficacy beliefs
for different aspects of each subject (e.g., algebra self-efficacy, geometry self-efficacy). The
finding of two distinct factors representing two levels of specificity allows me to answer the
question about the optimal level of specificity to measure mathematics self-efficacy in relation to
mathematics achievement (see below).
Convergent Validity
There were no gender differences on the general mathematics self-efficacy score or the
combined score, and the gender difference, favoring females, on the content-specific
mathematics self-efficacy score was statistically but not practically significant. These findings
are in line with Yeung, Lau, and Nie’s (2011) Singapore study that found that males tended to
have lower scores on motivational constructs but the findings go contrary to findings on meta-
analysis (Huang, 2013) which found that males had higher academic self-efficacy, although the
effect size was small. Overall, the state of the literature including this current study seems to
suggest that the relationship between gender and academic self-efficacy tends to be either
statistically insignificant or statistically but not practically significant. Of note, within this
study’s sample, there were also no significant differences between the genders in their past or
present mathematics achievement.
Next, as hypothesized, mathematics self-efficacy scores had positive correlations with
other attitudinal constructs, as well as current achievement. Findings also support further
hypotheses that mathematics self-efficacy scores would have stronger correlations with
constructs of similar grain size (i.e., mathematics attitude and mathematics grade) than with
constructs of different grain size (i.e., attitude towards school and overall performance). These
findings provide support for the convergent validity of the self-efficacy scores.
Of note, compared to combined self-efficacy score and content-specific self-efficacy
score, general mathematics self-efficacy score had the strongest correlations with four of the five
convergent variables (i.e., school climate score, mathematics attitude score, overall percentage,
and MSLQ self-efficacy score). The effect sizes ranged from medium to large. This finding
suggests that general self-efficacy may be the optimal level of specificity to measure
mathematics self-efficacy in relation to mathematics achievement. More evidence to support this
claim is provided in the following section.
 28

Concurrent Validity
Mathematics self-efficacy scores contributed significantly and substantially to the
variances in Elementary mathematics grade, even after taking into account the contributions of
other attitudinal constructs. In the first regression analysis, of all the attitudinal constructs,
mathematics attitude scores made the most contribution, followed by content-specific
mathematics self-efficacy scores, and then general mathematics self-efficacy; school climate
made the least contribution. This was contrary to the hypothesis, that mathematics self-efficacy
was not the greatest contributor to mathematics grade. One interpretation could be that
mathematics attitude is a more important variable than mathematics self-efficacy. However,
upon closer inspection of the results, the difference in contribution to variance between
mathematics attitude and mathematics self-efficacy was only .05.
Further, in the second regression analysis, when the two mathematics self-efficacy
variables were combined into one variable, this combined mathematics self-efficacy accounted
for more variance in Elementary mathematics grade than mathematics attitude. Results from the
second regression analysis is congruent with past research that found self-efficacy to be an
important variable in relation to academic achievement (e.g., Pajares & Kranzler, 1995;
Zimmerman et al., 1992) and in some studies, the variable with the more predictive validity than
other motivational constructs (e.g., Mone et al., 1995; Pajares & Miller, 1994). Comparing the
results from these two regression analyses, it seems like in order for mathematics self-efficacy to
be the variable that contributes the most to the variance of achievement, mathematics self-
efficacy should include items that are both content-specific and general.
In the third hierarchical multiple regression, past achievement in the same subject
contributed significantly to the variance in Elementary mathematics grade whereas past overall
achievement did not. Mathematics self-efficacy scores contributed significantly and
substantially to the variance in Elementary mathematics grade, even after taking into account the
contribution of past mathematics achievement. Specifically, general mathematics self-efficacy
scores contributed more than twice to the variance of Elementary mathematics grade than past
achievement, and content-specific mathematics self-efficacy scores made a smaller contribution
than past achievement, albeit still significant. These results suggest that self-efficacy scores
might be more predictive of current achievement than past achievement.
An interesting finding was gleaned from comparing the results from the first and third
hierarchical multiple regressions: the amount of variance accounted for by general mathematics
self-efficacy increased by .16 between regression one and three but the amount of variance
accounted for by content-specific mathematics self-efficacy decreased by .09. A possible
explanation for this is that in the third regression, content-specific mathematics self-efficacy’s
contribution could have been depressed by the significant contribution of past mathematics
achievement since both variables were assessing students’ content knowledge. Similarly, in the
first regression, general mathematics self-efficacy’s contribution could have been depressed by
the significant contribution of mathematics attitude since both variables were assessing students’
more general perception to mathematics.
Although the order of data collection in this study does not allow for the mathematics
self-efficacy variables to be considered predictors, the pattern of findings in this study is in line
with previous research that found that predictor variables have greater predictive utility when
they are commensurate with outcome variables in terms of content and levels of specificity (e.g.,
Choi, 2005; Lent et al., 1997). For example, mathematics attitude score and mathematics self-
efficacy scores accounted for more variance in current mathematics grade than school climate
 29

