You are on page 1of 11

See

discussions, stats, and author profiles for this publication at:


https://www.researchgate.net/publication/232549605

Role of Mathematics Self-Efficacy


in the Choice of Math-Related
Majors of College Women and
Men. A Path Analysis

Article in Journal of Counseling Psychology · January 1985


DOI: 10.1037/0022-0167.32.1.47

CITATIONS READS

274 1,267

1 author:

Gail Hackett
Virginia Commonwealth University
76 PUBLICATIONS 9,692 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related
projects:

As the provost,I am starting a new project now using Social cognitive


career theory to guide our interventions to improve retention and
graduation rates here at the VCU. View project

Psychotherapy and counseling with sexual and gender minorities View


project

All content following this page was uploaded by Gail Hackett on 25 June 2015.

The user has requested enhancement of the downloaded file.


Journal of Counseling Psychology Copyright 1985 by the American Psychological Association, Inc.
1985, Vol. 32, No. 1,47-56 0022-0167/85/$00.75

Role of Mathematics Self-Efficacy in the Choice of Math-


Related Majors of College Women and Men: A Path Analysis
Gail Hackett
Ohio State University

The purpose of the present study was to test the hypothesis that mathematics-
related self-efficacy mediates the effects of gender and mathematical prepara-
tion and achievement on math relatedness of college major choice. The re-
sponses of 117 undergraduates to a series of inventories and questionnaires
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

yielded seven variables descriptive of the math-related career choice process;


This document is copyrighted by the American Psychological Association or one of its allied publishers.

a causal model of the interrelationships of these variables was constructed


from predictions based on self-efficacy theory. A path analysis and conse-
quent refinement of the model resulted in a final path model that was con-
gruent with a self-efficacy approach to women's career development. Unex-
pected results and implications of the model are discussed.

The increasing realization on the part of role of cognitive-mediational factors, spe-


counseling psychologists that our knowledge cifically expectations of personal effective-
of women's vocational behavior and career ness or self-efficacy (Bandura, 1977,1982),
development is inadequate (Fitzgerald & in career choice and is already beginning to
Crites, 1980; Osipow, 1983; Vetter, 1978) has receive some empirical support. Betz and
been accompanied by a dramatic increase in Hackett (1981), in the first test of the self-
the empirical literature devoted to this topic. efficacy model, reported that career-related
Still missing, however, is a synthesizing efficacy expectations were related to the
framework, a relevant theoretical con- nature and range of career alternatives being
ceptualization of women's career develop- considered by women and men. They also
ment which is capable of integrating existing found that gender-related differences in ef-
knowledge and guiding future research and ficacy expectations, particularly lower self-
intervention efforts. efficacy expectations on the part of women,
In an attempt to fill this conceptual void, were strongly predictive of the traditionality
Hackett and Betz (1981) have outlined a of careers under consideration. Several
self-efficacy approach to women's career other studies investigating self-efficacy with
development. This approach stresses the regard to specific career-relevant behaviors,
such as professional-level career choice,
science career choice, and career decision-
This research was supported by a grant from Ohio
making skills, have uncovered a similar
State University Graduate School.
Appreciation is expressed to Tim Bray, Maxene Doty, pattern of results; gender differences in ca-
Shelley Rose, and Cindy Wilson for their assistance with reer-related self-efficacy exist, and these
the data collection and coding; to Nancy E. Betz for her differences in self-efficacy are predictive of
advice and assistance during all stages of the project;
gender differences in the career choice pro-
and to Robert Dolliver and three anonymous reviewers
for their helpful comments. cess (see Ayres, 1981; Kerns, 1981; Taylor &
An earlier version of this article entitled "Mathe- Betz, 1983).
matics Self-Efficacy and the Consideration of Math- Self-efficacy with regard to mathematics
Related Majors: A Preliminary Path Model" was has been chosen as the topic of investigation
presented as part of a Divisions 17, 15, and 35 sympo-
sium at the annual meeting of the American Psycho-
in the present study because of the increas-
logical Association, Los Angeles, August 1981. An ex- ing importance of adequate preparation in
panded report on the mathematics self-efficacy project mathematics to a broad range of career op-
including more extensive descriptions of the instru- tions. Gender differences in mathematics
ments and procedures is available from the author.
preparation have been termed a "critical
Requests for reprints should be sent to Gail Hackett,
who is now at the University of California, Santa Bar- filter" in the career development of women
bara, Counseling Psychology Program, Department of and minorities (Sells, 1980); lack of mathe-
Education, Santa Barbara, California 93106. matics preparation and consequent lower
47
48 GAIL HACKETT

