Content uploaded by Ziniya Zahedi
Author content
All content in this area was uploaded by Ziniya Zahedi on Nov 20, 2018
Content may be subject to copyright.
1 23
Journal of Economics and Finance
ISSN 1055-0925
J Econ Finan
DOI 10.1007/s12197-018-9450-1
The effects of role models on college
graduation rates
James V.Koch & Ziniya Zahedi
1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
Science+Business Media, LLC, part of
Springer Nature. This e-offprint is for personal
use only and shall not be self-archived in
electronic repositories. If you wish to self-
archive your article, please use the accepted
manuscript version for posting on your own
website. You may further deposit the accepted
manuscript version in any repository,
provided it is only made publicly available 12
months after official publication or later and
provided acknowledgement is given to the
original source of publication and a link is
inserted to the published article on Springer's
website. The link must be accompanied by
the following text: "The final publication is
available at link.springer.com”.
The effects of role models on college graduation rates
James V. Koch
1
&Ziniya Zahedi
1
#Springer Science+Business Media, LLC, part of Springer Nature 2018
Abstract
Many believe that the presence of role models positively influences the behavior and
success of students. We test one aspect of this contention by focusing upon the impact
that Black, Asian, Hispanic, White and women role models have upon graduation rates
at 176 four-year, public colleges. We find only limited evidence in favor of role model
hypotheses as they relate to individual institutional graduation rates. A 1% increase in
full-time Black faculty on a campus is associated with a .59% increase in the graduation
rate of Black students on that campus, but we do not find strong results for any other
student group. This suggests that we should be less sweeping and more evidence-based
in our approaches to questions involving campus diversity.
Keywords Graduationrates .Role model hypotheses.Influence of role modelson success.
Universities
JEL Classification I23
1 Introduction
Many individuals believe that the presence of faculty role models not only influences
the academic majors college students choose, but also stimulates their academic
performance (Bayer and Rouse 2016). Decades ago, Verdugo (1995)offeredasuccinct
version of the role model hypothesis through the lens of Hispanic students: BBy role
models is meant having Hispanics on campus who are in positions of status and power
within and outside the institution. The belief is that if Hispanic students are able to see
Hispanics in these kinds of positions, it will not only motivate them to remain in school
Journal of Economics and Finance
https://doi.org/10.1007/s12197-018-9450-1
*James V. Koch
jkoch@odu.edu
Ziniya Zahedi
zzahe001@odu.edu
1
Old Dominion University, Norfolk, VA 23529, USA
Author's personal copy
and achieve academically, but it will also provide them with a group that is a natural
sounding board for the many problems facing Hispanic students.^
Ehrenberg (1995) organized a symposium that focused substantially on role models
and concluded that the evidence in favor of these propositions was mixed. Boulware
(2011) arrived at a similar conclusion: Bthe small body of studies on this topic contains
methodological and conceptual inconsistencies and have not produced any consistent
results or conclusions.^
We offer fresh evidence on the influence of role models in higher education. However, in
contrast to previous studies that have concentrated on the choices of individual students in
specific academic majors, we adopt an alternative approach and focus upon institutional
graduation rates. We address both racial/ethnic characteristics and gender.
Our focus on institutions rather than individual students allows us to sidestep some
of the methodological problems that have plagued previous work, most prominent
among them the self-selection biases that have corroded the ability of researchers to
discern role model effects. To wit, most students have some capacity to choose the
specific courses they will take and the professors who teach those courses. Further, they
can drop courses. This has clouded results reported in most prior studies.
By no means do we provide the last word on the empirical validity of role model
hypotheses. Nevertheless, our work pushes the ball down the field because we focus on
the six-year graduation rates of specific racial groups (and women) and consider the
possibility influences of faculty, student and citizen role models on those graduation rates.