scores; mathematics self-efficacy scores and past mathematics achievement accounted for more
variance in current mathematics grade than past overall achievement.
To answer the question on what is the optimal level of specificity to measure
mathematics self-efficacy, results from the regressions analyses in this study make a few
suggestions. First, in the simultaneous regression comparing the two self-efficacy variables and
in the regression analysis comparing past achievement to mathematics self-efficacy, results
suggest that the optimal level of specificity may be at the level of general mathematics self-
efficacy. In the regressions comparing attitudinal factors with self-efficacy, results suggest that
there is not a big difference between content-specific or general mathematics self-efficacy and as
a single variable, combined mathematics self-efficacy made more contribution than other
attitudinal factors.
An implication of this finding is that it is beneficial to measure self-efficacy at both the
content-specific and general level. However, it can also be argued that general mathematics self-
efficacy may be a fruitful focus going forward. One reason for this is that a general mathematics
self-efficacy scale can be more easily adapted to and validated in different contexts. Second,
there was no significant gender differences in responses to the general mathematics scale. Third,
it is more time-efficient to administer a general mathematics scale. However more research on
the degree of specificity has to be done to substantiate theses implications.
Limitations and Future Research
The findings of this study suggest that excluding items with low factor loadings, the newly
developed scales had scores that were reliable and valid. The question about degree of
specificity and validity also arose in the area of gradations of challenges. Bandura (2006)
pointed out that since self-efficacy exists on a continuum, items should represent gradations of
challenges and should be pegged to different levels of task difficulties. Findings from this study
suggest that it is possible to create items that are so difficult that combine into a method factor
and these most challenging items then have to be excluded from analysis. In future research,
care has to be taken to construct items that represented a continuum of challenges but were still
within reach for a small group of students.
An important finding of this study was that Mathematics self-efficacy scores compared to
past achievement were found to have stronger relationships with current grade. In most cases,
general mathematics self-efficacy scores were found to have stronger relationship with various
attitudinal constructs than content-specific mathematics self-efficacy scores. This is one of the
most important contributions of this study, as it gives us some insight into what the optimal level
of specificity may be when assessing mathematics self-efficacy in relation to current
mathematics performance. More research on the level of specificity is needed to substantiate
that the optimal level is indeed at that of the general mathematics self-efficacy.
There is also some evidence from this study to suggest that the general mathematics self-
efficacy scale and the MSLQ have similar relations to many other constructs. More research is
needed to compare the reliability and validity of the scores of this newly developed scale
compared to the MSLQ, which is more widely-used but not contextualized. Such research may
elucidate the context-specificity of self-efficacy and how self-efficacy measures should be
designed.
A question that arose from the findings of this study was why there was a small gender
difference in content-specific self-efficacy scores but not in the other two self-efficacy scores. A
multi-national study by the Organisation for Economic Co-operation and Development (OECD,
2013) found no gender differences in self-efficacy when students were asked about doing tasks
 30

that clearly matched their curriculum (e.g., solve a linear equation). However, when students
were asked about their perceived ability to solve applied math problems that were associated
with stereotypical gender roles (e.g., calculating how many tiles would be needed to cover a
floor), 67% of boys reported feeling confident or very confident, compared to 44% of girls. The
findings of this study contradict the findings from OECD and suggest that more research has to
be conducted in this area to better understand the interaction among gender, types of self-
efficacy questions, and achievement.
One of the limitations of this study is that it was only administered to students in one grade
level and within one country. To further validate the scores, this new self-efficacy scale has to
be administered to more Secondary 3 students in Singapore. Given the preliminary nature of the
scale development, additional research can help to refine the scale in terms of efficiency (e.g.,
what is the minimal number of questions needed without compromising of the reliability an
validity of the scores). In addition, to establish the generalizability and usability of this scale, the
scale can be administered to students from different grades and possibly different countries.
Given the context-specificity of self-efficacy, I predict that the general mathematics self-efficacy
items would have better reliability and validity than the content-specific scores. Administering
this new mathematics self-efficacy scale to students of different grades or different countries can
also expand our understanding on whether self-efficacy may be used as a universal predictor of
academic achievement.
Another limitation of this study is that the scale was administered after the students had taken
the mathematics examination and some of them have even received their grades. This could
have affected the students’ perception of their abilities and this order of administration also
precluded any ability to make statements about the predictive utility of self-efficacy scores for
achievement. To make contributions regarding the predictive utility of self-efficacy scores,
future research can attempt to measure students’ self-efficacy beliefs before they take an
achievement test.
In conclusion, this study made several significant contributions but it also has some
limitations. Although self-efficacy has been studied extensively, this study shows us that there
exist many exciting research opportunities in which our understanding about the nature of self-
efficacy can be further expanded.
 31

References
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavior change. Psychological
Review, 84, 191–215. doi:10.1037/0033-295X.84.2.19
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory.
Englewood Cliffs, NJ: Prentice-Hall.
Bandura, A. (1989). Multidimensional scales of perceived self-efficacy. Unpublished test,
Stanford University, Stanford, CA.
Bandura, A. (1995). Exercise of personal and collective efficacy in changing societies. In A.
Bandura (Ed.), Self-efficacy in changing societies (pp. 1–45). Cambridge, UK:
Cambridge University Press.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: W. H. Freeman &
Company.
Bandura, A. (2002). Social cognitive theory in cultural context. Applied Psychology: An
International Review, 51, 269-290. doi:10.1111/1464-0597.00092
Bandura, A. (2006). Guide for constructing self-efficacy scales. In F. Pajares & T. Urdan (Eds.),
Self-efficacy beliefs of adolescents (pp. 307–338). Greenwich, CT: Information Age.
Bandura, A., & Schunk, D. H. (1981). Cultivating competence, self-efficacy, and intrinsic
interest through proximal self-motivation. Journal of Personality and Social Psychology,
41, 586–598. doi:10.1037/0022-3514.41.3.586
Betz, N. E., & Hackett, G. (1983). The relationship of mathematics self-efficacy expectations to
the selection of science-based college majors. Journal of Vocational Behavior, 23, 329–
345. doi:10.1016/0001-8791(83)90046-5
Busch, T. (1995). Gender differences in self-efficacy and academic performance among students
of business administration. Scandinavian Journal of Educational Research, 39, 311-318.
doi:10.1080/0031383950390403
Byrne, B. M. (2006). Structural equation modeling with EQS: Basic concepts, applications, and
programming (2nd ed.). Mahwah, NJ: Erlbaum.
Central Intelligence Agency. (2016, February). The world factbook. Retrieved from
https://www.cia.gov/library/publications/the-world-factbook/geos/sn.html
Chemers, M. M., Hu, L., & Garcia, B. F. (2001). Academic self-efficacy and first-year college
student performance and adjustment. Journal of Educational Psychology, 93, 55–64.
doi:10.1037/0022-0663.93.1.55
Chen, W., & Looi, C. (2011). Active classroom participation in a Group Scribbles primary
science classroom. British Journal of Educational Technology, 42, 676–686.
doi:10.1111/j.1467-8535.2010.01082.x
Choi, N. (2005). Self-efficacy and self-concept as predictors of college students’ academic
performance. Psychology in the Schools, 42, 197-205. doi:10.1002/pits.20048
Chong, W. H. (2007). The role of personal agency beliefs in academic self-regulation: An Asian
perspective. School Psychology International, 28, 63–76.
doi:10.1177/0143034307075681
Crockett, L. J., Schulenberg, J. E., & Petersen, A.C. (1987). Congruence between objective and
self-report data in a sample of young adolescents. Journal of Adolescent Research, 2,
383–392. doi:10.11.77/074355488724006
Dowling, D. M. (1978). The development of a mathematics confidence scale and its application
in the study of confidence in women college students. Unpublished doctoral dissertation,
Ohio State University, Columbus.
 32