levels of math achievement result in the matics (Pedro, Wolleat, Fennema, & Becker,
premature closure of options, effectively 1981; Sherman & Fennema, 1977).
serving to bar women and minorities from The affective and attitudinal variables
higher level, technological, and generally that have been found to predict mathemat-
male-dominated occupations (Fox, Brody, ics-related behavior include confidence in
& Tobin, 1980). Because of the importance learning mathematics, math anxiety, and the
of mathematics to the education and career perceived usefulness of mathematics. What
development of women (Sherman, 1979; Fennema and Sherman (1976) call "confi-
Wise, 1979), it becomes important to exam- dence in learning mathematics" is somewhat
ine the role played by mathematics-related similar to, yet distinct from, mathematics
self-efficacy expectations in the develop- self-efficacy. Confidence in learning
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ment of math-related careers. mathematics is a rather global estimate of


This document is copyrighted by the American Psychological Association or one of its allied publishers.

In any discussion of mathematics-related how well one expects him- or herself to do, or
career choice, the matter of gender differ- how well one has done, in math courses in
ences in mathematics achievement must be general. Mathematics self-efficacy is a more
addressed. A long-standing controversy still specific estimate of confidence in one's
reigns over the possible genetic origins of the ability to perform well with regard to par-
often-observed male superiority in mathe- ticular mathematics tasks or in particular
matics, and in fact this controversy has re- math and math-related courses. However,
cently been revived (Benbow & Stanley, the evidence supporting the relationship of
1980,1983). However, there is also strong confidence in learning mathematics to math
evidence for the existence of attitudinal and performance (Fennema & Sherman, 1977,
affective differences with regard to mathe- 1978), and the finding that men consistently
matics that can account for gender differ- score higher on the Fennema-Sherman Math
ences in math achievement (Fennema & Confidence Scale (Fennema & Sherman,
Sherman, 1977,1978; Sherman & Fennema, 1977), is suggestive of the importance of
1977). The finding of sex typing of mathe- mathematics-related self-efficacy to the
matics as a male domain on the part of male prediction of math-related career choice.
and female students (Fennema & Sherman, Findings with regard to the effects of
1978) further supports the notion that sex perceptions of the usefulness of mathematics
role socialization plays a major role in on math achievement are inconsistent, but
mathematics-related behavior. However, the relationship of math anxiety to math-
the effects of sex role socialization on math related performance and career choice is
achievement do not appear to be direct; that undeniable (Fennema, 1980). The results
is, sex role influences operate through other of studies by Betz (1978) and Hendel (1980)
factors, such as attittudes towards mathe- on very different populations were similar in
matics or election of math courses in high that math anxiety was found to be strongly
school (Sherman & Fennema, 1977). related to math performance. Based on her
Several studies uncovered a strong rela- research, Fennema (1980) has stated that
tionship between the nature and number of math confidence and math anxiety are
mathematics courses taken in high school highly interdependent, a conclusion in
and gender differences in math achievement keeping with the self-efficacy model, and in
(Fennema & Sherman, 1977; Hendel, 1980; fact Hendel (1980) found self-estimated
Wise, 1979). Fennema (1980) reported that, math ability to be highly correlated with
when mathematics background was con- math anxiety and also strongly related to
trolled for, something that few studies on math performance.
gender differences in mathematics have Goldman and Hewitt (1976) convincingly
done, few gender-related math achievement demonstrated the influence of mathematics
differences have been found. Furthermore, performance, as measured by the mathe-
the most important predictors of mathe- matics score of the Scholastic Aptitude Test
matics achievement, when math preparation (SAT), on math-related major choices in
is controlled for (i.e., persistence in math college. Their findings are helpful in ex-
courses and election of future courses in plaining why fewer women major in mathe-
mathematics), are attitudes toward mathe- matics and the sciences and provide further
MATH SELF-EFFICACY AND COLLEGE MAJOR CHOICES 49

evidence for the connection between math The original model was constructed ac-
background, performance, and attitudes and cording to predictions made on the basis of
math-related major and career choices. the self-efficacy approach to career devel-
Finally, two studies focusing on tests of opment (Hackett & Betz, 1981), tested for its
the application of self-efficacy theory to the explanatory adequacy, and revised.
career development process provide direct
support for the role of mathematics self- Method
efficacy in the career choice process. Betz
and Hackett (1983) found gender differences The development of the path model described in this
in math self-efficacy to be correlated with article is based on data collected as a part of a larger
gender differences in attitudes toward project on mathematics self-efficacy and math-related
career choice (see Betz & Hackett, 1983; Hackett &
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

mathematics and choice of math-related Betz, 1984). The discussion that follows has been ab-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