Our empirical work is based upon a diverse sample of 176 four-year public
institutions of higher education. We examine the six-year graduation rate of undergrad-
uate students at each institution and seek to determine if the presence or absence of role
models at each has an impact upon its graduation rates. The results are thought-
provoking. Where Black
1
students are concerned, a 1.0% increase in the percentage
of Black faculty on the typical campus evokes a .59% increase in that institution’s
graduation rate of Black students. However, we find no evidence to support a role
model/support effect for the relative presence of Black students on a campus, or for the
Black citizenry within that state. We present comparable evidence for Asian, Hispanic,
White and female students. In general, we uncover only mixed evidence of the
existence of role model effects.
2 Brief literature review
Until the mid-1990s, virtually all analyses of possible role model effects focused on K-
12 education.
2
Elementary education teachers traditionally have been considered influ-
ential role models for young students. Spurred on by work centered at the School of
Industrial Relations at Cornell University, researchers in the mid-1990s extended the
coverage of the role model hypothesis to college undergraduates by examining the
impact of the gender of college faculty members on students’choice of their major, the
courses they opted to take, whether they completed a course, the grade they received, if
1
We use the term Black because one of our major data sources, the Chronicle of Higher Education,usesthe
term Black rather than African-American to describe student and faculty in its data sets.
2
See Jones and Dindia (2004) for a summary of this evidence.
Journal of Economics and Finance
Author's personal copy
they subsequently took additional courses in this discipline, and subsequent labor
market outcomes (Canes and Rosen 1995; Rothstein 1995;Solnick1995; Dynan and
Rouse 1997; Robst et al. 1998; Robb and Robb 1999: Jensen and Owen 2000;
Ashworth and Evans 2010; Rask and Bailey 2002; Butler and Christensen 2003;
Bettinger and Long 2004,2005; Sonnert et al. 2007; Hoffman and Oreopoulos 2009;
Carrell and West 2010; Carrell et al. 2010).
The racial identities of faculty seldom were considered. Nor did these studies allow
for the possibility that that fellow students might serve as role and support models as
well as faculty. Boulware (2011) summarized these studies by observing that despite
their contributions, the fundamental questions concerning role models remained unre-
solved. For every study that concluded there was some type of role model effect at work
in higher education, there was another that asserted that such an effect was insignificant.
A promising methodological alternative to the previous studies has arisen, however,
that relies upon the use of randomized trials and natural experiments to generate
evidence. Carrell et al. (2010) utilized data from the United States Air Force Academy,
where the custom is to assign students randomly to professors. This allowed Carrell
et al. to bypass most self-selection effects.
3
They found female students earned higher
grades and were more likely to pursue STEM-oriented majors if they had been taught
by female faculty. Fairlie et al. (2014) arrived at a similar conclusion concerning
minority students at a community college. Kofoed and McGovney (2018)examined
randomly assigned mentors at the United States Military Academy and found small, but
significant evidence that those being mentored subsequently were more likely to choose
the military specialty of their mentors.
Porter and Serra (2017) randomly selected beginning economic class sections at
Southern Methodist University and then introduced into those class sections carefully
chosen female role models who were highly knowledgeable and presumably charis-
matic. They found that female students so exposed were more likely to pursue
economics as a major, though male students were not. Left unexamined was the
question of how the same students would have responded to highly knowledgeable,
charismatic male role models.
The randomized trials and natural experiments just cited do not provide evidence
concerning graduation rates. Still, they represent promising ways for researchers to
avoid the self-selection biases and specification problems that are present when indi-
vidual student data are used.
3 Model, data, and empirical results
Assume that for financial and other reasons that institutions wish to maximize their
graduation rates. Plausibly, they might do so by: (1) admitting students with superior
academic qualifications; (2) admitting students who do not face financial problems that
might deter their studies; (3) reducing the net price, after grants, that students must pay
in order to attend; (4) obtaining more general fund tax support from their states so that
3
The random assignment of students to classes eliminates the possibility of students self-selecting particular
class sections and specific instructors; however, this does not prevent students from dropping a course. It may
be, however, that the dropping courses at a military academy occurs less often than on conventional campuses.
Journal of Economics and Finance
Author's personal copy
they have the potential to provide a more resource rich education; and, (5) expending a
higher proportion of their funds on student services designed to coax students to
become more connected to their home campus.
