Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=vjeb20
Download by: [Universitas Maritim Raja Ali Haji] Date: 12 January 2016, At: 23:57
Journal of Education for Business
ISSN: 0883-2323 (Print) 1940-3356 (Online) Journal homepage: http://www.tandfonline.com/loi/vjeb20
Testing for Bias Against Female Test Takers of
the Graduate Management Admissions Test and
Potential Impact on Admissions to Graduate
Programs in Business
Robert E. Wright & Daniel G. Bachrach
To cite this article: Robert E. Wright & Daniel G. Bachrach (2003) Testing for Bias Against Female Test Takers of the Graduate Management Admissions Test and Potential Impact on Admissions to Graduate Programs in Business, Journal of Education for Business, 78:6, 324-328, DOI: 10.1080/08832320309598621
To link to this article: http://dx.doi.org/10.1080/08832320309598621
Published online: 31 Mar 2010.
Submit your article to this journal
Article views: 12
View related articles
Testing for Bias Against Female
Test Takers of the Graduate
Management Admissions Test and
Potential Impact on Admissions to
Graduate Programs in Business
ROBERT E. WRIGHT
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
University of Illinois at Springfield Springfield, Illinois
DANIEL G. BACHRACH
University
of
AlabamaTuscaloosa, Alabama he Graduate Management Admis-
T
sions Test (GMAT) provides a stan- dardized measure for college and univer- sity admissions committees to use in evaluating potential graduate students inbusiness (Graham, 1991; Johnson
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
&McLaughlin, 1993; Youngblood & Mar- tin, 1982). Business programs in all areas of graduate business instruction use the GMAT to help administrators make admissions decisions. In addition, the GMAT admissions criteria main- tained by colleges and universities also
are used by national forums such as
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
U.S.zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
News and World Report as a measure ofselectivity, which can have a strong influence on the caliber of students applying to those institutions and on the national and international reputations of those universities competing for this tal- ent. Empirical studies based on data from a wide array of colleges and uni- versities (Carver & King, 1994; Foxall, 1992; Paolillo, 1982) have shown the GMAT to be a significant predictor of graduate business school performance. Indeed, the majority of studies for which data on the GMAT have been collected
ABSTRACT. In this study, the authors examined the relationship between scores on the Graduate Man- agement Admissions Test (GMAT) and achievement in core academic courses. The authors used a sample
that revealed the potentially differen-
tial academic performance of male
and female MBA students ( N =
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
334)within particular GMAT scoring cohorts. Results showed that, for the top scoring GMAT cohort, there was statistically significant evidence of bias against the women. The potential effect of this bias against women may
affect admissions decisions to gradu-
ate programs in business, particularly at the most selective MBA programs. (Baird, 1975; Carver & King, 1994; Fisher & Resnick, 1990; Graham, 1991) show that GMAT scores are the single best predictor of academic performance, accounting for a significant percentage of the variance in academic perfor- mance in these studies. However, gen- der bias may be an important mitigating factor influencing the usefulness of this test (Hancock, 2000). Our goal in this study, therefore, was to determine whether there is such a bias, and, if so, whether it unfairly impedes women’s access to graduate business schools and consequently to the more rarified
heights of the business world (Hilgert, 1998; Simpson, 2000).
Gender Bias and Standardized Testing
Although the current research focus- es on the GMAT, investigations into the impact of bias against female test takers of the SAT, another ubiquitous stan- dardized admissions tool, also have been illuminating with regard to the bias issue (Bendlow & Stanley, 1980, 1983; Bridgeman & Wendler, 1991; Young, 1994). Those studies may pro- vide some perspective on the pervasive- ness of this factor. Specifically, results from those studies indicate that the math portions of the SAT consistently
underpredict female performance in college mathematics courses, despite the researchers’ controlling for such things as course selection. The results from these SAT studies are consistent with voluminous literature on the impact of gender bias on an array of standardized tests (e.g., GRE-Q, GMAT-Q, MCAT-Q, etc.) that all point
324 Journal
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
of Education for Businessto a consistent underprediction for female test takers in aptitude assess- ments containing a significant quantita- tive component (see Wilder and Powell,
1989 for a review).