college majors. Hackett and Betz (1984) breviated; only those components of the larger project
reported moderate correlations between that are directly relevant to the path analysis will be
mathematics performance and math self- discussed in depth.
efficacy, and significant gender differences
on both variables. Results of both studies Instruments
supported the superiority of math self-effi-
cacy over all indices of math attitudes and Mathematics Self-Efficacy Scale (MSES). The
Mathematics Self-Efficacy Scale (MSES), developed
math performance in predicting the math-
by Betz and Hackett (1983), contains 52 items identified
related major choices of college students. as relevant to the study of math-related self-efficacy
Although much is known about the vari- expectations. The scale is composed of three subscales:
ables that are thought to influence math (a) the math tasks subscale, consisting of 18 items in-
achievement and math career choice, the volving "everyday" math tasks (e.g., balancing a
checkbook); (b) the math courses subscale, consisting
documented correlations among these vari- of 16 math-related college courses; and (c) the math
ables do not allow strong statements about problems subscale, consisting of 18 arithmetic, algebra,
the causal relationships underlying them. and geometry problems.
Goldman and Hewitt (1976) and Hendel For the courses subscale, subjects were instructed to
rate their confidence in their ability to complete each
(1980) have suggested that prior perfor-
course with a grade of B or better. For the math tasks
mance and achievement are the factors that and math problems subscales, the subjects simply rated
are influenced by gender and consequently their confidence in their ability to successfully perform
directly influence math-related major and the task or solve the problem. Confidence ratings for
all scales were elicited on a 10-point continuum (0 = no
career choices. Self-efficacy theory (Ban-
confidence, 9 = complete confidence). Mean scores for
dura, 1977,1982; Hackett & Betz, 1981) and math self-efficacy were calculated across the three
consequent research investigating the role subscales. Betz and Hackett (1983) reported moderate
of mathematics self-efficacy in the career item-total score correlations for the MSES subscales
choice process (Betz & Hackett, 1983; and a high internal consistency reliability (coefficient
alpha) for the total scale (.96).
Hackett & Betz, 1984) provide support for Fennema-Sherman Mathematics Attitude Scales.
the view that mathematics-related self-ef- Five of the nine Mathematics Attitude Scales (Fennema
ficacy, as influenced by gender, socialization, & Sherman, 1976)—Confidence in Learning Mathe-
and math level and background, is more matics, Usefulness of Mathematics, Mathematics as a
strongly predictive of math-related major Male Domain, Effectance Motivation in Mathematics,
and the Math Anxiety scale (as revised by Betz, 1978,
and career choices than ability, math back- for use with college students)—were administered.
ground, or gender alone or in combination. However, only the scores from the Math Anxiety scale
Furthermore, math anxiety, according to the were employed in developing and testing the path
self-efficacy approach, is viewed as a conse- model. This decision was based on findings from pre-
quence of efficacy expectations with regard vious research (Betz & Hackett, 1983) indicating that
math anxiety was the only variable from the Fen-
to mathematics and is therefore a joint pre- nema-Sherman Mathematics Attitude Scales to sig-
dictor, along with math self-efficacy, of nificantly contribute to the prediction of math-related
math-related career choices. major choices.
The purpose of the present study, then, is Responses on the 10-item revised Math Anxiety scale
were scored on a 5-point Likert-type scale from 1
to develop and refine a causal model of (strongly disagree) to 5 (strongly agree). Half of the
math-related major choice utilizing math items were negatively worded and half were positively
self-efficacy as a major mediating variable. worded; those that were negatively worded were re-
50 GAIL HACKETT

verse-scored, and item scores were added to obtain the on math-related major choice. Causal modeling in
overall Math Anxiety score. High scores on the Math general, and path analytic techniques in particular, have
Anxiety scale are indicative of low levels of math anxi- been developed to examine and test causal relationships
ety. The split-half reliability of the revised Math in nonexperimental research (Asher, 1976; Blalock,
Anxiety scale was .92 (Betz, 1978). 1964; Heise, 1975). This approach to the analysis of
Background questionnaire. A series of questions correlational data has not been used extensively in
were added to the math attitude scales in order to gather psychological and educational research, but its potential
a range of demographic information as well as infor- applicability to these fields is beginning to be ac-
mation relative to mathematics preparation and career knowledged (Leclair, 1981).
plans. Three variables relevant to the path analysis Causal modeling is especially appropriate with regard
were elicited from each student: gender, years of high to research in the area of gender-related differences in
school mathematics, and the student's declared major, mathematical ability and math self-efficacy because of
if a major had been declared. The majors were then the consistent identification in the literature of several
rated according to Goldman and Hewitt's (1976) variables correlated with and hypothesized to influence
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

science/math continuum, from the fine arts (1) to the math ability, and the clear consequences of the observed
This document is copyrighted by the American Psychological Association or one of its allied publishers.