Eaton, M. J., & Dembo, M. H. (1997). Differences in motivational beliefs of Asian American
and Non-Asian students. Journal of Educational Psychology, 89, 433–440.
doi:10.1037/0022-0663.89.3.433
Ewart, C. K. (1992). Role of physical self-efficacy in recovery from heart attack. In R.
Schwarzer (Ed.), Self-efficacy: Thought control of action (pp. 287–304). Washington, DC:
Hemisphere.
Feltz, D. F., & Lirgg, C. D. (2001). Self-efficacy beliefs of athletes, teams, and coaches. In R. N.
Singer, H. A. Hausenblas, & C. M. Janelle (Eds.), Handbook of sport psychology (pp.
340–361). New York, NY: John Wiley & Sons.
Ferla, J., Valcke, M., & Cai, Y. (2009). Academic self-efficacy and academic self-concept:
Reconsidering structural relations. Learning and Individual Differences, 19, 499–505.
doi:10.1016/j.lindif.2009.05.004
Hampton, N. Z., & Mason, E. (2003). Learning disabilities, gender, sources of efficacy, self-
efficacy beliefs, and academic achievement in high school students. Journal of School
Psychology, 41, 101-112. doi:10.1016/S0022-4405(03)00028-1
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structural analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling, 6(1), 1-55.
doi:10.1080/10705519909540118
Huang, C. (2013). Gender differences in academic self-efficacy: A meta-analysis. European
Journal of Psychology of Education, 28, 1-35. doi:10.1007/s10212-011-0097-y
Kavanagh, D. J. & Bower, G. H. (1985). Mood and self-efficacy: Impact of joy and sadness on
perceived capabilities. Cognitive Therapy and Research, 9, 507–525.
Kitayama, S., Markus, H. R., Matsumoto, H., & Norasakkunkit, V. (1997). Individual and
collective processes in the construction of the self: Self-enhancement in the United States
and self-criticism in Japan. Journal of Personality and Social Psychology, 72, 1245-1267.
doi:10.1037/0022-3514.72.6.1245
Klassen, R. M. (2004). Optimism and realism: A review of self-efficacy from a cross-cultural
perspective. International Journal of Psychology, 39, 205-230.
doi:10.1080/00207590344000330
International Association for the Evaluation of Educational Achievement. (2011). Trends in
international mathematics and science study 2011 [Grade 8 student questionnaire].
Retrieved from http://timssandpirls.bc.edu/timss2011/downloads/T11_StuQ_8.pdf
Lau, S., Liem, A. D., & Nie, Y. (2008). Task- and self-related pathways to deep learning: The
mediating role of achievement goals, classroom attentiveness, and classroom
participation. British Journal of Educational Psychology, 78, 639–662.
Lent, R. W., Brown, S. D., & Gore, P. A. (1997). Discriminant and predictive validity of
academic self-concept, academic self-efficacy, and mathematics-specific self-efficacy.
Journal of Counseling Psychology, 44, 307-315. doi:10.1037/0022-0167.44.3.307
Lent, R. W., Lopez, F. G., & Bieschke, K. J. (1993). Predicting mathematics-related choice and
success behaviors: Test of an expanded social cognitive model. Journal of Vocational
Behavior, 42, 223–236. doi:10.1006/jvbe.1993.1016
Leung, F. K. S. (2002). Behind the high achievement of East Asian students. Educational
Research and Evaluation, 8, 87–108. doi:10.1076/edre.8.1.87.6920
Li, J. (2011). Cultural frames of children’s learning beliefs. In J. Li (Ed.), Bridging cultural and
developmental approaches to psychology: New syntheses in theory, research, and policy
(pp. 26–48). New York: Oxford University Press.
 33

