Ahmed (2018)

Journal of Adolescence 67 (2018) 158–166
Contents lists available at ScienceDirect
Journal of Adolescence
journal homepage: www.elsevier.com/locate/adolescence
T
Developmental trajectories of math anxiety during adolescence:
Associations with STEM career choice
Wondimu Ahmed∗
Department of Educational Foundations and Leadership, The University of Akron, USA
A R T IC LE I N F O ABS TRA CT
Keywords: The purpose of this study was to determine if there were distinct developmental trajectories of
Adolescence math anxiety during adolescence and if these trajectories predicted later STEM career choice. The
Math anxiety study also evaluated whether the trajectories varied in relation to gender and race/ethnicity.
Developmental trajectories Data were drawn from Longitudinal Study of American Youth-younger cohort that consisted of
STEM career
3116 nationally representative sample of 7th grade students (48% Female, 70% European
American) that were followed for 7 years. The results revealed four heterogeneous math anxiety
trajectory groups: consistently low (34.68%), decreasing (23.72%), increasing (21.90%), and con-
sistently high (20.12%). Trajectories varied with regard to race/ethnicity but not gender.
Membership in consistently low or decreasing trajectory predicted later STEM career choice. The
findings suggest that preventive interventions may benefit from targeting math anxiety during
adolescence.
Math anxiety is defined as “a feeling of tension and anxiety that interferes with the manipulation of numbers and the solving of
mathematical problems in a wide variety of ordinary life and academic situations” (Richardson & Suinn, 1972, p. 551). Math anxiety
is one of the most prevalent types of academic anxieties in school (see Cassady, 2010). For instance, the 2012 Program for Inter-
national Student Assessment report (Organization for Economic Co-operation and Development [OECD], 2015) indicated that on
average about 30% of 15-year students across 65 nations experience high levels of math anxiety. Math anxiety is negatively asso-
ciated with math performance (e.g., Ashcraft & Moore, 2009; Hembree, 1990; Lee, 2009; Ma, 1999). In addition to its negative
cognitive consequences, math anxiety can lead to avoidance of math and related tasks (Ashcraft, Krause, & Hopko, 2007). Indeed,
evidence shows that math anxiety is positively associated with avoidance of math courses thereby influencing individuals' career
pursuit (Hembree, 1990). These patterns are acutely troubling given the potential of math anxiety to deplete the Science, Technology,
Engineering and Mathematics (STEM) pipeline in the U.S. and elsewhere (Beilock & Maloney, 2015; Foley et al., 2017). Despite the
prevalence and its debilitating effects on math performance and its potential negative impact on a nation's resource base in STEM, we
know little about the development of math anxiety during adolescence— a developmental period during which a more realistic
occupational identity emerges (Porfeli & Lee, 2012). Moreover, we know little about how the development of math anxiety relates to
core demographic variables such as gender and race/ethnicity— variables that play a key role in math participation (Catsambis,
1994). More importantly, little is known about how math anxiety during adolescence is associated with later adult outcomes such as
STEM career choice. A better understanding of the development of math anxiety during adolescence is essential in providing targets
for prevention intervention to alleviate math anxiety and enhance youth's interest in STEM— an ongoing national effort (Morrell &
Parker, 2015).
This study attempted to address the aforementioned gaps in the literature. To this end, the study had three objectives. The first
∗
The University of Akron, Department of Educational Foundations and Leadership, Zook Hall 421, Akron, OH, USA.
E-mail address: wahmed@uakron.edu.
https://doi.org/10.1016/j.adolescence.2018.06.010
Available online 02 July 2018