physical sciences and mathematics (5), in order to ob- differences in math ability for the career choice process.
tain an index of the math relatedness of the major In other words, the following are related in some way to
choice. Gender was transformed into a dummy variable math-related career choices: gender and socialization
for the regression equations, with females coded 1 and influences, amount of high school mathematics, level
males coded 0. of math achievement, math anxiety, and confidence in
Bern Sex Role Inuentory (BSRI). The mean mas- learning math. Path analytic techniques allow us to
culinity scale score from the BSRI was employed in this move beyond simple or multiple correlations to testing
investigation as an index of the effects of sex role or- the causal ordering of these variables that is hypothe-
ientation (Bern, 1974). The BSRI consists of 60 ad- sized on the basis of self-efficacy theory.
jectives for which respondents are asked to rate them- Path analysis consists of three stages: (a) develop-
selves according to a 7-point scale (1 = never or almost ment of a causal scheme or path model specifying the
never true; 7 = always or almost always true). The hypothesized causal relationships among the variables,
masculine subscale comprises 20 adjectives that have (b) computation of path coefficients and elimination of
been judged as more socially desirable for men in our nonsignificant paths in the original model, and (c)
culture. specification of a reduced path model consistent with
the data. Path coefficients are computed via a series
of multiple regression analyses based on the hypothe-
Participants and Procedure sized model, are statistically identical to standardized
multiple regression coefficients, and represent the direct
The original 262 subjects who participated in the effects of one variable in the model on another, when the
study (153 females, 109 males) were undergraduate influences of all other causally prior variables are con-
volunteers enrolled in introductory psychology courses trolled (Leclair, 1981). Path analytic procedures also
who received course credit for their participation. allow for the decomposition of zero-order correlations
Because introductory psychology is required as a part between variables into indirect as well as direct causal
of the basic educational requirements for most major effects.
fields, participants represented a wide variety of aca- Thus, based on self-efficacy theory and related re-
demic majors. search on mathematics attitudes and achievement, the
The order of administration of the two math ques- full path model to be tested is as follows: Gender is
tionnaires was randomly assigned to small groups of hypothesized to influence all other variables directly
subjects, and the BSRI was always administered last. and indirectly; sex role socialization, as measured by the
Mathematics scale scores from the American College BSRI masculinity scale, is hypothesized to mediate
Test (ACT) were obtained from university records for some of the influence of gender and is, in turn, hy-
70% of the subjects. pothesized to affect years of high school mathematics
One hundred seventeen subjects (72 females, 45 courses taken, ACT math scores, and finally math
males), or 45% of the original subject pool, were ulti- self-efficacy. Math self-efficacy is hypothesized to
mately selected for inclusion in the path analysis. mediate these causally prior influences and conse-
Reduction in the sample size resulted from missing data; quently affect math anxiety and math-related major
ACT scores were unavailable for 30% of the sample, and choice.
only 69% of the subjects had declared a major. Exam-
ination of the means and standard deviations from the Results
full and reduced sample revealed a surprisingly close
correspondence between these sets of scores and pro- Zero-order correlations were obtained for
moted confidence in the representativeness of the
subject sample employed in the path analysis. all variables included in the analysis, and
these correlations, along with the respective
means and standard deviations for each
Data Analysis variable, are presented in Table 1.
Path analysis was chosen as the most appropriate
Correlations between all variables in-
procedure for examining the possible causal relation- cluded in the model, except for the mascu-
ships among the variables identified as having an effect linity score, were significant in relation to the
MATH SELF-EFFICACY AND COLLEGE MAJOR CHOICES 51

Table 1
Means, Standard Deviations, and Zero-Order Correlations of Variables Included in the Path
Analysis

Variable M SD

1. Gender"
2. BSRI masculine score 4.85 0.69 -.28***
3. Years of high school math 4.05 0.86 -.18* .06
4. ACT math score 20.21 6.6 -.19* -.01 .46***
5. Mathematics self-efficacy 6.40 1.18 -.25** .26** .47*** .66***
6. Math Anxiety score 30.30 8.77 -.23** .02 .33*** .47*** .58***
7. Math relatedness of declared majorb 3.62 0.85 -.37*** .10 .43*** .38*** .50*** .44***
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Note. N = 117 (72 females and 45 males). BSRI = Bern Sex Role Inventory; ACT = American College Test.
a
Gender was coded 0 for males, 1 for females; negative correlations indicate higher scores for men.
b
Math/science relatedness of major was coded on a continuum from 1 (fine arts) to 5 (physical sciences/mathe-
matics).
*p<.05. **p<.01. ***p<.001.