4
Among additional factors that influ-
ence graduation rates, but which institutions have little or no influence over in a typical
six-year time frame, are their missions and locations.
The 176 four-year public institutions in our sample represent a diverse sample
consisting of 71 flagships, 67 regional state colleges and universities, and 38 urban
institutions that are neither flagships nor regionals.
Our six-year graduation rate (SIXYRGRAD) for each institution is the percent of
first-time freshmen who entered an institution in 2008 and subsequently earned a
baccalaureate degree at the same institution by 2014. These graduation rates come
from the U.S. Department of Education’s College Navigator (https://nces.ed.
gov/collegenavigator/?s=all).
To capture student qualifications and the rigors of the admissions process, we utilize
two variables: the percent of freshmen applicants who were admitted (PCTAD) as
reported by the College Navigator, and the Brookings Institution’s SAT/ACT composite
mathematics scale (BROOKMATH).
5
Student financial circumstances are measured in
two ways: the College Navigator’s percent of undergraduates receiving a Pell Grant
(PCTPELL), and the St. Louis Federal Reserve’s(https://fred.stlouisfed.org)reporting
of real median household income in the state in which the institution is located
(REALMEDHHINC).
6
We consider the possible impact of the price of each institution’s undergraduate
education upon graduation rates by utilizing the College Navigator’s average net price
paid annually by an in-state undergraduate student (REALNETPRICE). We take
account of possible effects of state support on the graduation rates of each institution
by means of a Delta Cost Study variable (https://deltacostproject.org/delta-cost-data)
recording each institution’s general fund tax support per 100 full-time equivalent
students (REALSTATESUPPFTE). The extent of support services provided to students
may influence graduation rates and we seek to capture this possibility by means a Delta
Cost Study variable that measures the percent of each institution’s education and
general budget spent on student services (STUDSERVPCTEG).
We utilize several role model variables based on the ethnic and gender composition
of the faculty, the student body, and the state in which the institution resides. The
percentages of full-time faculty at an institution that are Black, Asian, Hispanic, White,
or Female (FACPCTBLACK, FACPCTASIAN, FACPCTHISP, FACPCTWHITE,
FACEPCTFEMALE) were obtained from the Chronicle of Higher Education.
7
The
percentages of the student bodies of an institution that are Black, Asian, Hispanic,
White, or Female (STUDPCTBLACK, STUDPCTASIAN, STUDPCTHISP,
STUDPCTWHITE, STUDPCTFEMALE) come from the College Navigator. Finally,
4
A significant literature exists that suggests the degree of involvement of a student on a campus is positively
correlated with that student’s academic performance, persistence and graduation.
5
See S. Kulkarni and J. Rothwell, BBeyond College Rankings: AValue-Added Approach to Assessing Two-
and Four-Year Schools,^Brookings Foundation (April 29, 2015), www.brookings.edu/research/beyond-
college-rankings-a-value-added-approach-to-assessing-two-and-four-year-schools.
6
All financial variables are valued in terms of 2016 prices.
7
Race, Ethnicity, and Gender of Full-time Faculty at More Than 3700 Institutions.^Chronicle of Higher
Education, www.chronicle.com/interactives/faculty-diversity.
Journal of Economics and Finance
Author's personal copy
the percentage of the state’s population that is Black, Asian, Hispanic, or White
(STATEPCTBLACK, STATEPCTASIAN, STATEPCTHISP, STATEPCTWHITE)
was obtained from the U.S. Census.
8
Finally, the institutional mission and location is captured by several dummy vari-
ables. We identify regional institutions (REGIONAL) by means of a 0,1 dummy
variable where 1 = regional. Urban institutions (URBAN) are similarly represented
with 1 = urban.
9
Flagship institutions are the excluded institutional category.
Our cross-sectional model in Eq. (1) seeks to explain the six-year graduation rate of
student group Bj^at institution Bi.^for each ethnicity or gender (j= Black, Asian,
Hispanic, White, or Female).