Received 12 August 2017; Received in revised form 21 June 2018; Accepted 22 June 2018
0140-1971/ © 2018 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.
10959254, 2018, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1016/j.adolescence.2018.06.010 by Universite De Sherbrooke, Wiley Online Library on [09/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
W. Ahmed Journal of Adolescence 67 (2018) 158–166
objective was to investigate the developmental trajectories of math anxiety in a large sample of adolescents from 7th to 12th grade
using a person-centered approach (Laursen & Hoff, 2006) based on socio-cognitive models of math anxiety (e.g., Ashcraft et al., 2007;
Beilock, Schaeffer, & Rozek, 2017). These models hold that math anxiety emerges as a result of complex person-by-situation
transaction. Thus, the models stress the importance of both personal (e.g., low self-confidence) and environmental (e.g., negative
math classroom climate) factors in the development and maintenance of math anxiety (Ashcraft et al., 2007; Beilock et al., 2017;
Newstead, 1998). Given the diversity of individual characteristics and the dynamic nature of social and learning environments
students encounter in school, there might exist diverse developmental trajectories of math anxiety (Ashcraft et al., 2007).
Although there is an extensive body of literature on the nature and cognitive consequences of math anxiety (for reviews see
Ashcraft & Moore, 2009; Dowker, Sarkar, & Looi, 2016; Suárez-Pellicioni, Núñez-Peña, & Colomé, 2016), we know little about the
developmental changes in math anxiety. Indeed, there is a widespread claim that math anxiety increases overtime (see Dowker et al.,
2016). Nevertheless, to date, to the best of the researcher's knowledge, only three longitudinal studies have examined changes in
math anxiety. In the first study, Wigfield and Eccles (1989) investigated changes in math anxiety over two years in a sample of early
adolescents transitioning to junior high school. In this study, the authors found a decrease in math anxiety from grade 6 (elementary
school) to grade 7 (junior high school). In the second study, Ahmed, van der Werf, Minnaert and Kuyper (2013) investigated the
changes in math anxiety at three measurement occasions over a year among a sample grade 7 students. The authors found that math
anxiety was fairly stable over time. In the third study, Madjar, Zalsman, Weizman, Lev-Ran and Shoval (2016) examined changes in
math anxiety from beginning of sixth grade to the end of seventh grade, with two measurement occasions in between, and found no
change in math anxiety overtime. Such inconsistencies in the previous research along with the often used group-based approach to
identify distinct groups of math anxious individuals as low, medium and high based on arbitrary cut of scores on math anxiety
measures (Ashcraft & Moore, 2009) suggest that there may be clusters of individuals following distinct developmental trajectories.
To more fully understand diverse developmental trajectories of math anxiety, it is imperative to examine group differences in the
trajectories. Thus, the second objective of this study was to examine whether and how two core demographic variables: gender and
race/ethnicity are associated with trajectories of math anxiety. The extant literature suggests that students' levels of math anxiety
may vary across gender and racial/ethnic groups. An extensive body of literature indicates that females experience higher levels of
math anxiety than males (see Dowker et al., 2016; Else-Quest, Hyde, & Linn, 2010; Hembree, 1990). For instance, in their meta-
analysis of gender differences in math achievement and math attitude, Else-Quest et al. (2010) found that despite very small gender
differences in mathematics performance, girls reported significantly higher math anxiety than boys. Evidence regarding racial/ethnic
differences in math anxiety, albeit limited, shows interesting diversity. For instance, Hembree (1990) meta-analysis found that
whereas Hispanics reported higher levels of math anxiety than Whites, there were no significant differences between Blacks and
Whites. In a recent study that utilized a nationally representative sample of adolescents, Hispanics reported higher levels of math
anxiety than their White counterparts but Blacks did not differ from Whites (Cheema & Sheridan, 2015). Despite such group var-
iations in levels of math anxiety, no study has examined whether gender and race/ethnicity are associated with adolescents having
distinct developmental trajectories. Thus, the current study investigated whether and how math anxiety trajectories varied in relation
to gender and race/ethnicity. To avoid possible confounding effect of socioeconomic status (SES), the current study controlled for
parental education—a stable indicator of SES (Sirin, 2005). Understanding how the development of math anxiety varies as a function
of sociodemographic characteristics is essential for advances in math anxiety etiology and prevention.
The third objective of this study was to assess the extent to which trajectories of math anxiety during adolescence are associated