math/science relatedness of major choice; presented in Table 2, along with the break-
self-efficacy was the variable most highly down of the causal effects and the R2 from
correlated with math-related major choice the full and reduced models. The decom-
(r = .50, p < .001). The ACT mathematics position of causal effects is a means for de-
score, math anxiety, and years of high school termining the true versus the spurious, or
mathematics variables demonstrated the unexplained, covariance. The direct effects
highest correlations with math self-efficacy represent the path coefficients, or the true
(r = .66, .58, and .47, respectively), although effects of the predictor variables on the de-
correlation coefficients between math self- pendent variable within each equation; the
efficacy and the remaining variables indirect effects represent the influence of the
'achieved statistical significance as well. predictor variables mediated by other vari-
Masculinity scores bear the lowest direct ables in that equation. As can be seen from
relationships to the variables in the path Table 2, the reduced model explained almost
model; significant correlations were only as much of the variance as the full model for
found between masculinity and self-efficacy all regression equations, therefore providing
and gender (rs = .26 and —.28), indicating a good representation of the observed data
that males tended to score higher on the in a more parsimonious form.
BSRI masculinity scale and that higher Although a large amount of unexplained
masculinity scores were related to higher variance exists in the prediction equations
levels of math self-efficacy. for masculine sex role and years of high
The correktion between ACT math scores school mathematics, later equations yielded
and years of high school math was statisti- a moderate to high ft2 (amount of variance
cally significant (r = .46, p < .001), but both explained) for the reduced path model.
scores were also strongly correlated with Evidently, some exogenous variables that
math self-efficacy and math anxiety. have not been incorporated into the model
To test the hypothesized causal ordering are contributing to the variability in the
of these variables, path coefficients were prediction of masculine sex role orientation
estimated via a series of multiple regressions. and years of high school mathematics
Each endogenous variable in the model was scores.
regressed on all causally prior variables, and Gender and years of high school mathe-
beta (standardized regression) coefficients matics adequately predict math ACT scores,
were examined (Asher, 1976). In an attempt and ACT math, years of high school math,
to control for practical as well as statistical and masculine scores in combination explain
significance, all path estimates of less than a great deal of the variability in math self-
.10 were excluded from the model, and the efficacy (ft 2 = .54, accounting for 54% of the
regression analyses were performed again on variance). These data support the conten-
the reduced model. These analyses are tion that amount of high school math prep-
52 GAIL HACKETT

aration influences self-efficacy directly, but and math anxiety is statistically nonsignifi-
also indirectly, that is, through math cant (r = .02).
achievement, as measured by math ACTs. Finally, the science/math relatedness of
There appear to be two separate paths of college major choice is predicted directly by
influence from gender to math self-efficacy: gender, years of high school mathematics,
The first path, through masculinity, might math self-efficacy, and math anxiety (R2 =
be termed "direct socialization influences"; .38). It is clear from the decomposition of
the second, or indirect influence of sociali- the causal effects that gender, years of high
zation, reflects math preparation and school math, and math self-efficacy have
achievement. Gender did not influence indirect as well as direct effects on major
math self-efficacy directly. choice. The relatively high degree of co-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

A moderate amount of the variability in variation between math anxiety and major
This document is copyrighted by the American Psychological Association or one of its allied publishers.

math anxiety scores was accounted for by choice that is left unaccounted for by the
gender, masculine sex role, and math self- model leads to the .suspicion that math
efficacy (fl 2 = .37), although the direct ef- anxiety may have indirect effects on major
fects of masculinity scale scores on math choice as well; these indirect influences could
anxiety were contrary to expectations. The be transmitted by another variable that
coefficient for the masculinity/math anxiety ought to be included in the model, or more
path was —.17, indicating that low mascu- probably, they are the result of a reciprocal
linity scores were related to lower levels of interaction between math anxiety and math
math anxiety (high scores on the Fennema- self-efficacy.
Sherman Math Anxiety scale). Yet the in- Figure 1 presents the reduced path model
direct effects of masculinity on math anxiety, with all significant path coefficients, along
as mediated by math self-efficacy, are posi- with path coefficients for the residual vari-
tive; that is, higher masculinity scores are ables. In the figure, the path coefficients (p
related to lower levels of math anxiety. The values) represent the direct influence of
zero-order correlation between masculinity causally prior variables on succeeding vari-

Table 2
Decomposition of the Effects from the Path Analysis Using the Reduced Model

fl2
Zero- causal eriecre
order Noncausal Full Reduced
Effect r Direct Indirect Total covariation model model

On masculine sex role:


Of gender -.277 -.277 .000 -.277 .000 .076 .076
On years of high school math:
Of gender -.180 -.180 .000 -.180 .000 .033 .032
On ACT math scores:
Of gender -.193 -.115 -.078 -.193 .000 .226 .221
Of years math .457 .436 .000 .436 .021 — —
On math self-efficacy:
Of masculine sex role .262 .259 .000 .259 .003 .541 .540
Of years math .472 .194 .251 .445 .027 — —
Of ACT math .661 .576 .000 .576 .085 —