Liem, A. D., Lau, S., & Nie, Y. (2008). The role of self-efficacy, task value, and achievement
goals in predicting learning strategies, task disengagement, peer relationship, and
achievement outcome. Contemporary Educational Psychology, 33, 486–512.
doi:10.1016/j.cedpsych.2007.08.001
Luo, W., Paris, S. G., Hogan, D., & Luo, Z. (2011). Do performance goals promote learning? A
pattern analysis of Singapore students’ achievement goals. Contemporary Educational
Psychology, 36, 165–176. doi:10.1016/j.cedpsych.2011.02.003
Marsh, H. W., Hau, K., & Wen, Z. (2004). In search of golden rules: Comment on hypothesis
testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing
Hu and Bentler’s (1999) findings. Structural Equation Modeling: A Multidisciplinary
Journal, 11, 320-341. Doi:10.1207/s15328007sem1103_2
Martin, J. J., & Gill, D. L. (1995). The relationship of competitive orientations and self-efficacy
to goal importance, thoughts, and performance in high school distance runners. Journal
of Applied Sports Psychology, 7, 50–62. doi:10.1080/10413209508406300
Midgley, C., Maehr, M. L., & Urdan, T. (1993). Pattern of adaptive learning survey (PALS).
Ann Arbor, MI: University of Michigan, Combined Program in Education and
Psychology.
Ministry of Education. (2006). Secondary Mathematics Syllabuses. Retrieved from
http://www.moe.gov.sg/education/syllabuses/sciences/files/maths-secondary.pdf
Ministry of Education. (2012, November). School terms and holidays 2013. Retrieved from
http://www.moe.gov.sg/schools/terms-and-holidays/2013/
Ministry of Information, Communications, and the Arts. (2009, November). Singapore at a
glance. Retrieved from http://app.www.sg/who/default.aspx
Mone, M. A., Baker, D. D., & Jeffries, F. (1995). Predictive validity and time dependency of
self-efficacy, self-esteem, personal goals, and academic performance. Educational and
Psychological Measurement, 55, 716–727. doi:10.1177/0013164495055005002
Martin, M. O., Mullis, I. V. S., Foy, P., & Stanco, G. M. (2012). TIMSS 2011 international
results in science. Retrieved from
http://timssandpirls.bc.edu/timss2011/downloads/T11_IR_Science_FullBook.pdf
Mullis, I. V. S., Martin, M. O., Foy, P. & Arora, A. (2012). TIMSS 2011 international results in
mathematics. Retrieved from
http://timssandpirls.bc.edu/timss2011/downloads/T11_IR_Mathematics_FullBook.pdf
Mullis, I. V. S., Martin, M. O., Foy, P., & Drucker, K. T. (2012). PIRLS 2011 international
results in reading. Retrieved from
http://timssandpirls.bc.edu/pirls2011/downloads/P11_IR_FullBook.pdf
Nahas, M. V., Goldfine, B., & Collins, M. A. (2003). Determinants of physical activity and
young adults: the basis for high school and college physical education to promote active
lifestyles. Physical Educator, 60, 42–56.
Nie, Y., & Lau, S. (2010). Differential relations of constructivist and didactic instruction to
students’ cognition, motivation, and achievement. Learning and Instruction, 20, 411–423.
doi:10.1016/j.learninstruc.2009.04.002
Nie, Y., Lau, S., & Liau, A. K. (2011). Role of academic self-efficacy in moderating the relation
between task importance and test anxiety. Learning and Individual Differences, 21, 736–
741. doi:10.1016/j.lindif/2011/09/005
 34

Nielsen, I. L., & Moore, K. A. (2003). Psychometric data on the mathematics self-efficacy scale.
Educational and Psychological Measurement, 63, 128–138.
doi:10.1177/0013164402239321
Norman, G. (2010). Likert scales, levels of management, and the “laws” of statistics. Advances
in Health Sciences Education, 15, 625-632. doi: 10.1007/s10459-010-9222-y
Organisation for Economic Co-operation and Development. (2012). Program for international
student assessment 2012 [Student questionnaire Form B]. Retrieved from
http://pisa2012.acer.edu.au/downloads/MS12_StQ_FORM_B_ENG.pdf
Organisation for Economic Co-operation and Development. (2013). PISA 2012 results: Ready to
Learn: Students’ engagement, drive and self-beliefs (Volume III), PISA, OECD
Publishing. doi:10.1787/9789264201170-en
Pajares, F. (1997). Current directions in self-efficacy research. In M. Maaehr & P. R. Pintrich
(Eds.), Advances in motivation and achievement (pp. 1-49). Greenwich, CT: JAI Press.
Pajares, F., Hartley, J., & Valiante, G. (2001). Response format in writing self-efficacy
assessment: Greater discrimination increases prediction. Measurement and Evaluation in
Counseling and Development, 33, 214–221.
Pajares, F., & Kranzler, J. (1995). Self-efficacy beliefs and general mental ability in
mathematical problem-solving. Contemporary Educational Psychology, 20, 426–443.
doi:10.1006/ceps.1995.1029
Pajares, F. & Miller, M. D. (1994). The role of self-efficacy and self-concept beliefs in
mathematical problem-solving: A path analysis. Journal of Educational Psychology, 86,
193–203. doi:10.1037/0022-0663.86.2.193
Pajares, F., & Valiante, G. (2006). Self-efficacy beliefs and motivation in writing development.
In F. Pajares, & G. Valiante (Eds.), Handbook of writing research (pp. 158–170). New
York: Guilford Press.
Pintrich, P. R., & De Groot, E. V. (1990). Motivational and self-regulated learning components
of classroom academic performance. Journal of Educational Psychology, 82, 33–40.
doi:10.1037/0022-0663.82.1.33
Pintrich, P. R., Smith, D. A. F., Garcia, T., & McKeachie, W. J. (1991). The Motivated Self-
Regulated Learning Questionnaire (MSLQ). Ann Arbor, MI: National Center for
Research to Improve Postsecondary Teaching and Learning, University of Michigan.
Pintrich, P. R., Smith, D. A. F., Garcia, T., & McKeachie, W. J. (1993). Reliability and
predictive validity of the Motivated Strategies for Learning Questionnaire (MSLQ).
Education and Psychological Measurement, 53, 801–813.
doi:10.1177/0013164493053003024
Salili, F., Chiu, C., & Lai, S. (2001). The influence of culture and context on students’
motivational orientation and performance. In F. Salili, C. Chiu, & Y. Hong (Eds.),
Student motivation: The culture and context of learning (pp. 221–247). New York, NY:
Kluwer Academic/Plenum.
Schunk, D. H. (1989). Self-efficacy and achievement behaviors. Educational Psychology Review,
1, 173–208.
Schunk, D. H. (1995). Self-efficacy and education and instruction. In J. E. Maddux (Ed.), Self-
efficacy, adaptation, and adjustment: Theory, research, and application (pp. 281–303).
New York: Plenum Press.
Schunk, D. H., & Zimmerman, B. J. (1994). Self-regulation in education: Retrospect and
prospect. In D. H. Schunk, & B. J. Zimmerman (Eds.), Self-regulation of learning and
 35