with later STEM career choice. Theoretically, math anxiety is posited to lead to avoidance of math and related courses and careers
(Ashcraft, 2002; Ashcraft et al., 2007). Although a considerable amount of literature has been published on math anxiety and its
relationship with math performance, only limited numbers of studies have examined the relations between math anxiety and be-
havioral intentions such as intent to take more math courses. Hembree (1990) meta-analysis of a handful of studies showed that math
anxiety is negatively associated with intent to take more math courses in high school and beyond. Based on such studies, scholars
claim that math anxiety steers individuals away from STEM careers (Ashcraft, 2002; Beilock & Maloney, 2015). Yet, to the best of the
researcher's knowledge no longitudinal study to date has examined the relationship between math anxiety and STEM career at-
tainment. Whereas previous studies have certainly added to our knowledge of the negative impact of math anxiety on intentions to
take more or advanced math courses, we know little about whether a high and stable math anxiety during adolescence is associated
with lesser likelihood of choosing STEM careers during adulthood. The current study attempted to fill this gap by examining the
associations between math anxiety trajectory membership and STEM career choice. Because evidence shows that parents with STEM
occupations influence their children's college major choice in STEM (Moakler & Kim, 2014), parental STEM occupation was taken
into account when predicting STEM career choice by math anxiety trajectories.
1. The present study
The objectives of this study were:1) to identify groups of individuals who follow distinct development trajectories of math anxiety,
2) to examine if gender and race/ethnicity differentiate math anxiety trajectory groups and 3) to examine the relationship between
membership in math anxiety trajectory groups and STEM career choice. With regard to the first objective, because no previous study
has examined math anxiety trajectory clusters, no specific hypothesis was formulated regarding the specific number of expected
trajectory groups. Nevertheless, based on previous cross-sectional (see Dowker et al., 2016; Hembree, 1990) and longitudinal
(Ahmed, van der Werf, Kuyper, & Minnaert, 2013; Madjar, Zalsman, Weizman, Lev‐Ran, & Shoval, 2016; Wigfield & Eccles, 1989)
studies as well as existing empirical approaches to the study and classification of math anxiety (Ashcraft & Moore, 2009; Ashcraft
et al., 2007), four trajectories were expected: decreasing, increasing, low and high trajectories. Concerning the second objective,
159
based on the literature on gender differences favoring males (e.g., Else-Quest et al., 2010), females would be more likely than males
to belong to maladaptive math trajectory groups. Again based on the literature on racial/ethnic differences in math anxiety (e.g.,
Hembree, 1990), Hispanic adolescents would be more likely than White adolescents to belong to maladaptive math trajectory groups
but Blacks would be as likely as Whites to belong to any trajectory group. Finally, based on prior research on math anxiety and math
course taking intentions (e.g., Heembree, 1990), for the last objective, it was hypothesized that membership in adaptive math
trajectory groups during adolescence would be associated with higher likelihood of later employment in STEM occupation.
2. Methods
2.1. Participants
The data used in the current study were drawn from Longitudinal Study of American Youth ([LSAY] Miller, 2014). The LSAY
followed two cohorts of students (i.e., grade 7 and grade 10 cohorts) beyond high school. Because the objective of this study was to
identify developmental trajectories of math anxiety during adolescence, data for the present study were from the grade 7 cohort. This
cohort consisted of a nationally representative sample of 3116 seventh grade students (60 students randomly selected from 52 middle
schools across eight strata) who were followed up for an initial period of seven years (Miller, 2014). The seventh wave was conducted
one year after the end of high school. The sample approximately consisted of 48% females, 70% European American, 9% Hispanic,
11% African American, 3% Asian American, and 1% Native American. Sixty-nine percent of students in the sample had at least one
parent who did not complete college. The main focus of LSAY was on math and science education. Each year starting in 1987 students
completed a number of questionnaires and took achievement test in both math and science. In 2007, more than 95% of the original
sample (for both cohorts) were located and 78% of the respondents completed a survey regarding their education and occupation
when they were between 33 and 37 years old (Kimmel & Miller, 2008; Miller, 2014).
2.2. Measures
The LSAY used many standard demographic questions about the youth and their families including gender, race/ethnicity,
parental education and whether parents were employed in a STEM career. In addition to the demographic measures, LSAY used,
among others, measures that pertain to assessment of math anxiety during the first 6 waves as well as occupational choices of the
participants in 2007.
2.2.1. Math anxiety