On math anxiety:
Of gender -.234 -.134 -.039 -.173 -.061 .380 .374
Of masculine sex role .022 -.172 .154 -.018 .040 — —
Of math self -efficacy .583 .595 .000 .595 -.012 — —
On math relatedness of major:
Of gender -.374 -.236 -.127 -.363 -.011 .379 .377
Of years math .429 .215 .152 .367 .062 — —
Of math self -efficacy .502 .244 .100 .344 .158 —

Of math anxiety .437 .168 .000 .168 .269 — —

Note. ACT = American College Test. Direct causal effects = path coefficients; Rz = timount of variance ex-
plained.
MATH SELF-EFFICACY AND COLLEGE MAJOR CHOICES 53

R=98
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

R=.96 R=.79

Figure 1. Reduced path model representing causal relationships among variables predicting math-
related college major choice (Path coefficients, ps, represent direct causal influences in direction of
arrows; R values represent path coefficients of residuals, i.e., variables not included in the model; gender
was coded 0 for males and 1 for females so that negative correlations indicate higher scores for males;
BSRI = Bern Sex Role Inventory; HS = high school; ACT = American College Test.)

ables; their relative size is indicative of their by the results of the current investigation.
predictive importance in the model. Fur- Gender-related socialization influences in
thermore, squaring any path coefficient combination with amount of mathematics
yields the percentage of explained variance preparation predict level of mathematics
for each path. The path coefficients for the achievement, which is, in turn, predictive of
residual variables, representing the effects mathematics-related self-efficacy. Math
of variables not included in the model, are self-efficacy, as influenced by these causally
represented by the Rs in Figure 1. These prior variables, is consequently predictive of
effects have been calculated according to the both math anxiety and math-related major
formula (1 — R2)1/2, that is, the square root choices. The correlational data are very
of 1 minus the percentage of explained similar to the findings of previous research
variance from each regression equation on math achievement (Fennema & Sherman,
(Asher, 1976). 1977, 1978; Sherman, 1979; Sherman &
Fennema, 1977); the path analysis supports
Discussion the central mediational role of mathematics
self-efficacy in the development of math-
The hypothesized causal ordering of the related careers.
variables correlated with mathematics Unexpected results emerging from the
achievement was at least partially supported study point to several deficiencies in the
54 GAIL HACKETT

proposed path model. Obviously, gender lated to math anxiety when a simple corre-
alone is insufficient to explain mathemat- lation was computed, yet masculinity con-
ics-related behavior in high school. The tributed significantly but negatively to the
relatively low proportion of the variance regression equation predicting math anxiety
accounted for by the first several variables (path coefficient = —.17). This puzzling
suggests the existence of at least one, if not finding may also be due to measurement
several, exogenous variables that should be problems.
added to increase the model's explanatory Finally, the proposed model does a good
capability and predictive power. For ex- job of predicting math relatedness of college
ample, socioeconomic status (Wise, 1979), major, accounting for 38% of the variance.
parental background and attitudes (Fenne- However, as discussed in the Results section,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ma & Sherman, 1977), and the influences of there is also a great deal of covariation not
This document is copyrighted by the American Psychological Association or one of its allied publishers.