performance and educational applications (pp. 305–314). Hillsdale, NJ: Lawrence

Erlbaum Associates.
Smith, R. E. (1989). Effects of coping skills training on generalized self-efficacy and locus of
control. Journal of Personality and Social Psychology, 56, 228–233. doi:10.1037/0022-
3514.56.2.228
Stankov, L., Lee, J., Luo, W., & Hogan, D. J. (2012). Confidence: A better predictor of academic
achievement than self-efficacy, self-concept and anxiety? Learning and Individual
Differences, 22, 747–758. doi:10.1016/j.lindif.2012.05.013
Statistics Singapore. (2009, September). Time series on population (Mid-year estimates).
Retrieved from http://www.singstat.gov.sg/stats/themes/people/hist/popn.html
Steinberg, L., Dornbusch, S. M., & Brown, B. B. (1992). Ethnic differences in adolescents
achievement: An ecological perspective. American Psychologist, 47, 723-729. doi:
10.1037/0003-066X.47.6.723
Stevenson, H. W., & Lee, S. Y. (1999). The academic achievement of Chinese students. In M. H.
Bond (Ed.), The handbook of Chinese psychology (pp. 124-142). Hong Kong: Oxford
University Press.
Tabachnick, B. G., & Fidell, L. S. ( 2001). Using multivariate statistics (4th ed.). Needham
Heights, MA: Allyn & Bacon.
Tan, H. H., & Chee, D. (2005). Understanding interpersonal trust in a Confucian-influenced
society: An exploratory study. International Journal of Cross Cultural Management, 5,
197–212. doi:10.1177/1470595805054493
Tan, J. B., & Yates, S. (2011). Academic expectations as sources of stress in Asian students.
Social Psychology of Education, 14, 389–407. doi:10.1007/s11218-010-9146-7
Taylor, R. (1990). Interpretation of the correlation coefficient: A basic review. Journal of
Diagnostic Medical Sonography, 6, 35–39.
Usher, E. L., & Pajares, F. (2008). Sources of self-efficacy in mathematics: A validation study.
Contemporary Educational Psychology, 34, 89-101. doi:10.1016/j.cedpsych.2008.09.002
Watkins, M. W. (2000). MonteCarlo PCA for parallel analysis [Computer software]. State
College, PA: Ed & Psych Associates.
Wu, X. Y., Pender, N., & Noureddine, S. (2003). Gender differences in the psychosocial
cognitive correlates of physical activity among Taiwanese adolescents: A structural
equation modeling approach. International Journal of Behavioral Medicine, 10, 93–105.
doi:10.1207/S15327558IJBM1002_01
Yeung, A. S., Lau, S., & Nie, Y. (2011). Primary and secondary students’ motivation in learning
English: Grade and gender differences. Contemporary Educational Psychology, 36, 246–
256. doi:10.1016/j.cedpsych.2011.03.001
Zimmerman, B. J., Bandura, A., & Martinez-Pons, M. (1992). Self-motivation for academic
attainment: The role of self-efficacy beliefs and personal goal setting. American
Educational Research Journal, 29, 663–676. doi:10.2307/1163261
Zimmerman, B. J., & Cleary, T. J. (2006). Adolescents’ development of personal agency: The
role of self-efficacy beliefs and self-regulatory skills. In F. Pajares & T. Urdan (Eds.),
Self-efficacy beliefs of adolescents (pp. 45–70). Greenwich, CT: Information Age.
Zimmerman, B. J., & Kitsantas, A. (2002). Acquiring writing revision and self-regulatory skill
through observation and emulation. Journal of Educational Psychology, 94, 660–668.
doi:10.1037/0022-0663.94.4.660
 36

Zimmerman, B. J., & Ringle, J. (1981). Effects of model persistence and statements of
confidence on children’s efficacy and problem-solving. Journal of Educational
Psychology, 73, 485–493. doi:10.1037/0022-0663.73.4.485

 37

Table 1
Descriptive Statistics for Independent Variables in the Study (N = 1,587)
Variables M SD
School Climate Item 1 4.16 1.17
School Climate Item 2 5.38 1.12
School Climate Item 3 3.98 1.28
School Climate Item 4 3.68 1.30
School Climate Item 5 4.86 1.26
School Climate Item 6 4.29 1.33
School Climate Item 7 4.58 1.27
School Climate Item 8 4.56 1.28
School Climate Item 9 3.78 1.36
School Climate Item 10 4.49 1.46
School Climate Combined Mean 4.37 0.76
Mathematics Attitude Item 1 3.76 1.63
Mathematics Attitude Item 2 4.33 1.48
Mathematics Attitude Item 3 4.18 1.52
Mathematics Attitude Item 4 4.84 1.24
Mathematics Attitude Item 5 3.26 1.54
Mathematics Attitude Item 6 4.32 1.54
Mathematics Attitude Item 7 4.50 1.29
Mathematics Attitude Item 8 3.50 1.45
Mathematics Attitude Item 9 3.95 1.39
Mathematics Attitude Item 10 4.43 1.47
Mathematics Attitude Combined Mean 4.11 0.93
Mathematics Self-Efficacy Item 1 4.06 1.18
Mathematics Self-Efficacy Item 2 4.37 1.46
Mathematics Self-Efficacy Item 3 4.00 1.70
Mathematics Self-Efficacy Item 4 4.75 1.16
Mathematics Self-Efficacy Item 5 4.93 1.32
Mathematics Self-Efficacy Item 6 3.18 1.63
Mathematics Self-Efficacy Item 7 4.54 1.25
Mathematics Self-Efficacy Item 8 3.98 1.61
Mathematics Self-Efficacy Item 9 3.26 1.62
Mathematics Self-Efficacy Item 10 3.26 1.51
Mathematics Self-Efficacy Item 11 4.56 1.44
Mathematics Self-Efficacy Item 12 3.55 1.59
Mathematics Self-Efficacy Item 13 2.78 1.49
Mathematics Self-Efficacy Item 14 4.89 1.46
Mathematics Self-Efficacy Item 15 4.82 1.29
Mathematics Self-Efficacy Item 16 4.52 1.45
Mathematics Self-Efficacy Item 17 3.57 1.70
Mathematics Self-Efficacy Item 18 4.57 1.47
Mathematics Self-Efficacy Item 19 4.53 1.47
Mathematics Self-Efficacy Item 20 4.19 1.63
Mathematics Self-Efficacy Item 21 2.71 1.54
 38