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Evidence of Gender Bias of the GMAT and Implications
Despite the predictive characteristics of tests such as the SAT and GMAT, there is increasing speculation among peda- gogical researchers regarding the fairness
of these tests (Dobson, Krapljan-Ban;
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
&Vielba, 1999), specifically regarding gender bias (Hancock, 2000). For exam- ple, Hancock (2000) has reported results suggesting that the GMAT may be biased against female test takers. Hancock’s conclusion was based on the differential performance that he found on the GMAT for admitted male and female master’s of business administration (MBA) students relative to their grade point average (GPA) in those same programs. Specifi- cally, Hancock reported that although men and women completing a particular MBA program did not have significantly different overall graduate GPAs, the male students had statistically significantly higher GMAT scores than the female stu- dents completing the program. Hancock concluded that although students of both genders were performing similarly in their degree course work, the business aptitude tests were clearly biased toward male test takers.
As indicated previously, a number of published studies show that the GMAT is a good predictor of performance in graduate business schools, although issues of gender bias in this predictive tool continue to be raised by scholars in the field. For example, in a finding sim- ilar to that reported by Hancock (2000), Deckro and Wounderberg (1 977), hold- ing constant the GMAT scores reported by male and female students, found that female MBAs significantly outper- formed their male counterparts in their graduate course work. Fisher and Resnick (1990) found that the ideal weighting for the GMAT in a regression equation for determining performance in a graduate business school differed between men and women, with the GMAT a better predictor of perfor- mance for men than for women. How-
ever, Carver, Jr. and King (1994), Gra- ham (1991), and Paolillo (1982) all found that gender was not a significant determinant of performance in graduate business schools.
Recent data published by the Gradu- ate Management Admission Council shows that men consistently outscore women on the GMAT by approximate- ly 40 points (GMAC, 2001). Clearly, any bias inherent in the GMAT poten- tially could be a major deterrent to female entry into graduate schools of business. Because the MBA is increas- ingly a major entry credential to upper- level management positions in many corporations both within the United States and abroad (Joyce, 2002), bias in the GMAT may result in decreased ability of qualified women to climb the corporate ladder. Given the paucity of women in upper-level management today (Dalton & Daley, 1998; Ragins, Townsend, & Mattis, 1998), this deter- rent is a potentially major societal problem. Bias against women, in a cli- mate currently characterized by dra- matic increases in applications to MBA programs, could have a deleterious effect, particularly given the increased selectivity of top MBA programs (Alsop, 2001).
Purpose
The issue of gender bias in the GMAT is an important contemporary issue (Hilgert, 1998; Simpson, 2000; Sinclair, 1997), and the results reported by Hancock (2000) and others demon- strate a potential bias in the GMAT for graduate admissions decisions. Given previous controversial findings con- cerning possible gender bias of the GMAT, we sought to investigate the fol- lowing question: Is there bias, and, if so, does the extant bias in the GMAT against female test takers have the potential to affect negatively the entry of women into graduate business schools, and as a consequence, into the business world (Hilgert, 1998; Simpson, 2000)? Using course performance as the criteria, we sought to examine, in par- ticular, the performance of male and female MBAs to determine (a) the level,
or
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
cohort, of GMAT performance at which any bias against women was evi-dent and (b) whether any such bias was problematic for all cohorts of GMAT performance (as suggested by Han- cock). We measured course perfor- mance by grade point average (GPA). (Although we used this measure to uncover any evidence of bias, it is not clear whether graduate GPA necessarily is predictive of a student’s potential pro- fessional performance.)
Method
Participants
Three hundred and thirty-four MBA students enrolled in the standard gradu- ate business curriculum at a small, mid- western university served as the sub- jects in this study. Fifty-six percent of the sample was male ( N = 190). The
average age of the sample was 36.4
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(SD
= 7.01 years). Male and female respondents did not differ significantly in terms of age, years of work experi- ence, or percentage of full versus part- time work experience.Procedures
We collected data from the MBA records.office at the university and exam- ined the relationship between GMAT and graduate GPA in core courses. We used core GPA as the basis of comparison between cohorts because all students in the program are obligated to take the same core courses. Thus, this metric pro- vided a standardized basis of comparison for making judgments about bias and effect of bias. To eliminate possible bias resulting from differential dropout rates, we used records of all students who were admitted to the MBA program and took at least one core course.