teachers and school systems (Fennema & accounted for by the model, especially with
Sherman, 1977,1978) all may affect mathe- regard to the math anxiety variable. Evi-
matics-related attitudes and behaviors in dently, math anxiety influences math rela-
high school. tedness of major choice indirectly as well as
Another interesting outcome of the study directly, and the indirect influence, which is
was the failure of the BSRI masculinity score not depicted in the model, must be mediated
to significantly contribute to the prediction by some other variable, possibly math self-
of years of high school mathematics courses efficacy. Alternatively, math anxiety may
or math ACT scores. Gender-related dif- not be a unique contributor to math-related
ferences in mathematics preparation and major choice. Rounds and Hendel (1980)
achievement were expected to be mediated have suggested that math anxiety is actually
by the masculinity scale score, an index of nothing more than a reflection of various
the extent of sex role socialization. That is, other factors, such as mathematics back-
gender was thought to affect math prepara- ground and test anxiety. This hypothesis
tion and achievement indirectly, via mas- has already begun to generate empirical
culine sex role orientation. This evidently scrutiny (e.g., Dew, Galassi, & Galassi,
is not the case. 1983).
While the results with regard to the mas- The above findings highlight one of the
culinity scores may very well be valid, there major problems with the analysis—that is,
is always the possibility that they may be an that the model is static and only captures to
artifact of the instrumentation or design of a limited degree the processes underlying the
the study. There is no generally satisfactory development of math-related competencies
instrument for the assessment of the effects and attitudes. Self-efficacy theory suggests
of sex role socialization. Theoretical and an ongoing and reciprocal interaction be-
psychometric problems with the concept of tween mathematics-related efficacy expec-
androgyny and sex role orientation abound, tations and math anxiety, mathematics
and the BSRI in particular has been heartily achievement, vocational interests, and
criticized (e.g., Locksley & Colten, 1979; math-related major and career choice
Pedhazur & Tetenbaum, 1979). Further- (Hackett & Betz, 1981). Therefore, nonre-
more, the BSRI was developed to categorize cursive models incorporating these hypoth-
respondents into a fourfold sex role classi- esized reciprocal linkages must be developed
fication through the use of both feminine and tested, and studies of a longitudinal
and masculinine scale scores. The em- nature need to be conducted to build a model
ployment of the masculine scale alone, ne- that will come closer to reflecting the process
cessitated because the path analysis required of career development and mathematics-
interval data, may account for the prob- related career choice.
lematic findings. In conclusion, the strongest statement
The suspicion that the BSRI masculinity that can be made on the basis of these results
score was not an appropriate measure of sex is that the observed data are consistent with
role socialization was reinforced by the self-efficacy theory. A major limitation of
finding that masculinity scores were unre- the analysis, however, is that the self-efficacy
MATH SELF-EFFICACY AND COLLEGE MAJOR CHOICES 55

model was not compared to other, rival the- References


oretical models, nor was an overall signifi-
cance test conducted for the goodness of fit Asher, H. B. (1976). Causal modeling. Beverly Hills,
of the reduced model. Due to the limited CA: Sage.
Ayres, A. (1981, August). Self-efficacy expectations
scope of this preliminary investigation, ex-
with respect to occupationally specific behaviors.
pansion and refinement of the model and the Paper presented at the annual meeting of the
testing of alternative models await future American Psychological Association, Los Angeles.
research. Nonetheless, the present inves- Bandura, A. (1977). Self-efficacy: Toward a unifying
tigation yielded results congruent with the theory of behavioral change. Psychological Review,
84,191-215.
results of previous correlational studies on Bandura, A. (1982). Self-efficacy mechanism in
the self-efficacy approach to women's career human agency. American Psychologist, 37,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

development (Ayres, 1981; Betz & Hackett, 122-147.


This document is copyrighted by the American Psychological Association or one of its allied publishers.

1981, 1983; Hackett & Betz, 1984; Kerns, Bern, S. L. (1974). The measurement of psychological
androgyny. Journal of Consulting and Clinical
1981; Taylor & Betz, 1983), thus encouraging
Psychology, 42,155-162.
further work in the area. Benbow.C. P., & Stanley, J.C. (1980). Sex differences
The implications for counselors are also in mathematical ability: Fact or artifact? Science,
clear: Counselors working with clients with 210,1262-1264.
career concerns, particularly female clients Benbow, C. P., & Stanley, J.C. (1983). Sex differences
in mathematical reasoning ability: More facts.
with career concerns, must work not only Science, 222,1029-1031.
with the interests, values, and abilities of Betz, N. E. (1978). Prevalence, distribution, and
each client but also with the individual's correlates of math anxiety in college students.
perceptions of his or her abilities and per- Journal of Counseling Psychology, 25, 441—448.
ceptions of the gender-appropriateness of Betz, N. E., & Hackett, G. (1981). The relationship
of career-related self-efficacy expectations to per-
the options that may be available. In fact, ceived career options in college women and men.
the findings of the present study and those Journal of Counseling Psychology, 28, 399-410.
of Hackett and Betz (1984) imply that, at Betz, N. E. & Hackett, G. (1983). The relationship of
least with college-aged women and men, mathematics self-efficacy expectations to the selec-
tion of science-based college majors. Journal of
self-efficacy expectations with regard to Vocational Behavior, 23,329-345.
occupations and career-related domains are Blalock, H. M. (1964). Causal inferences in nonex-
much more important than measured abili- perimental research. New York: Norton.
ty. Self-efficacy expectations encompass Dew, K. M. H., Galassi, J. P., & Galassi, M. D. (1983).
ability information but are significantly Mathematics anxiety: Some basic issues. Journal
of Counseling Psychology, 30, 443^146.
more predictive of career choice behavior Fennema, E. (1980). Sex-related differences in
than ability or performance. mathematics achievement: Where and why. In L.
Finally, math-related majors and careers H. Fox, L. Brody, & D. Tobin (Eds.), Women and the
are still viewed as inappropriate or less ap- mathematical mystique (pp. 76-93). Baltimore:
Johns Hopkins University Press.
propriate for women by both women and
Fennema, E., & Sherman, J. A. (1976). Fennema-
men. Signs that these stereotypes are Sherman Mathematics Attitude Scales: Instruments
breaking down exist, but there remain the designed to measure attitudes toward the learning of
subtle and indirect effects of sex role so- mathematics by females and males. JSAS Catalog
cialization, that is, lack of confidence or low of Selected Documents in Psychology, 6, 31. (Ms.
No. 1225)
self-efficacy with regard to mathematics and Fennema, E., & Sherman, J. A. (1977). Sex-related
other vocational domains traditionally differences in mathematics achievement, spatial vi-
dominated by men (see, e.g., Betz & Hackett, sualization, and affective factors. American Edu-
1981), which may still effectively serve to cational Research Journal, 14, 51-71.
Fennema, E., & Sherman, J. A. (1978). Sex-related
hold women back from the pursuit of these
differences in mathematics achievement and related
nontraditional and generally higher level, factors: A further study. Journal for Research in
higher paying occupations. Therefore, Mathematics Education, 9,189-203.
counselors and educators must actively work Fitzgerald, L. F., & Crites, J. 0. (1980). Toward a
to bolster women's career and math-related career psychology of women: What do we know?
What do we need to know? Journal of Counseling
self-efficacy expectations if dramatic shifts Psychology, 27, 44-62.
toward more informed and nonstereotyped Fox, L. H., Brody, L., & Tobin, D. (Eds.). (1980).
career choices are to be effected. Women and the mathematical mystique. Balti-
56 GAIL HACKETT