Mathematics Self-Efficacy Item 22 3.52 1.62

Mathematics Self-Efficacy Item 23 4.40 1.32
Mathematics Self-Efficacy Item 24 4.64 1.49
Mathematics Self-Efficacy Item 25 4.13 1.69
Mathematics Self-Efficacy Item 26 4.47 1.49
Mathematics Self-Efficacy Item 27 3.42 1.70
Mathematics Self-Efficacy Item 28 3.37 1.68
Mathematics Self-Efficacy Item 29 4.46 1.36
Mathematics Self-Efficacy Item 30 4.43 1.56
Mathematics Self-Efficacy Combined Mean 4.07 0.96
a
Mathematics Self-Efficacy (General) Mean 4.17 1.02
Mathematics Self-Efficacy (Content) Mean b 4.01 1.02
MSLQ Item 1 4.14 1.71
MSLQ Item 2 3.78 1.64
MSLQ Item 3 4.90 1.57
MSLQ Item 4 3.95 1.60
MSLQ Item 5 4.11 1.62
MSLQ Item 6 4.67 1.77
MSLQ Item 7 4.35 1.64
MSLQ Item 8 4.24 1.67
MSLQ Item Combined Mean 4.27 1.46
Note. School climate, mathematics attitude, and mathematics self-efficacy items ranged from 1
to 6. MSLQ items ranged from 1 to 7.
a
Ten mathematics self-efficacy items contributed to the General combined mean: items 1, 4, 7,
10, 11, 12, 17, 19, 23, and 29. bTwenty mathematics self-efficacy items contributed to the
Content combined mean: items 2, 3, 5, 6, 8, 9, 13, 14, 15, 16, 18, 20, 21, 22, 24, 25, 26, 27, 28,
and 30.


Table 2

Correlations Matrix of Major Variables in the Study (N = 1,587)

Variables 1 2 3 4 5 6 7 8 9 10
1. School Climate 1.00*
2. Mathematics Attitude .36* 1.00*
3. Mathematics Self-Efficacy Combined .32* .60* 1.00*
4. Mathematics Self-Efficacy General .37* .69* .92* 1.00*
5. Mathematics Self-Efficacy Content .25* .48* .96* .76* 1.00*
6. MSLQ .34* .66* .70* .74* .59* 1.00*
7. Elementary Mathematics Grade .09* .50* .57* .56* .52* .55* 1.00*
8. Overall Percentage .13* .43* .49* .49* .44* .49* .73* 1.00*
9. PSLE Mathematics Grade .04* .18* .41* .30* .45* .22* .35* .28* 1.00*
10. PSLE T-score .03* .06* .41* .23* .49* .10* .27* .21* .71* 1.00*
*p < .01
39

40

Table 3
Reliability Estimates for Independent Variables in the Study (N = 1,587)
Variables α
School Climate .76
Mathematics Attitude .86
Mathematics Self-Efficacy Combined .94
Mathematics Self-Efficacy General .88
Mathematics self-Efficacy Content .92
MSLQ .96



41

Table 4
Measures of Factorability of Scores (N = 414)
Variables KMO Barlett’s Test of Sphericity
School Climate .83 χ2(45) = 1073.81
Mathematics Attitude .84 χ2(45) = 1635.82
Mathematics Self-Efficacy .95 χ2(435) = 7564.05
MSLQ .94 χ2(28) = 13400.78


Table 5

Fit Indices Derived from Confirmatory Factor Analyses (Maximum Likelihood Robust)
Model χ2s-b df χ2/df TLI CFI RMSEA 90% CI SRMR
School Climate 38.99* 9 0.11 .960 .976 .054 .038, .072 .029
(1-factor; 6 items)
Mathematics Attitude 110.60* 9 12.29 .941 .965 .098 .082, .115 .031
(1-factor; 6 items)
Mathematics Self-Efficacy 1358.32* 170 7.99 .867 .881 .077 .073, .081 .050
(1-factor; 20 items)
Mathematics Self-Efficacy 784.27* 169 4.64 .931 .938 .056 .052, .060 .036
(2-factors: Content-Specific &
General; 20 items)
Note. N = 1,173; s-b = Satorra-Bentler; df = degrees of freedom; NNFI = non-normed fit index; CFI = comparative fit index; RMSEA =
root mean square error of approximation; CI = confidence interval; SRMR = standardized root mean square residual.
*p < .001
42

 43

Table 6
t-tests on Gender, Mathematics Achievement, and Mathematics Self-Efficacy
M
Variable Male Female t df p
n = 749 n = 823
PSLE Mathematics Grade 3.12 (1.08) 3.09 (1.08) .60 1570 .549
Elementary Mathematics 5.79 (2.50) 5.67 (2.58) .89 1570 .375
Grade
Mathematics Self-Efficacy 4.27 (1.07) 4.39 (.93) -2.40 1570 .016
Combined
Mathematics Self-Efficacy 4.25 (1.02) 4.22 (.95) 0.65 1570 .514
General
Mathematics Self-Efficacy 4.29 (1.21) 4.54 (1.05) -4.34 1570 .000
Content-Specific
Note. PSLE Mathematics grade ranged from 1 to 5 and Elementary Mathematics grade ranged
from 1 to 9, where a lower score represented a better grade. Self-efficacy variables ranged from
1 to 6, where a higher score represented higher self-efficacy.