Results
We focused on the differences in graduate school performance as a func- tion of scores on the Graduate Manage- ment Admissions Test. To examine these kinds of between-groups mean differences, we conducted a series of planned comparisons. We conducted the first set of tests to establish that the pro- file of the sample in the current study, with regard to GMAT performance and
July/August 2003 325
GPA,
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
was consistent with those samples examined in previous studies demon-strating evidence of bias (Hancock, 2000). Initially, we analyzed the sample as a whole, without regard to GMAT
cohort. The average GMAT score for
the sample as a whole was 491
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(SD =zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
89.80) with a range from 280 to 760. The average GMAT score for the men in
the sample was 503 (SD = 91.98), whereas the average GMAT score for
the women was 474.65 (SD = 84.57). In the sample as a whole, the GMAT
scores for male and female students were significantly different from one
another at the p < .005 level
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
( r
= 2.92, p< .005, df = 332).
The overall GPA in the sample as a
whole was 3.48 (SD = .37). The average
GPA for the male students in the sample
was 3.49 (SD = .37), whereas the aver- age GPA for the female students was
3.48 (SD = .37). There was no signifi- cant difference in the GPAs between
male and female students at the p < .05
level (t = .40, p = .69). So, although
course performance was not significant- ly different between male and female students, male students in the study had been admitted to the program with sig- nificantly higher GMAT scores than
female students. This result is consistent with Hancock’s findings, suggesting that although female students are admit- ted to business programs with lower average GMAT scores, these same stu-
dents perform as well as male students on the course component of their degrees.
After determining that in the gross analysis the profile of the sample with respect to relative GMAT score and GPA
across males and females conformed to that reported by Hancock (ZOOO), we arrived at the next question: How might the bias against female test takers affect female students’ entry into graduate schools? It was not possible to answer this question directly in the current study, because of the cross-sectional nature of the data. However, we were able to draw some strong inferences regarding the potential influence of that bias.
To evaluate the influence of bias, we
split our sample of MBAs into three
cohorts based on GMAT scores. We
defined the cohorts as those students
scoring (a) from 400 to 500 ( N = 150),
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
326
(b) from 500 to 600
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(N
=72), and (c) above 600 (N = 45).The findings from the data analysis were very interesting and provide some insight into which constituency of female test takers was most affected by the bias in the GMAT. The GPA of stu-
dents in the cohort between 400 and 500 (cohort 1) was 3.45 (SD = .30), the GPA
of students in the cohort between 500 and 600 (cohort 2) was 3.54 (SD = .43), and the GPA of students in the cohort
above 600 (cohort 3) was 3.70
(SD
=.34). Using one-tailed t tests across both men and women, as might be expected, we found significant differences in GPA
between cohorts 1 and 2 ( t = 1.67, p <
.05, df = 221), between cohorts 2 and 3
(t = 2.22 p
<
.05, df = 1 16), and between cohorts 1 and 3 (t = 4.76, p < .001, df =193). These finding provided support for the gender-based predictive utility of the
GMAT. We then examined these differ-
ences in GPA within the male and
female gender groups. For men, al- though a significant difference was found for the comparison of the cohorts as a whole, there were no significant dif- ferences in GPA between cohorts 1 and
2 ( t = .29, p = .37, df = 122). However, like the results from the cohorts as a whole, significant differences in GPA
were found between cohorts 2 and 3 ( t =
1.80, p < .05, df = 76) and between cohorts 1 and 3 ( t = 2.75, p < .005, df =
1 lo). For women, significant differences in GPA were found between cohorts 1
and 2 (t = 2.22, p < .05, df =97), cohorts
1 and 3 ( t = 4.40, p < .001, df = 38), and cohorts 2 and 3 ( t = 1.62, p < .05, df =
38). These results generally suggest that there are differences in GPA, for both
men and women, across the three cohort groups. These results were as expected and confirm the usefulness of the GMAT
in predicting performance in graduate business programs.
In our next set of tests, we examined differences in core GPA between male
and female students within cohorts, to uncover the cohort(s) in which the bias in the GMAT was having its greatest
effect. Beginning with the first cohort, we found no significant differences in core GPA between male ( m =3.48, SD =
.30) and female students (m = 3.43, SD
= .30) (t = 1.01, p = .16, df = 148). In addition, there were no significant dif-
ferences in core GPA between male ( m
= 3.51, SD = .42) and female students
(m = 3.60, SD = .45) ( t = 1.02, p = .15,
df = 71) in cohort 2. However, there were significant differences in GPA between men ( m = 3.66, SD = .38) and women ( m = 3.82, SD = .01) ( t = 2.14, p
< .05, df = 40) in the third cohort, with
females outperforming males with simi- lar GMAT scores.