Math anxiety was measured using two items. Example item: “Doing math makes me nervous or upset”. Students were asked to
rate the items on a 5-point Likert type scale ranging from “strongly agree (1)” to “strongly disagree (5)”. The items were reverse
scored so that high scores indicate high math anxiety. The scale presents satisfactory reliability estimates (αs = .63 - .76). The intra-
class correlation coefficient (ICC) was .47 suggesting the presence of both individual differences as well as time specific fluctuations
in math anxiety. The average of the two items was used to indicate the participants' level of math anxiety.
2.2.2. STEM occupation

In 2007, respondents were asked about their occupation. LSAY staff coded the jobs categories into a dichotomous variable
indicating whether a respondent was currently employed in a STEM occupation (physical sciences, medical science, life science,
computer science, mathematics, or engineering) or not. For the current study, STEM occupations were coded “1” and non-STEM
occupations were coded “0”.
2.2.3. Demographic variables

Four important sociodemographic variables available in LSAY dataset were used in the predictive analyses. To predict trajectory
class membership, gender (male = 0; female = 1) and race/ethnicity (Black = 1; Hispanic = 2; White = 0) were employed. In this
same model, parental education (1 = high school or less, 0 = some college or higher) was used as a control variable. To predict STEM
career choice by math anxiety trajectories, gender (male = 0; female = 1), race/ethnicity (Black = 1; Hispanic = 2; White/
Asian = 0) and parental employment in STEM occupation (1 = neither parent worked in technical or professional STEM occupations,
0 = either parent worked in technical or professional STEM profession) were controlled.
2.3. Data analytic strategy
To identify optimal number of math anxiety trajectory classes, k-means for longitudinal data (KmL) statistical package (Genolini &
Falissard, 2011; Genolini, Alacoque, Sentenac, & Arnaud, 2015) as implemented in R (R Core Team, 2013) was used. KmL was
selected because evidence from Monte Carlo simulations (McLachlan, 2011; Steinley & Brusco, 2011) suggest that cluster analysis
performs as well as or better than mixture modeling methods particularly when the number of clusters is unknown. KmL uses a
special type of expectation-maximization algorithm in which, first each observation is assigned to a trajectory cluster, and then, the
seeds of each trajectory cluster are computed based on which each observation is reassigned to the closest trajectory class. This
procedure is repeated until convergence (Genolini et al., 2015). To avoid local maxima, this algorithm is run many times (1000 times
in the current study). To determine the best solution, several criteria were used including: Calinksi and Harabatz (CH), Davies and
160
Boudin (BD), Ray and Turi (RT) and Bayesian Information criterion (BIC). Whereas BIC is based on log liklihood function, the CH is a
relative metric that compares the different group solutions based on the ratio of the between-group variance to the within-group
variance (Genolini & Falissard, 2011; Genolini et al., 2015). DB is the ratio of the within cluster scatter to the between cluster
separation and TR is the ratio of average intra-cluster distance (i.e., distance of a point from the centroid) to inter-cluster distance
(Genolini et al., 2015). Thus, lower values of BIC, DB and TR but higher values of CH indicate a better solution. In addition to these
indexes, the investigator relied on cluster interpretability and substantive theory. Although KmL does not make any assumptions
regarding the shape of trajectories, once the trajectories were identified growth curve models were fit separately to each group to
describe the trajectory groups using MLwIN 2.33 (Rasbash, Steele, Browne, & Goldstein, 2015). After the trajectory groups were
identified, multinomial logistic regression was used to examine if gender and race/ethnicity predicted trajectory group membership
after controlling for parental education. Then, binary logistic regression was used to examine the associations between trajectory
group membership and STEM career choice. The major concern with long-term longitudinal studies such as LSAY is attrition. Because
KmL is capable of handling intermittent missing data as long as data are available at two-time points (Genolini & Falissard, 2011),
cases with missing values for math anxiety on five or 6 waves were excluded from the analysis. This reduced the analytic sample to
2912. The final analytic sample comprised of almost identical demographic groups to the original sample: 48% females and 23%
minority. Little's (Little & Rubin, 2002) Missing Completely At Random (MCAR) test on the analytic sample indicated that data are
not missing completely at random (χ2 = 212.24, df = 179, p = .045). To handle the intermittent missing data, KmL uses an approach
that combines linear interpolation with trajectory mean imputation (Genolini & Falissard, 2011; Genolini, Écochard, & Jacqmin-
Gadda, 2013). This approach has been shown to efficient in reducing bias in longitudinal data analysis under three missing data
mechanisms: missing completely at random, missing at random, missing not at random (Genolini & Falissard, 2011; Genolini et al.,
2013). The predictive analyses were conducted using SPSS Complex Sample procedure to account for sampling and attrition. Thus,
appropriate weights available in the LSAY dataset were utilized.
3. Results
The purpose of this study was threefold. First, trajectories of math anxiety were identified. Second, multi-nominal logistic re-
gression was used to determine if three important demographic variables were associated with trajectory membership. Third, binary
logistic regression was run to determine whether trajectories of math anxiety predicted STEM career attainment. The results are
presented in such an order.
3.1. Math anxiety trajectory classes
Several successive models, with three to six classes were tested. Table 1 presents the four indexes used to evaluate the validity of
the clustering along with the proportion of each class. Whereas CH, DB and RT favored a four-cluster solution, BIC favored a six-cluster
solution. Inspection of the trajectories of the six-cluster solution indicated that two of the additional trajectories split the low and
decreasing groups into additional low and decreasing classes thereby reducing the heuristic value of the classes. The second best
lowest BIC score was for the four-cluster solution. Thus, the four-cluster solution was chosen as an optimal solution. As noted above,
growth curve analyses were run for each of the classes to describe the shape of the trajectories. The four math anxiety trajectory
classes are illustrated in Fig. 1. These four classes1 are labeled: Consistently Low (34.68%), Decreasing (23.72%), Increasing (21.9%),
and Consistently High (20.12%). The Consistently Low class includes adolescents presenting low levels of math anxiety (Inter-
cept = 1.71) that appears to remain stable over the course of 6 years (Slope = .03, p = .11). The Decreasing group includes adolescents
presenting high levels of math anxiety at the beginning of the study period (Intercept = 3.41) with decreasing trajectory
(Slope = −.23, p < .01). The most distinctive class is the Increasing group. Adolescents in this group appear to present low levels of
math anxiety at the beginning of the study period (Intercept = 1.96) similar to the Consistently Low group with significant change
overtime (Slope = .26, p < .01). At the end of the study period, this group had already caught up with the Consistently High group.
The last group, Consistently High, includes adolescents presenting high levels of math anxiety at the beginning of the study period and
with stable trajectories (Intercept = 3.51; Slope = −.02, p = .26). The results show significant qualitative differences between classes.
3.2. Predicting math anxiety trajectory class membership
Table 2 presents the results from multi-nominal logistic regression analysis predicting trajectory group memberships. The multi-
nominal logistic regression analysis was conducted with gender, race/ethnicity and parental education as joint predictors of tra-
jectory membership. The odds ratios refer to the likelihood of membership in target group compared to the reference class (i.e.,
Consistently Low trajectory group). Contrary to the expectation, the results indicated that female youth were as likely as male youth to
belong to either Increasing or Consistently High group. With regard to racial/ethnic groups, both Black youth and Hispanic youth were
more likely than White youth to belong to both Increasing and Consistently High groups. Concerning parental education, youth born to
parents with high school or less education level were more likely to belong to both either Increasing or Consistently High math anxiety
trajectory groups.
1
The percentages in the parentheses are the actual percentages of the sample that fell into those trajectory classes.
161
Table 1
Math anxiety trajectory classes and associated clustering indexes.
Criteria Trajectory Classes
3 4 5 6
Calinski Harabatz 1430.38 1507.37 1481.95 1467.22