 44

Table 7
Hierarchical Multiple Regression on Attitudinal Constructs, Content-Specific, and General Self-
Efficacy Variables (N = 1,587)
B SE β Adjusted
R2
Step 1 .26
Constant -2.13 .29
School Climate -0.30 .07 -0.10*
Mathematics Attitude 1.22 .05 0.54*
Step 2 .38
Constant -0.14 .29
School Climate -0.46 .06 -0.16*
Mathematics Attitude 0.63 .06 0.28*
Content-Specific Mathematics 0.57 .07 0.25*
Self-Efficacy
General Mathematics Self- 0.59 .10 0.23*
Efficacy
*p < .001

 45

Table 8
Hierarchical Multiple Regression on Attitudinal Constructs and Combined Self-Efficacy
Variable (N = 1,587)
B SE β Adjusted
R2
Step 1 .26
Constant -2.13 .29
School Climate -0.30 .07 -0.10*
Mathematics Attitude 1.22 .05 0.54*
Step 2 .38
Constant -0.13 .29
School Climate -0.46 .06 -0.16*
Mathematics Attitude 0.66 .06 0.29*
Combined Mathematics Self- 1.13 .06 0.45*
Efficacy
*p < .001
 46

Table 9
Hierarchical Multiple Regression on Past Achievement and Self-Efficacy (N = 1,587)
B SE β Adjusted
R2
Step 1 .12
Constant -2.52 .47
PSLE Mathematics Grade 0.72 .08 0.31*
PSLE Overall Score 0.01 .00 0.05*
Step 2 .35
Constant -1.00 .45
PSLE Mathematics Grade 0.40 .07 0.17*
PSLE Overall Score -0.00 .00 - .02*
Content-Specific Mathematics 0.36 .08 0.16*
Self-Efficacy
General Mathematics Self- 0.99 .08 0.39*
Efficacy
*p < .001

 47

Appendix A
School Climate Survey
The following statements are about how a student may feel about school. Read each statement
and decide how much you agree. Fill in the bubble of your corresponding choice. There are no
right or wrong responses, so pick the first response that comes to mind. Please answer every
question.

1 = 0% 2 = 20% 3 = 40% 4 = 60% 5 = 80% 6 = 100%

Agree Agree Agree Agree Agree Agree
1 2 3 4 5 6
1. Most of the teachers in this school are interested
¡ ¡ ¡ ¡ ¡ ¡
in students’ well being.
2. I do not have friends in this school. ¡ ¡ ¡ ¡ ¡ ¡
3. I am happy at school. ¡ ¡ ¡ ¡ ¡ ¡
4. School has helped give me confidence to make
¡ ¡ ¡ ¡ ¡ ¡
decisions.
5. I feel like an outsider at school. ¡ ¡ ¡ ¡ ¡ ¡
6. Trying hard at school will help me get a good
¡ ¡ ¡ ¡ ¡ ¡
job.
7. My teachers do not listen to what I have to say. ¡ ¡ ¡ ¡ ¡ ¡
8. When I have problems in school, I have friends
¡ ¡ ¡ ¡ ¡ ¡
who can help me out.
9. I am proud to be in this school. ¡ ¡ ¡ ¡ ¡ ¡
10. School has been a waste of time. ¡ ¡ ¡ ¡ ¡ ¡
 48

Appendix B
Mathematics Attitude Survey
The following statements are about how a student may feel about mathematics. Read each
statement and decide how much you agree. Fill in the bubble of your corresponding choice.
There are no right or wrong responses, so pick the first response that comes to mind. Please
answer every question.

1 = 0% 2 = 20% 3 = 40% 4 = 60% 5 = 80% 6 = 100%

Agree Agree Agree Agree Agree Agree
1 2 3 4 5 6
1. Mathematics is one of my favorite subjects. ¡ ¡ ¡ ¡ ¡ ¡
2. I do not understand why I have to learn
¡ ¡ ¡ ¡ ¡ ¡
mathematics.
3. Studying mathematics makes me feel nervous. ¡ ¡ ¡ ¡ ¡ ¡
4. If I put in effort I can succeed in mathematics. ¡ ¡ ¡ ¡ ¡ ¡
5. I am happier in a mathematics class than in any
¡ ¡ ¡ ¡ ¡ ¡
other class.
6. I do badly in mathematics whether or not I study
¡ ¡ ¡ ¡ ¡ ¡
for my exams.
7. I need to do well in mathematics to help me get a
¡ ¡ ¡ ¡ ¡ ¡
good job.
8. I look forward to my mathematics lessons. ¡ ¡ ¡ ¡ ¡ ¡
9. I am willing to study harder for mathematics than
¡ ¡ ¡ ¡ ¡ ¡
for any other subject.
10. I have never enjoyed studying mathematics in
¡ ¡ ¡ ¡ ¡ ¡
school.
 49

Appendix C
Mathematics Self-Efficacy Survey
Students have different levels of confidence about math. Based on your current level of
confidence, read each statement and decide how much you agree. Fill in the bubble of your
corresponding choice. There are no right or wrong responses, so pick the first response that
comes to mind. Please answer every question. Do not attempt to solve the examples given.