Discussion
Our study results suggest that the
GMAT tends to be a good predictor of
performance in graduate business pro- grams, given the significant differences in GPA across the three GMAT cohorts
as a whole, as well as in GPA across the
three cohorts within both the male and female groups (excepting the case of nonsignificance for men between cohorts
1 and 2). This finding, consistent with previous research indicating that the
GMAT is a strong predictor of academic
performance in graduate business pro- grams, serves to illustrate the generaliz- ability of the results from the current study.
In addition, consistent with the results reported by Hancock (2000), our results also indicate that there is a sig- nificant bias effect in the GMAT.
Specifically, our results indicate that although there were no differences in
GPA across the sample as a whole for
male and female students, male students had significantly higher GMAT scores
than did female students. This finding suggests that although men and women tended to have similar levels of objec- tive success during the course of their
MBA programs, the GMAT scores
reported by the students in this group would have underpredicted the success of the female students.
We reasoned that it would be impor- tant to examine the results from this study in detail to establish where, exactly, the effects of the underpredic- tion in the GMAT were most manifest.
This specificity was important for us to determine which constituency would be most affected by the bias. We found that although there is a bias effect from the GMAT in this sample, the point at
which the bias effect occurs is among students reporting GMAT scores above
600. The bias at the upper-end of per- formance on the GMAT may have been a result of gender-based, differential perceptions of performance-level access issues. Specifically, in a recent national survey of 461 female execu- tives asked to rank 13 strategies for career success, the top-ranking strategy was identified as “consistently exceed- ing performance expectations” (Ran-
gins, Townsend,
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
& Mattis, 1998). Those authors reported that femaleexecutives consistently felt that they needed to work harder and outperform their male counterparts to receive simi- lar levels of recognition. It may be that the top female MBA students in our sample, experiencing similar feelings of gender-based access inequity, were more motivated to work hard than their male counterparts during the course of their MBA programs because they felt the need to outperform the competition to gain similar levels of recognition. The top female students would be cog- nizant of their ability to perform at a top level, based on their high GMAT score, in both graduate business school and in the business world. Female stu- dents in the other cohorts may have felt less confident in their ability to com- pete with men at those levels and there- fore may have been less motivated to exert the extra effort necessary to out- perform men.
If the finding of bias against women among the top performers on the GMAT were shown to generalize to other uni- versities, the implications for admis- sions personnel at top business schools
in the United States could be dramatic.
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Best Graduate Schools for 2002 (U.S. News and World Report, 2001) lists its
ranking of the top
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
25 business schools in the United States, including GMATscores. Average GMAT scores range
from a low of 641 to a high
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
of 730 forthese schools. Clearly, these elite schools are forced to choose among applicants with very high GMAT scores. To the extent that women with slightly lower GMAT scores may be rejected in favor of men with higher scores, any possible bias in the GMAT would have a negative effect on women. In addition, because these are top schools, they would seem more likely to produce future top-level executives than
would lower-ranked graduate business schools. Thus, bias could, as Hancock (2000) suggested, have a negative impact on women who are attempting to break the “glass ceiling” and move into upper management levels in the busi-
ness world.
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Suggestions for Future Research
This study’s results indicate that it is imperative for admissions staffs at col- leges and universities across the country to determine whether there is any bias in the GMAT at their institution, and whether and to what extent they need to make provisions-in the form of indexed performance schemes, for example-to control for this negative influence. Some institutions seeking to contribute to the diversity of their stu- dent body “discount” GMAT scores for international students and members of “protected” groups. This policy may be worth considering for female applicants unfairly affected by unintended biases in the GMAT.
In addition, future researchers should seek to understand how, exactly, female test takers are affected negatively by biased selection tools. This impact may take the form of lowered expectations for success upon matriculation to a par- ticular program (Bandura, 1986) and, as a result, lower levels of both academic and even professional performance (Staj kovic, 1998).
The GMAT is the single best predic- tor of graduate school performance as measured by graduate students’ GPAs. However, given mounting evidence of potential systematic bias against wom- en, validation studies designed to uncover instances in which the test may be a biased predictor would seem worthwhile at all universities offering the MBA and using GMAT scores as admission criteria-expecially because it is possible for business programs to make adjustments that reduce the impact of that bias.
REFERENCES
Alsop, R. (2001, October 16). Business schools expect big jump in applications. The Wall Street Journal, pp. B1, B12.