more: Johns Hopkins University Press. (1981). Election of high school mathematics by fe-
Goldman, R. D., & Hewitt, B. N. (1976). The Scho- males and males: Attributions and attitudes.
lastic Aptitude Test "explains" why college men American Educational Research Journal, 18,
major in science more often than college women. 207-218.
Journal of Counseling Psychology, 23, 50-54. Rounds, J. B., & Hendel, D. D. (1980). Measurement
Hackett, G., & Betz, N. E. (1981). A self-efficacy ap- and dimensionality of mathematics anxiety. Journal
proach to the career development of women. Jour- of Counseling Psychology, 27,138-149.
nal of Vocational Behavior, 18, 326-339. Sells, L. W. (1980). The mathematics filter and the
Hackett, G., & Betz, N. E. (1984). Mathematics education of women and minorities. In L. H. Fox, L.
performance, mathematics self-efficacy, and the Brody, & D. Tobin (Eds.), Women and the mathe-
prediction of science-based college majors. Man- matical mystique (pp. 66-75). Baltimore: Johns
uscript submitted for publication. Hopkins University Press.
Heise, P. R. (1975). Causal analysis. New York: Sherman, J. (1979). Predicting mathematics perfor-
Wiley. mance in high school girls and boys. Journal of
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Hendel, D. D. (1980). Experimental and affective Educational Psychology, 71,242-249.


This document is copyrighted by the American Psychological Association or one of its allied publishers.

correlates of math anxiety in adult women. Psy- Sherman, J., & Fennema, E. (1977). The study of
chology of Women Quarterly, 5, 219-230. mathematics by high school girls and boys: Related
Kerns, E. M. (1981). Chemistry self-efficacy and Us variables. American Educational Research Journal,
relationships to the avoidance of chemistry-related 14,159-168.
majors and careers. Unpublished undergraduate Taylor, K. M., & Betz, N. E. (1983). Applications of
honors thesis, Ohio State University. self-efficacy theory to the understanding and treat-
Leclair, S. W. (1981). Path analysis: An informal ment of career indecision. Journal of Vocational
introduction. Personnel and Guidance Journal, 59, Behavior, 22,63-81.
643-646. Vetter, L. (1978). Career counseling for women. In
Locksley, A., & Colten, M. E. (1979). Psychological L. H. Hansen & R. S. Rapoza (Eds.), Career devel-
androgyny: A case of mistaken identity? Journal opment and counseling of women (pp. 7-26).
of Personality and Social Psychology, 37, Springfield, IL: Charles C Thomas.
1017-1031. Wise, L. L. (1979, April). Long-term consequences
Osipow, S. H. (1983). Theories of career development of sex differences in high school mathematics edu-
(3rded.). Englewood Cliffs, NJ: Prentice-Hall. cation. Paper presented at the annual meeting of the
Pedhazur, E. J., & Tetenbaum, T. J. (1979). Bern Sex American Educational Research Association, San
Role Inventory: A theoretical and methodological Francisco.
critique. Journal of Personality and Social Psy-
chology, 37, 996-1016. Received March 26,1984
Pedro, J. E., Wolleat, P., Fennema, E., & Becker, A. D. Revision received July 24,1984 •

View publication stats

You might also like