SIXYRGRADj
i¼β0þβ1REGIONALiþβ2URBANiþβ3PCTADMi
þβ4BROOKMATHiþβ5PCTPELLiþβ6REALMEDHHINCi
þβ7REALNETPRICEiþβ8REALSTATESUPPFTEiþβ9STUDSERVPCTEGi
þβ10FACPCT j
iþβ11STUDPCT j
iþβ12STATEPCT j
iþεi
ð1Þ
The hypothesized coefficients the βi's, are as follows: β1,β2,β5,β7< 0 and β3,β4,β6,
β8,β9,β10,β11,β12 > 0. The white noise error term is given by εi. Table 1presents the
results from the above cross-section model(s).
Column A of Table 1reports the regression results in which the dependent variable
is the six-year institutional graduation rate for Black students. We utilize White’s(1980)
correction for heteroskedascity.
Two of the three Brole model^coefficients are statistically significant at the .05 level
(two-tailed tests). The coefficient on the FACPCTBLACK variable tells us that a 1 %
increase in the percent of full-time Black faculty on a campus generates a .59% increase
in the graduation rate of Black students on that campus.
10
This is notable support for
the role model hypothesis and suggests that the relative presence of Black faculty
makes a difference in the academic achievement and graduation of Black students.
On the other hand, the coefficient on STUDPCTBLACK not only is negative, but
also is statistically significant at the .05 level. A 1 % increase in the percent of Black
students on campus leads to a .36% decline in Black student graduation rates on that
campus. At first glance, this may seem a perverse result, but could be explained by
freely chosen student social relationships. As the number of students in a specific racial
group on a campus increases from zero, the students in that racial group may volun-
tarily segregate themselves socially and residentially (DeRuy 2016;Anonymous2017).
Such students may seek safe spaces where they need not worry about discrimination.
Or, they may prefer the company of friends who have similar tastes for food and music,
8
Data USA, United States Census, State Population Totals and Components of Change: 2010–2017,
https://www.census.gov/data/tables/2017/demo/popest/state-total.html. The percentage of the state’s
population that is female was not used in modeling female graduation rates.
9
The Office of Management and Budget of the United State Government defines metropolitan and micro-
politan statistical areas, https://www.whitehouse.gov/sites/whitehouse.gov/files/omb/bulletins/2017/b-17-01.
pdf.
10
It is possible that the inclusion of part-time faculty would change our results. However, racial data on part-
time faculty are not widely available on an institutional basis.
Journal of Economics and Finance
Author's personal copy
or perhaps wish to be near fellow students who provide support for their points of view.
Speaking broadly, they seek ease and comfort (Walton and Cohen 2011; Gannon 2018).
In the case of Black students, they often come to their chosen institution with
standardized test scores below the institutional average (Reeves and Halikias 2017).
Further, on average, Black students also emanate from households whose incomes are
below the average of their state and their campus and they are more likely to be Pell
Grant recipients. These are real environmental circumstances that exert downward
pressure on academic performances and plausibly diminish graduation rates if these
students self-segregate.