Baird, L. L. (1975). Comparative prediction of first year graduate and professional school grades in six fields. Educational and Psycho-
logical Measurement, 35, 941-946.
Bandura, A. (1986). Social foundations
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
of thoughtand action: A social cognitive theory. Engle- wood Cliffs, NJ: Prentice Hall.
Bendlow, C. P., & Stanley, J. C. (1980). Sex dif- ferences in mathematical ability: Fact or arti- fact? Science, 210, 1,262-1,264.
Bendlow, C . P., & Stanley, J. C. (1983). Differen- tial course-taking hypothesis revisited. Ameri-
can Educational Research Journal, 20,
469-473.
Bridgeman, B., & Wendler, C. (1991 ). Gender dif- ferences in predictors of college mathematics performance and in college mathematics course grades. Journal of Educational Psychology, 83,
Carver, M. R., Jr., & King, T. E. (1994). An empir- ical investigation of the MBA admission crite- ria for nontraditional programs. Journal of Edu- cation for Business, 69, 95-98.
Dalton, D. R., & Daley, C. M. (1998). Not there yet. Acmss the Board? 35, 1 6 2 0 .
Deckro, R. F., & Wounderberg, H. W. (1977). MBA admission criteria and academic success. Decision Sciences, 8, 765-769.
Dobson, P., Krapljan-Ban, P., & Vielba, C. ( 1 999). An evaluation of the validity and fairness of the graduate management admissions test (GMAT) used for selection in a UK business school. International Journal of Selection and Assess- ment, 7, 196-202.
Fisher, J. B., & Resnick, D. A. (1990). Standard- ized testing and graduate business school admission: A review of issues and an analysis of a Baruch college MBA cohort. College and Universio, 65, 137-148.
Foxall,
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
G. R. (1992). Gender differences in cogni-tive styles of MBA students in three countries. Psychological Reports, 70, 169-1 70.
Graduate Management Admission Council. (2001). Profile of Graduate Management Admissions Test candidates-Five year sum-
mary. Princeton, NJ: Author.
Graham, L. D. (1991). Predicting academic suc- cess of students in a master of business admin- istration program. Educational and Psycholog-
ical Measurement, 51, 721-727.
Hancock, T. (2000). The gender difference: Valid- ity of standardized admission tests in predicting mba performance. Journal of Education for Business, 75, 91-93.
Hilgert, A. (1998). Professional development of women and the executive MBA. Journal of
Management Development, 17, 629-643.
Johnson, T. R., & McLaughlin, S. D. (1993). Declining numbers of female MBAs: An analy- sis from the GMAT registrant survey. Selec- tions, Winter, 22-36.
Joyce, A. (2002, May 7). First MBA lesson, sup- ply and demand. Washington Post. Retrieved July 18, 2002 from http:/www.gmat.com Paolillo, J. G. P. (1982). The predictive validity of
selected admissions variables relative to grade point average earned in a master of business administration program. Educational and Psy- chological Measurement, 42, 1,163-1 ,I 67.
Ragins, B.
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
R., Townsend, B., & Mattis, M. (1998).Gender gap in the executive suite: CEOs and female executives report on breaking the glass ceiling. The Academy of Management Execu- tive, 12, 2842.
Simpson, R. (2000). A voyage of discovery or a fact track to success: Men, women and the MBA. Journal of Management Development. 19,164-782.
Sinclair, A. (1997). The MBA through womens’ eyes: Learning and pedagogy in management 275-284.
July/August 2003 327
education.
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Management Learning, 28, Wilder, G . Z.,zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
& Powell, K.zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(1989).zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Sex di$er- replication study. Educational and Psychologi- 313-330. ences in test performance: A survey of the liter- cal Measurement, 54, 1,022-1,029.V.S. News
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
and World Report. (2001). Best gradu- ature. College Board Report No. 89-3, ETSRR Youngblood, S. A,., & Martin, B.zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
J. (1982). Abil-ate schools for 2002. Washington, DC: Author. No. 89-4. New York: College Entrance Exami- ity testing and graduate admissions decision
Stajkovic, A. D. (1998). Self-efticacy and work- nation Board. process modeling and validation. Educational related performance: A meta analysis. Psycho- Young, J. W. (1994). Differential prediction of and Psychological Measurement, 4 2 .
logical Bulletin, 124, 240-26 1. college grades by gender and by ethnicity: A I , 153-1,162.