Davies Bouldin −1.59 −1.52 −1.61 −1.58
Ray Turi -.15 -.13 -.20 -.18
BIC −35677.17 −35629.08 −35661.69 −35575.25
Class 1 40.21% 34.68% 26.85% 22.90%
Class 2 33.58% 23.72% 20.29% 19.02%
Class 3 26.20% 21.46% 19.61% 16.27%
Class 4 20.12% 18.75% 15.41%
Class 5 14.49% 14.28%
Class 6 12.08%
Fig. 1. Trajectories of math anxiety.
Table 2
Results from multi-nominal logistic regression predicting trajectory group membership.
Trajectory groups B SE OR 95%CI
LL UL
Decreasing
Female -.10 .10 .90 .73 1.11
Black .00 .23 1.00 .63 1.58
Hispanic .05 .20 1.05 .71 1.55
LowParEduc .09 .13 1.09 .85 1.41
Increasing
Female -.18 .11 .83 .70 1.04
Black .25 .17 1.28 .91 1.80
Hispanic .10 .19 1.11 .75 1.64
LowParEduc .33 .13 1.38 1.07 1.79
Consistently High
Female -.12 .14 .88 .67 1.17
Black .46 .16 1.58 1.14 2.18
Hispanic 49 .22 1.64 1.04 2.57
LowParEduc .33 .12 1.40 1.09 1.79
Note: OR= Odds Ratio; CI = confidence interval; LL = lower limit; UL = upper limit; LowParentEduc = Low parental education (i.e., high school or
less); Reference group is Consistently Low group.
162
Table 3
Results from logistic regression predicting STEM career choice.
Variables Unadjusted Model Adjusted Model
95%CI 95%CI
B SE OR LL UL B SE OR LL UL
Consistently Low 2.00 .43 7.40 3.08 17.79 1.93 .44 6.89 2.81 16.91
Decreasing 1.86 .45 6.42 2.57 16.01 1.70 .47 5.46 2.10 14.18
Increasing .61 .41 1.82 .82 3.03 .54 .41 1.72 .77 2.83
Female −1.00 .33 .37 .19 .72
Black −1.13 .56 .32 .10 .99
Hispanic −1.06 .51 .35 .12 .98
NParSTEM -.59 .28 .55 .31 .98
Note: OR= Odds Ratio; CI = confidence interval; LL = lower limit; UL = upper limit; NParSTEM = neither parent employed in technical or pro-
fessional STEM occupation; Reference group is Consistently High group.
3.3. Math anxiety trajectory group membership and STEM career choice
The final objective of the study was to examine the relationship between math anxiety trajectory group membership and STEM
career choice. Binary logistic regression was performed to examine the contribution of math anxiety trajectories in predicting STEM
career choice (see Table 3). Table 3 presents both unadjusted and adjusted effects of trajectory membership on STEM career choice.
As indicated in the unadjusted model, adolescents in Consistently Low math anxiety class are about 7.4 times as likely to be employed
in STEM as those in Consistently High math anxiety group. Similarly, adolescents in the Decreasing math anxiety trajectory group are
6.4 times as likely to be employed in STEM as those in the Consistently High math anxiety trajectory group. The Increasing class was not
significantly different from that of Consistently High class in predicting STEM career choice. The results of the adjusted model, in
which gender, race/ethnicity and parental STEM occupation were taken into account still showed that the significance of class
membership remained the same although the coefficients were somewhat reduced. With regard to the control variables, female
youth, Black and Hispanic were less likely to be employed in STEM. Similarly, not having a parent who worked in STEM professional
or technical support occupations was associated with less likelihood of being employed in STEM.
4. Discussion
The purpose of this study was three-fold: 1) to identify developmental trajectories of math anxiety, 2) to examine whether gender
and race/ethnicity, after adjusting for parental education, are associated with math anxiety trajectory classes, and 3) to examine if
membership in math anxiety trajectory classes predicted STEM career choice, above and beyond important control variables. The
results demonstrated that four distinct trajectories of math anxiety could be identified. The results also indicated that memberships in
math anxiety trajectory groups are associated with gender and race/ethnicity as well as with STEM career choice. Given these
findings, the current study contributes to the literature on the development of math anxiety in three important ways.
First, despite the prevailing belief that math anxiety generally increases overtime (Dowker et al., 2016; Hembree, 1990), the
results of this study indicated that there exists heterogeneity in the developmental patterns of math anxiety during adolescence.
Although math anxiety increases overtime for a significant number of adolescents in the current sample (i.e., about 22%), a much
larger portion of adolescents follow diverse developmental trajectories. For instance, whereas about 35% of adolescents belong to a
Consistently Low math anxiety group, about 20% of them belong to a Consistently High math anxiety trajectory group and about 24%
belongs to a Decreasing math anxiety group. For Consistently Low and Consistently High groups, math anxiety levels appear to be stable
overtime suggesting that for some adolescents, math anxiety does not change. This is consistent with previous studies that found math
anxiety to be stable overtime (Ahmed et al., 2013; Madjar et al., 2016). However, the findings also suggest that stability of math
anxiety may not characterize everyone, because about 46% of the current sample show either a decreasing or an increasing pattern of
development. The presence of these qualitatively distinct developmental trajectories of math anxiety suggests that one size may not
fit all.
Second, the study identified several instances of gender and race/ethnicity differences and similarities with regard to membership
in math anxiety trajectory groups. Whereas previous research typically found that females report higher levels of math anxiety than
males (e.g., Else-Quest et al., 2010; Hembree, 1990), in the current study, gender did not predict membership in the Consistently High
math anxiety trajectory group (i.e., compared to the reference category: Consistently Low) suggesting that both male and female
adolescents have similar likelihood of membership in this group. The findings also showed that females are as likely as males to
belong to either the Decreasing or the Increasing trajectory groups (i.e., compared to the reference category: Consistently Low). These