Table 1 Cross-section regression models
Column A Column B Column C Column D Column E
Black Asian Hispanic White Female
CONSTANT 90.51
(5.79)a
98.91
(6.73)a
93.10
(7.82)a
83.31
(5.10)a
113 .36
(7.63)a
FAC P CT 0.594
(2.79)b
−0.083
(−0.24)
0.223
(0.35)
0.155
(1.55)
0.151
(0.74)
STUDPCT −0.356
(−2.63)b
0.237
(1.65)c
0.0011
(0.01)
0.052
(0.95)
−0.301
(−1.73)c
STATEPCT 0.154
(0.95)
0.047
(0.13)
−0.0278
(−0.22)
−0.011
(−0.14)
REALNETPRICE 0.000178
(0.44)
0.001
(2.60)b
0.00011
(0.35)
0.0003
(1.11)
0.00018
(0.60)
PCTPELL 0.003
(0.02)
0.07
(0.40)
−0.103
(−0.72)
−0.99
(0.64)
0.013
(−1.07)
REALMEDHHINC 0.00021
(1.38)
−0.00024
(−1.57)
0.0001
(1.03)
0.000005
(0.47)
0.000007
(0.65)
BROOKMATH 5.71
(2.23)b
3.72
(1.42)
4.82
(2.16)b
4.99
(2.44)b
4.94
(2.44)b
PCTADMIT −0.662
(−8.28)a
−0.459
(−5.70)a
−0.535
(−8.17)a
−0.524
(−8.33)a
−0.507
(−8.44)a
STUDSERVPCTE&G −0.472
(1.57)
−0.346
(−1.19)
−0.232
(0.97)
−0.087
(−0.39)
−0.204
(0.91)
REALSTATESUPPFTE −0.00062
(−2.51)b
−0.0006
(2.25)b
−0.00036
(−1.69)c
−0.0004
(−2.33)b
0.0005
(2.67)b
REGIONAL −4.099
(−1.20)
−4.69
(−1.40)
−3.92
(−1.42)
−2.64
(−1.04)
−3.81
(−1.49)
URBAN −4.272
(−1.40)
2.66
(0.85)
−3.09
(−1.24)
0.324
(0.14)
−2.42
(−1.06)
R20.472 0.364 0.483 0.515 0.509
Adjusted R20.434 0.317 0.445 0.479 0.476
Overall F-statistic 12.16
[0.00]a
7.74
[0.00]a
12.69
[0.00]a
14.43
[0.00]a
15.48
[0.00]a
Number of observations 176 176 176 176 176
t-statistics are given in parentheses with probability values in brackets. Significance levels denoted as follows:
a(1%), b(5%), and c(10%). Regression models estimated via ordinary least squares with White’s correction
for heteroskedasticity. Bold face type signifies coefficient estimates significant at a least the 10% level
Journal of Economics and Finance
Author's personal copy
State-level role model affects are minimal. The STATEPCTBLACK variable has a
small positive sign, but is not statistically significant.
Three of the variables in the equation—NETREALPRICE, PCTPELL, and
REALMEDHHINC—represent students’ability to pay. None of the coefficients on these
variables is statistically significant, suggesting that cost and family financial circumstances
are not critical to graduation. This may initially seem a puzzling result, but one must
remember that the sample of students in this study represent individuals who already have
chosen their institution and are in attendance. One would expect price differentials to
influence which institution a student attends, but once he/she knows an institution’scosts
and decides to attend it, financial considerations already have been taken into account.
Two of the variables in the equation, BROOKMATH and PCTADMIT, represent the
measured academic abilities of the students on each campus. Few would argue with the
proposition that student who come to a campus with higher academic abilities are more
likely to graduate. BROOKMATH is a Brookings Institution variable that recognizes that
some institutions emphasize the SAT, while others the ACT. BROOKMATH is a combi-
nation index with a mean of 0 and a standard deviation of 1. A positive value indicates that
the average SAT or ACT mathematics score of entering students on a campus is above the
national average, while a negative value indicates a below national average score. For
example, the BROOKMATH datum for Clemson University is 1.44, but is −.28 for South
Dakota’s Northern State University. In any case, both coefficients for the academic ability
variables are statistically significant, with the PCTADMITcoefficient especially so. These
are hardly surprising results. Students who come to college with strong academic abilities
are more likely to graduate (holding other things constant).
A refrain sometimes heard on campuses is that students are more likely to identify with
their institution, be retained, and graduate, if they become involved in student activities
and/or take advantage of student services (Roberts and Styron 2010). Our results provide
no support for this hypothesis. Indeed, the coefficient on the STUDSERVPCTEG variable
is negative and barely misses being statistically significant at the .05 level. Why?
Competition for funds often means that more student affairs spending results in less being
spent on instruction (Desrochers and Hurlburt 2015). It might also be the case that student
activities ultimately are not that influential and for students represent time away from
academic pursuits. This may reduce their chances of graduating.