1 = 0% 2 = 20% 3 = 40% 4 = 60% 5 = 80% 6 = 100%

Agree Agree Agree Agree Agree Agree
1 2 3 4 5 6
1. I can understand new mathematics topics when
¡ ¡ ¡ ¡ ¡ ¡
taught.
2. I can plot a graph of a quadratic function (e.g., y
¡ ¡ ¡ ¡ ¡ ¡
= ax2 + b)
3. I can calculate the arc length of a circle. ¡ ¡ ¡ ¡ ¡ ¡
4. I can put in greater efforts to get better results in
¡ ¡ ¡ ¡ ¡ ¡
math.
5. I can calculate the gradient of a linear graph. ¡ ¡ ¡ ¡ ¡ ¡
6. I can differentiate when to add probabilities and
¡ ¡ ¡ ¡ ¡ ¡
when to multiply probabilities.
7. When I am stuck on a mathematics problem, I
¡ ¡ ¡ ¡ ¡ ¡
can solve it by asking people for help.
8. I can plot a graph for an exponential function
¡ ¡ ¡ ¡ ¡ ¡
(e.g., y = kax) and find x when y = 4.
9. I can use the mean and standard deviation to
¡ ¡ ¡ ¡ ¡ ¡
compare two sets of data.
10. When I am studying for a mathematics test, I
can plan a study schedule and finish revising all the ¡ ¡ ¡ ¡ ¡ ¡
topics before the test.
11. I can do 50% of my mathematics homework on
¡ ¡ ¡ ¡ ¡ ¡
my own.
12. I can do 100% of my mathematics homework
¡ ¡ ¡ ¡ ¡ ¡
on my own.
13. I can tell when two events are mutually
¡ ¡ ¡ ¡ ¡ ¡
exclusive and how that will affect probability.
14. I can solve simple inequality (e.g., 3x ≤ !).
¡ ¡ ¡ ¡ ¡ ¡
Note: ! represents a number.
15. I can determine whether two triangles are
¡ ¡ ¡ ¡ ¡ ¡
congruent or similar.
16. I can calculate or identify the mean, mode, and
¡ ¡ ¡ ¡ ¡ ¡
median for a set of data.
17. I can get an “A” in the next mathematics class
¡ ¡ ¡ ¡ ¡ ¡
test.
! !!!
18. I can solve fractional equations (e.g., + =
! ! ¡ ¡ ¡ ¡ ¡ ¡
!) that can be reduced to linear equations.
 50

Note: ! represents a number.

19. I can pass the next mathematics class test. ¡ ¡ ¡ ¡ ¡ ¡
20. I can calculate the area of a segment. ¡ ¡ ¡ ¡ ¡ ¡
21. I can calculate the standard deviation for a set of
¡ ¡ ¡ ¡ ¡ ¡
data.
22. I can list two properties of similar polygons. ¡ ¡ ¡ ¡ ¡ ¡
23. When I do badly in a class test I can learn from
¡ ¡ ¡ ¡ ¡ ¡
my mistakes and try harder in the next test.
24. I can solve quadratic equations by completing
¡ ¡ ¡ ¡ ¡ ¡
the square.
25. I can interpret stem-and-leaf diagrams. ¡ ¡ ¡ ¡ ¡ ¡
26. I can calculate the volume and surface area of a
¡ ¡ ¡ ¡ ¡ ¡
sphere.
27. I can reduce a plane figure by a scale factor. ¡ ¡ ¡ ¡ ¡ ¡
28. I can calculate probability by listing all the
¡ ¡ ¡ ¡ ¡ ¡
possible outcomes in a simple chance situation.
29. When I am studying for a mathematics test, I
can tell which are the concepts or skills I am weak ¡ ¡ ¡ ¡ ¡ ¡
at.
!
30. I can solve fractional equations (e.g., !!! = x +
!) that can be reduced to quadratic equations. ¡ ¡ ¡ ¡ ¡ ¡
Note: ! represents a number.


 51

Appendix D
Motivated Strategies for Learning Questionnaires (MSLQ) – Self-Efficacy
Students learn mathematics differently. Based on your performance in math class, read each
statement and decide how much you agree. Fill in the bubble of your corresponding choice.
There are no right or wrong responses, so pick the first response that comes to mind. Please
answer every question.

Not Very
at all true
true of
of me me
1 2 3 4 5 6 7
1. I believe I will receive an excellent
¡ ¡ ¡ ¡ ¡ ¡ ¡
grade in this class.
2. I’m certain I can understand the most
¡ ¡ ¡ ¡ ¡ ¡ ¡
difficult material present in this course.
3. I’m confident I can understand the
¡ ¡ ¡ ¡ ¡ ¡ ¡
basic concepts taught in this course.
4. I’m confident I can understand the
most complex materials presented by ¡ ¡ ¡ ¡ ¡ ¡ ¡
the instructor in this course.
5. I’m confident I can do an excellent
job on the assignments and tests in this ¡ ¡ ¡ ¡ ¡ ¡ ¡
course.
6. I expect to do well in this class. ¡ ¡ ¡ ¡ ¡ ¡ ¡
7. I’m certain I can master the skills
¡ ¡ ¡ ¡ ¡ ¡ ¡
being taught in this class.
8. Considering the difficulty of this
course, the teacher, and my skills, I ¡ ¡ ¡ ¡ ¡ ¡ ¡
think I will do well in this class.


 52

Appendix E
Background Information
The following questions are important in helping researchers understand how students’
backgrounds relate to their attitude and approach towards mathematics. There are no right or
wrong responses. Your responses will not be used to personally identify you.

1. Age ____________ years ____________ months

2. Gender ¡ ¡
Female Male

3. Academic Stream ¡ ¡ ¡ ¡
Special Express Normal Normal
Academic Technical

4. PSLE Score __________________ (out of 300)

5. PSLE Math Grade ¡ ¡ ¡ ¡ ¡

A* A B C D

6. Mathematics subjects taken in

¡ ¡
Secondary 3
Elementary Mathematics Elementary and
Additional Mathematics
7. Secondary 3 End of the Year
Overall Percentage (or Expected ¡ ¡ ¡ ¡ ¡ ¡ ¡
Percentage)
90- 80- 70- 60- 50- 41- <41
100% 89% 79% 69% 59% 49% %

8. Secondary 3 End of the Year Elementary Mathematics Grade (or Expected Grade)
¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡
A1 A2 B3 B4 C5 C6 D7 E8 F9

9. Secondary 3 End of the Year Additional Mathematics Grade (or Expected Grade)
¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡ ¡
A1 A2 B3 B4 C5 C6 D7 E8 F9

10. Any comments or feedback on this survey?