finding are consistent with some cross-sectional studies that found no gender differences in math anxiety (e.g., Birgin, Baloğlu,
Çatlıoğlu, & Gürbüz,2017; Chiu & Henry, 1990; Newstead, 1998). Overall, given the large sample size and considerable statistical
power, these findings suggest gender similarity in terms of the development of math anxiety.
The associations between race/ethnicity and membership in math anxiety trajectory groups is more complex. Contrary to the
expectations, Black adolescents tend to belong the Consistently High math anxiety trajectory group compared to their White counter
163
parts. Previous cross-sectional studies have consistently indicated that there are no significant Black-White differences in levels of
math anxiety (e.g., Cheema & Sheridan, 2015; Hembree, 1990). Perhaps the inconsistency with previous research is due to the fact
that prior studies mostly were cross-sectional by design, which makes difficult to identify distinct developmental path ways that Black
students might follow. As hypothesized and consistent with previous cross-sectional studies (e.g., Cheema & Sheridan, 2015;
Hembree, 1990), Hispanic adolescents also tend to belong the Consistently High math anxiety. It is important to note that these racial/
ethnic differences and similarities remained significant even after taking parental education into account. Given recent evidence that
the impact of math anxiety may be stronger among racial/ethnic minorities such as African American (Young & Young, 2015) these
findings suggest that the prevailing belief in Black-White similarity in math anxiety may carry substantial costs. One reason why
Blacks and Hispanics are more likely to be members of the less favorable math anxiety trajectory group may be due to stereotype
threat. Because racial/ethnic minorities including Blacks and Hispanics endure negative stereotypes regarding their math ability (e.g.,
Cvencek, Nasir, O'connor, Wischnia, & Meltzoff, 2015) they may experience stereotype threat—fear and apprehension that they will
confirm the negative stereotype; and consequently, experience elevated levels of anxiety in a domain such as math (see Spencer,
Logel, & Davies, 2016). Negative stereotypes about a social group's ability in a specific domain negatively impact individual's self-
concept of ability in the domain (Lewis & Sekaquaptewa, 2016) thereby leading to negative emotional experiences including anxiety.
Another possible reason is that Black and Hispanic adolescents may not have the opportunity to learn math— which includes quality
teaching and learning experience. Indeed, evidence shows that lack of opportunity to learn formal math is an important correlate of
math anxiety (Schmidt, Zoido, & Cogan, 2013). Opportunity to learn math is also is associated with SES (Schmidt, Burroughs, Zoido,
& Houang, 2015). The findings of the current study showed that youth from less educated families tend to belong to either increasing
and consistently high math anxiety trajectory groups. This finding advances our knowledge about importance of SES in math anxiety
which is largely lacking in the literature (see Dowker et al., 2016).
Third, the study examined the contributions of adolescents' math anxiety trajectories to their later STEM career choice. As
hypothesized, students in the Consistently Low math anxiety trajectory group were more likely than students in either Increasing or
Consistently High group to be employed in STEM as adults. This finding extends previous research on math anxiety and career
aspirations (e.g., Hembree, 1990) by demonstrating that consistently high levels of anxiety or increasing levels of math anxiety
during adolescence may be one reason why many college students turn away from STEM majors and occupations. The effect of the
maladaptive math anxiety trajectories remained significant even after several demographic variables were taken into account. To our
knowledge, this is the first study to examine longitudinal links between math anxiety and STEM career attainment. Math anxious
students are more likely to avoid math-related courses and activities (Ashcraft, 2002). Unfortunately, this avoidance behavior is likely
to have a snowballing effect— avoidance leads to less skill development which erodes confidence thereby increasing anxiety. This
process may be one mechanism through which high math anxiety is maintained overtime.
Taken together the findings of this study have important implications for theory and practice. This study, to our knowledge, is the
first to examine diverse developmental trajectories of math anxiety during adolescence. The fact that there are distinct trajectory
groups of math anxiety provides important new insights into its development. This finding also implies that educators and researchers
should be aware of the developmental heterogeneity of math anxiety. This type of knowledge is important in the design and de-
velopment of prevention or intervention programs. This research also extends our knowledge about the importance of core demo-
graphic variables in the development of math anxiety which is largely lacking in the literature (Dowker et al., 2016). Contrary to
previous cross-sectional research, the results indicated gender similarity and Black-White difference in terms of membership in the
Consistently High trajectory group. Moreover, this study is the first to investigate the associations between the developmental tra-
jectories of math anxiety during adolescence to later STEM career attainment. The results indicated that experiencing consistently
high levels or increasing levels of math anxiety during adolescence appear to lead to avoidance of STEM careers. To increase STEM
participation much of the research to date has focused mainly on cognitive and motivational factors (Wang & Degol, 2013), the
findings of this study suggest that affective variables such as math anxiety should not be ignored (see also Beilock & Maloney, 2015).
The current study has some limitations that should be taken into account when interpreting the findings. First, although the study
identified distinct developmental trajectories of math anxiety, factors that led to these trajectories, beyond the demographic variables