Another theme oft-sounded on campuses whose state funding has been reduced is
that additional state funding would produce improved academic outcomes. At least
where institutional graduation rates are concerned, we find no evidence of this. The
coefficient on the REALSTATESUPPFTE variable is both negative and statistically
significant. Why? More generous state support may be used to reduce faculty teaching
loads, provide greater research support, and increase administrative overhead—factors
that may change the culture on campus, have little to do with student graduation rates,
and divert attention from undergraduates.
The coefficients for the two dummy variables, REG and URB, take on negative
signs. The excluded dummy variable category is flagship institutions. The negative
signs (neither statistically significant at the .05 level) suggest that holding all other
variables constant, there are academic advantages associated with attending a flagship
institution. This is mild evidence that the same student performs better at a flagship
institution than at a regional institution or urban institution. Of course, it seems likely
that unobserved factors are at work here. Consider, for example, that higher proportions
Journal of Economics and Finance
Author's personal copy
of students at regional and urban institutions hold jobs and have families than at
flagship institutions. In addition, higher percentages of flagship university students live
on campus than hold true at regional and urban institutions and on-campus living often
has been connected to better academic performance (Hanover Research 2014). Thus, it
is possible that what we are observing here is the result of non-academic characteristics
of student bodies.
It is appropriate here to insert a caution: one should remember that the regression
coefficients represent marginal changes from the average values of each variable.
Consequently, one cannot use them to extrapolate the effects of much larger changes
in any variable. Thus, while a 1 % increase in the percent of Black faculty on a campus
may elicit a .59% increase in Black student graduation rates on that campus, it does not
follow that a 20% increase in the percent of Black faculty would cause a 12% increase
in Black student graduation rates.
Column B of Table 1presents a regression for Asian students that is strictly comparable
to that presented in Table 1for Black students. Here, however, we find that there is no role
model impact of Asian faculty on Asian student graduation rates, but there is a statistically
significant impact of the presence of Asian students on Asian student graduation rates.
There is no role model effect coming from the Asian citizenry in a state.
The coefficient on the REALNETPRICE variable in the Asian equation takes a
positive sign. One doubts that Asian students prefer to pay higher prices, but they may
self-select institutions whose costs are higher and financial aid prospects are lower.
Once again, admissions tightness (PCTADMIT) and REALSTATESUPPFTE are sta-
tistically significant. All other coefficients except URB assume the same signs as in the
Black equation, but none is statistically significant. Notably, however, the adjusted
coefficient of determination (R2) for this equation is only .317, substantially below the
comparable statistics for the other student groups. There may be unobserved differences
between Asian students (and their subgroups) and other student groups that we have
not been able to capture in our estimating equation.
Column C of Table 1presents a regression for Hispanic students that is strictly
comparable to those presented for Black and Asian students. Hispanic here represents
those students who identify as Latinos as well as they identify as Hispanic. Notably, none
of the coefficients of the role model variables either is large or statistically significant. All
other coefficients follow the patterns established for Black and Asian students.
Column D of Table 1presents a regression for White students that is strictly
comparable to those presented for Black, Asian and Hispanic students. These results
also follow the patterns noted above.
Column E of Table 1presents a regression for female students that is comparable to
that presented the other student groups. The estimating equation for women students
does not include a PCTSTATEFEMALE variable because there is little variation in it
among the states and hence it has little impact on graduation rates. Interestingly, the
while none of the role model coefficients attains statistical significance where women
students are concerned, the STUDPCTWOMEN variable is negative and of large size.
While the coefficient is not statistically significant at the .05 level, it would be
statistically significant at the .10 level. Given that 57% of college students today are
women (Marcus 2017), this suggests some version of the Law of Diminishing Returns
could be at work with respect to the relative representation of women in co-educational,
public university student bodies. All other coefficients imitate our previous results.
Journal of Economics and Finance
Author's personal copy
In summary, we created 14 test opportunities to demonstrate the presence of role
model effects. In only one of the five cases involving faculty, this being
FACPCTBLACK, did we find a statistically significant coefficient. In only two of
the five cases involving students did we find statistically significant coefficients, and in
one of those situations, involving Black students, the coefficient was negative. None of
the coefficients involving the state representation of a student group was statistically
significant and none came remotely close to being so.