considered in this study, are not known. Lack of perceived control over success in math may be an important vulnerability factor in
the development of emotions such as math anxiety (Pekrun, 2017). In addition, negative math classroom climate as well as exposure
to anxious role models at home and school may be source of math anxiety (Beilock et al., 2017). These and other personal and
environmental factors should be further explored to fully understand the development of math anxiety. Second, although math
anxiety was measured from grade 7 to grade 12 covering a good temporal window in the participants' life, the data were collected
between 1987 and 1992. Since the early 1990's a number of significant changes have occurred in the US math education (Woodward,
2004). Thus, only future research will ascertain if the distinct trajectories found in the current study hold among recent generation of
adolescents. In relation to this, the associations of gender and race/ethnicity with math anxiety might have changed overtime. Third,
although this study examined the link between math anxiety trajectories and STEM career choices, it did not control for related
variables including SES, test anxiety and math performance. Future research should investigate how math anxiety trajectories relate
to subsequent STEM career choices after controlling for SES and math performance. Fourth, though the study examined the potential
impact of maladaptive math anxiety trajectories on later STEM career attainment, mechanisms through which this effect operates is
unclear. Indeed, scholars have speculated that math anxiety results in avoidance of and decreases in interest in math-related activities
(Ashcraft et al., 2007). Future research is needed on these and other motivational mechanisms to fully understand how math anxiety
influences STEM career choice.
164
5. Conclusion
This study identified diverse developmental trajectories of math anxiety during adolescence that have important implications for
later STEM career choice. The findings showed that having consistently high levels of math anxiety or increasing levels of math
anxiety during adolescence is likely to lead to avoidance STEM careers. The findings also showed that racial/ethnic minorities are
more likely to experience maladaptive pathways of math anxiety. Overall, the findings suggest that any initiative that aims to
increase diverse STEM workforce should not ignore math anxiety.
References
Ahmed, W., van der Werf, G., Kuyper, H., & Minnaert, A. (2013). Emotions, self- regulated learning, and achievement in mathematics: A growth curve analysis. Journal
of Educational Psychology, 105(1), 150. https://doi.org/10.1037/a0030160.
Ashcraft, M. H. (2002). Math anxiety: Personal, educational, and cognitive consequences. Current Directions in Psychological Science, 11, 181–185. https://doi.org/10.
1111/1467-8721.00196.
Ashcraft, M. H., Krause, J. A., & Hopko, D. R. (2007). Is math anxiety a mathematical learning disability? In D. B. Berch, & M. M. M. Mazzocco (Eds.). Why is math so
hard for some children? The nature and origins of mathematical learning difficulties and disabilities (pp. 329–348). Baltimore, MD: Brookes.
Ashcraft, M. H., & Moore, A. M. (2009). Math anxiety and the affective drop in performance. Journal of Psychoeducational Assessment, 27, 197–205. https://doi.org/10.
1177/0734282908330580.
Beilock, S. L., & Maloney, E. A. (2015). Math anxiety: A factor in math achievement not to be ignored. Policy Insights from the Behavioral and Brain Sciences, 2(1), 4–12.
https://doi.org/10.1177/2372732215601438.
Beilock, S. L., Schaeffer, M. W., & Rozek, C. S. (2017). Understanding and addressing performance anxiety. Handbook of competence and Motivation: Theory and application.
155.
Birgin O, Baloğlu M, Çatlıoğlu H, Gürbüz R. An investigation of mathematics anxiety among sixth through eighth grade students in Turkey. Learning and Individual
Differences, 20, 654–658. doi: 10.1016/j.lindif.2010.04.006.
Cassady, J. C. (2010). Anxiety in schools: The causes, consequences and solutions for academic anxieties. New York, NY: Peter Lang.
Catsambis, S. (1994). The path to math: Gender and racial-ethnic differences in mathematics participation from middle school to high school. Sociology of Education,
199–215. https://doi.org/10.2307/2112791.
Cheema, J. R., & Sheridan, K. (2015). Time spent on homework, mathematics anxiety and mathematics achievement: Evidence from a US sample. Issues in Educational
Research, 25(3), 246–259.
Chiu, L. H., & Henry, L. L. (1990). Development and validation of the mathematics anxiety scale for children. Measurement and Evaluation in Counseling and
Development, 23, 121–127. https://doi.org/10.1177/0748175612459198.
Cvencek, D., Nasir, N. I. S., O'connor, K., Wischnia, S., & Meltzoff, A. N. (2015). The development of math–race stereotypes: “They say Chinese people are the best at
math. Journal of Research on Adolescence, 25(4), 630–637. https://doi.org/10.1111/jora.12151.
Dowker, A., Sarkar, A., & Looi, C. Y. (2016). Mathematics anxiety: What have we learned in 60 years? Frontiers in Psychology, 7, 508. http://dx.doi.org/10.3389/fpsyg.
2016.00508.
Else-Quest, N. M., Hyde, J. S., & Linn, M. C. (2010). Cross-national patterns of gender differences in mathematics achievement, attitudes, & affect: A meta-analysis.
Psychological Bulletin, 136, 103–127. https://doi.org/10.1037/a0018053.
Foley, A. E., Herts, J. B., Borgonovi, F., Guerriero, S., Levine, S. C., & Beilock, S. L. (2017). The math anxiety-performance link: A global phenomenon. Current Directions
in Psychological Science, 26(1), 52–58. https://doi.org/10.1177/0963721416672463.
Genolini, C., Alacoque, X., Sentenac, M., & Arnaud, C. (2015). Kml and kml3d: R packages to cluster longitudinal data. Journal of Statistical Software, 65, 1–34.
Genolini, C., Écochard, R., & Jacqmin-Gadda, H. (2013). Copy mean: A new method to impute intermittent missing values in longitudinal studies. Open Journal of
Statistics, 3(04), 26. https://doi.org/10.4236/ojs.2013.34a004.
Genolini, C., & Falissard, B. (2011). Kml: K-means for longitudinal data. Computational Statistics, 25, 317–328. https://doi.org/10.1016/j.cmpb.2011.05.008.
Hembree, R. (1990). The nature, effects, and relief of mathematics anxiety. Journal for Research in Mathematics Education, 21(1), 33–46. https://doi.org/10.2307/
749455.
Kimmel, L. K., & Miller, J. D. (2008). The longitudinal study of American Youth: Notes on the first 20 years of tracking and data collection. Survey Practice, 1(5).
Laursen, B., & Hoff, E. (2006). Person-centered and variable-centered approaches to longitudinal data. Merrill-Palmer Quarterly, 52(3), 377–389. https://doi.org/10.
1353/mpq.2006.0029.
Lee, J. (2009). Universals and specifics of math self-concept, math self-efficacy, and math anxiety across 41 PISA 2003 participating countries. Learning and Individual
Differences, 19(3), 355–365. https://doi.org/10.1016/j.lindif.2008.10.009.
Lewis, N. A., & Sekaquaptewa, D. (2016). Beyond test performance: A broader view of stereotype threat. Current Opinion in Psychology, 11, 4–43. http://dx.doi.org/10.
1016/j.copsyc.2016.05.002.
Little, R. J. A., & Rubin, D. B. (2002). Statistical analysis with missing data (2nd ed. .). New York, NY: John Wiley & Sons.
Ma, X. (1999). A meta-analysis of the relationship between anxiety toward mathematics and achievement in mathematics. Journal for Research in Mathematics
Education, 30, 520–540. https://doi.org/10.2307/749772.
Madjar, N., Zalsman, G., Weizman, A., Lev‐Ran, S., & Shoval, G. (2016). Predictors of developing mathematics anxiety among middle‐school students: A 2‐year
prospective study. International Journal of Psychology. http://dx.doi.org/10.1002/ijop.12403.
McLachlan, G. J. (2011). Commentary on Steinley and Brusco (2011): Recommendations and cautions. Psychological Methods, 16, 80–81. http://dx.doi.org/10.1037/
a0021141.
Miller, J. D. (2014). Longitudinal study of american youth, 1987-1994, 2007-2011. Ann Arbor, MI: Inter-university Consortium for Political and Social Research. http://
doi.org/10.3886/ICPSR30263.v6.
Moakler, M. W., & Kim, M. M. (2014). College major choice in STEM: Revisiting confidence and demographic factors. The Career Development Quarterly, 62(2),
128–142. https://doi.org/10.1002/j.2161-0045.2014.00075.x.
Morrell, C., & Parker, N. C. (2015). Solving the education equation: A report developed for the multi-stakeholder coalition for building a diverse US STEM workforce.
Newstead, K. (1998). Aspects of children's mathematics anxiety. Educational Studies in Mathematics, 36(1), 53–71. https://doi.org/10.1023/a:1003177809664.
OECD (2015). The ABS of gender Equality in Education: Aptitude, behavior, confidence, PISA. OECD Publishinghttps://doi.org/10.1787/9789264229945-en.
Pekrun, R. (2017). Emotion and achievement during adolescence. Child Development Perspectives. http://dx.doi.org/10.1111/cdep.12237.
Porfeli, E. J., & Lee, B. (2012). Career development during childhood and adolescence. New Directions for Youth Development, 134, 11–22. https://doi.org/10.1002/yd.
20011.
R Core Team (2013). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. http://www.R-project.org/.
Rasbash, J., Steele, F., Browne, W. J., & Goldstein, H. (2015). A user's guide to MLwiN. Bristol, UK: Centre for Multilevel Modeling, University of BristolVersion 2.33.
Richardson, F. C., & Suinn, R. M. (1972). The mathematics anxiety rating scale: Psychometric data. Journal of Counseling Psychology, 19, 551–554.
Schmidt, W. H., Burroughs, N. A., Zoido, P., & Houang, R. T. (2015). The role of schooling in perpetuating educational inequality: An international perspective.
Educational Researcher, 44(7), 371–386. https://doi.org/10.3102/0013189x15603982.
Schmidt, W. H., Zoido, P., & Cogan, L. S. (2013). Schooling matters: Opportunity to learn in PISA 2012 (OECD Education Working Papers No. 95)Paris, France:
Organization for Economic Co-operation and Development.
165
Sirin, S. R. (2005). Socioeconomic status and academic achievement: A meta-analytic review of research. Review of Educational Research, 75(3), 417–453. https://doi.
org/10.3102/00346543075003417.
Spencer, S. J., Logel, C., & Davies, P. G. (2016). Stereotype threat. Annual Review of Psychology, 67, 415–437. https://doi.org/10.1146/annurev-psych-073115-103235.
Steinley, D., & Brusco, M. J. (2011). Evaluating mixture modeling for clustering: Recommendations and cautions. Psychological Methods, 16, 63–97. http://dx.doi.org/
10.1037/a0022673.
Suárez-Pellicioni, M., Núñez-Peña, M. I., & Colomé, À. (2016). Math anxiety: A review of its cognitive consequences, psychophysiological correlates, and brain bases.
Cognitive, Affective, & Behavioral Neuroscience, 16(1), 3–22. https://doi.org/10.3758/s13415-015-0370-7.
Wang, M. T., & Degol, J. (2013). Motivational pathways to STEM career choices: Using expectancy–value perspective to understand individual and gender differences
in STEM fields. Developmental Review, 33(4), 304–340. https://doi.org/10.1016/j.dr.2013.08.001.
Wigfield, A., & Eccles, J. S. (1989). Anxiety and worries about math and English before and after the junior high transition. Paper presented at the annual meeting of the
American educational research association, San Francisco, CA.
Woodward, J. (2004). Mathematics education in the United States: Past to present. Journal of Learning Disabilities, 37(1), 16–31.
Young, J. R., & Young, J. L. (2015). Anxious for answers: A meta-analysis of the effects of anxiety on African American K-12 students' mathematics achievement.
Journal of Mathematics Education at Teachers College, 6(2), 1–8.
166

Ahmed (2018)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ahmed (2018)

Uploaded by

Copyright:

Available Formats

Journal of Adolescence 67 (2018) 158–166

Contents lists available at ScienceDirect

Available online 02 July 2018

1. The present study

2.2.1. Math anxiety

2.2.2. STEM occupation

2.2.3. Demographic variables

2.3. Data analytic strategy

3.1. Math anxiety trajectory classes

3.2. Predicting math anxiety trajectory class membership

Calinski Harabatz 1430.38 1507.37 1481.95 1467.22

Fig. 1. Trajectories of math anxiety.

You might also like