Thus, we discover only limited evidence in favor of role model hypotheses as they
relate to graduation rates. On the other hand, none of our adjusted R2statistics was
greater than .479. This strongly suggests that unobserved student and campus
circumstances play substantial roles in determining campus graduation rates.
4 Concluding remarks
Many believe that role models are critically important influences upon the behavior and
performance of underrepresented and less powerful groups in society.
11
This belief is a
recurrent theme in widely read higher education outlets such as the Chronicle of Higher
Education. As one individual put it from the standpoint of students, BSeeing that there
are people from similar situations who have made it makes all the difference^(Field
2017). Nevertheless, we find only limited evidence of such if graduation rates at public
four-year colleges are the criterion. We hasten to note that graduation rates are only one
possible measure of the effects of diversity and colleges and universities might benefit
from diversity in many other ways.
Many institutions have made major resource commitments to support programs
designed to increase faculty and student diversity and provide role models. In this vein,
the University of Michigan announced an $85 million plan to promote diversity, equity
and inclusion on its Ann Arbor campus (Jesse 2016). Michigan’s program is far-
reaching, and most would regard it as a noble effort to change the environment at that
institution. Whether the program will have any noticeable impact on that institution’s
already admirable graduation rates of 80% for African-American students, 91% rate for
Asian students, and 89% rate for Hispanic students (College Navigator 2018)isunclear.
A cautionary lesson to be drawn from our analysis is that campuses would do well to
be more evidence-based as they discuss the impact of diversity on academic perfor-
mance variables. The influence of diversity may vary in its impact depending upon
which ethnic group is being considered and what aspect of performance is being
examined. Our work suggests the need for additional research in this arena.
References
Anonymous (2017) When students self-segregate. Inside Higher Education (March 24, 2017), www.
insidehighered.com/views/2017/03/24/professor-explores-why-his-students-self-segregate-essay
Ashworth J, Evans JL (2010) Modeling student subject choice at secondary and tertiary level: a cross-section
study. J Econ Educ 32:311–320
11
Myers (2016), for example, asserts with respect to flagship institutions, that students want the faculties of
those institutions to be diverse and that student bodies should reflect the demographics of the areas they serve.
Journal of Economics and Finance
Author's personal copy
Bayer A, Rouse CE (2016) Diversity in the economics profession: a new attack on an old problem. J Econ
Perspect 30:221–242
Bettinger EP, Long BT (2004) Do college instructors matter? The effects of adjuncts and graduate assistants on
students’interests and success, National Bureau of Economic Research Working Paper No. 10370
https://doi.org/10.3386/w10370
Bettinger EP, Long BT (2005) Do faculty serve as role models? The impact of instructor gender on female
students. Am Econ Rev 95:152–157
Boulware J (2011) The role model effect in higher education. Master’s Paper, Department of Economics,
Kansas State University
Butler DM, Christensen R (2003) Mixing and matching: the effect of student performance of teaching
assistants of the same gender. PS: Political Science and Politics 36:781–786
Canes B, Rosen H (1995) Following in her footsteps? faculty gender composition and women’s choices of
college majors. Ind Labor Relat Rev 48:486–504
Carrell SE, West JE (2010) Does professor quality matter? Evidence from random assignment of students to
professors. J Polit Econ 118:409–432
Carrell SE, Page ME, West JE (2010) Sex and science: how professor gender perpetuates the gender gap. Q J
Econ 125:1101–1144
College Navigator (2018) 6-year graduation rate by race/ethnicity for students pursing Bachelor’s degrees.
National Center for Education Statistics, https://nces.ed.gov/collegenavigator/?q=
university+of+michigan&s=all&id=170976 Accessed 1 Mar 2018
DeRuy E (2016) The fine line between space and segregation. Atlantic Magazine (August 17, 2016), www.
theatlantic.com/education/archive/2016/08/finding-the-line-between-safe-space-and-segregation/496289
Desrochers DM, Hurlburt S (2015) Trends in College Spending: 2003–2013, Delta Cost Project, www.
deltacostproject.org/sites/default/files/products/15-4626 Final01 Delta Cost Project College Spending
11131.406.P0.02.001 ....pdf
Dynan KE, Rouse CE (1997) The underrepresentation of women in economics: a study of undergraduate
economics students. J Econ Educ 28:350–368
Ehrenberg RG (1995) Introduction. Ind Labor Relat Rev 48:482–485
Fairlie F, Hoffman F, Oreopoulos P (2014) A community college instructor like me: race and ethnicity
interactions in the classroom. Am Econ Rev 104:2567–2591
Field K (2017) A professor ‘Who looks like them.’Chronicle of Higher Education (July 2, 2017),
www.chronicle.com
Gannon K (2018) The case for inclusive teaching. Chronicle of Higher Education (March 3, 2018),
www.chronicle.com
Hanover Research (2014) Strategies for improving student retention.Hanover Research (September 2014)
Hoffman F, Oreopoulos P (2009) A professor like me: the influence of instructor gender on college
achievement. J Hum Resour 44:479–494
Jensen EJ, Owen AL (2000) Why are women such reluctant economists? Evidence from liberal arts colleges.
Am Econ Rev 90:466–470
Jesse D (2016) University of Michigan to Invest $85 Million in Diversity Programs. Detroit Free Press
(October 6, 2016)
Jones SM, Dindia K (2004) A meta-analytic perspective on sex equity in the classroom. Rev Educ Res 74:
443–471
Kofoed M, McGovney E (2018) The effect of same-gender and same-race models on occupation choice:
evidence from randomly assigned mentors at west point. J Hum Resour (forthcoming)
Marcus J (2017) Why men are the new college minority. The Atlantic (August 8, 2017), www.theatlantic.
com/education/archive/2017/08/why-men-are-the-new-college-minority/536103
Myers B (2016) The flagship diversity divide. Chronicle of Higher Education (January 5, 2016), www.
chronicle.com
Porter C, Serra D (2017) Gender differences in the choice of major: the importance of female role models,
unpublished paper, December 5, 2017
Rask KN, Bailey EM (2002) Are faculty role models? Evidence from major choice in an undergraduate
institution. J Econ Educ 33:99–124
Reeves RV, Halikias D (2017) Race gaps in SAT scores highlight inequality and hinder upward mobility.
Brookings Institution Social Mobility Papers (February 2, 2017), www.brookings.edu/research/race-gaps-
in-sat-scores-highlight-inequality-and-hinder-upward-mobility
Robb RE, Robb AL (1999) Gender and the study of economics: the role of gender of the instructor. J Econ
Educ 30:3–19
Journal of Economics and Finance
Author's personal copy
Roberts J, Styron R (2010) Student satisfaction and persistence: factors vital to student retention. Research in
Higher Education Journal 9:3–6
Robst J, Keil J, Russo D (1998) The effect of gender composition of faculty on student retention. Econ Educ
Rev 17:429–439
Rothstein DS (1995) Do female faculty influence students’educational and labor market attainments? Ind
Labor Relat Rev 48:515–530
Solnick S (1995) Changes in Women’s majors from entrance to graduation at Women’s and coeducational
colleges. Ind Labor Relat Rev 48:505–514
Sonnert G, Fox MF, Adkins M (2007) Undergraduate women in science and engineering: effects of faculty,
fields and institutions over time. Soc Sci Q 88:1334–1356
Verdugo RR (1995) Racial stratification and the use of Hispanic faculty as role models: theory, policy and
practice. J High Educ 66:669–685
Walton GM, Cohen GL (2011) A brief social-belonging intervention improves academic and health outcomes
of minority students. Science 331(2011):1447–1451
White H (1980) A Heteroskedasticity-consistent covariance matrix estimator and a direct test for
Heteroskedasticity. Econometrica 48:817–838
Journal of Economics and Finance
Author's personal copy