Embed Size (px)
Citation preview
Exploring the Use of Credit Scores in Selection Processes: Bewareof Adverse Impact
Sabrina D. Volpone • Scott Tonidandel •
Derek R. Avery • Safiya Castel
Published online: 14 June 2014
� Springer Science+Business Media New York 2014
Abstract
Purpose The use of credit checks or credit scores in
personnel selection has received widespread media atten-
tion of late. Though there is speculation that basing hiring
decisions (even partially) on credit-related variables may
produce or increase adverse impact, virtually no empirical
literature exists to support or refute this claim. The present
study explores the impact of using credit scores, in the
context of a larger selection system, on adverse impact.
Design/Methodology/Approach We conducted Monte
Carlo simulations representing various real-world selection
systems (i.e., multiple hurdle, multiple hurdle with cut-off
score, single hurdle). In addition to applicant credit scores,
each simulation included variables that organizations
commonly use during selection (i.e., educational back-
ground, personality).
Findings Results showed that in a majority of simulated
hiring scenarios, using credit scores (as opposed to a ran-
dom, race-neutral variable) widened the Black-White gap
in hiring, producing more violations of the 4/5ths rule and
statistically significant adverse impact.
Implications These results imply that organizations
should be cautious when using credit scores to evaluate
potential or current employees for jobs.
Originality/Value This is one of the first studies to provide
empirical evidence of a relationship between credit scores in
selection and adverse impact. The use of simulations helps
organizations be proactive in regards to choosing selection
practices. Our results in particular pinpoint the situations
where implementing credit scores as part of a larger selection
process might be most problematic in terms of adverse
impact, thereby providing much needed guidance to those
considering credit scores for their selection processes.
Keywords Credit score � Monte Carlo simulation �Selection systems � Adverse impact � 4/5ths rule �Employment discrimination � Disparate impact
Introduction
Selection decisions are a critical part of organizational
functioning, as few things are as important as choosing the
right employees to hire (Pfeffer 1998). Accordingly, the
tools used in selection are under constant scrutiny, particu-
larly regarding their ability to screen applicants both validly
and fairly (Ababneh et al. 2013; Outtz 2009). Recently, the
use of credit-related variables (e.g., credit checks, credit
scores) in the hiring process has been debated widely in the
popular press (SHRM 2010). Partial interest in this topic
stems from concerns that the inclusion of credit checks and
S. D. Volpone (&)
Anderson School of Management, University of New Mexico,
1 University of New Mexico, Albuquerque, NM 87131-0001,
USA
e-mail: [email protected]
S. Tonidandel
Department of Psychology, PO Box 7061, Davidson,
NC 28035-7061, USA
e-mail: [email protected]
D. R. Avery
Fox School of Business, Temple University, 1801 Liacouras
Walk, Philadelphia, PA 19122, USA
e-mail: [email protected]
S. Castel
Anderson School of Management, University of California,
Los Angeles, Los Angeles, CA 90095, USA
e-mail: [email protected]
123
J Bus Psychol (2015) 30:357–372
DOI 10.1007/s10869-014-9366-5
scores in selection systems may be unfair, in that it may
produce adverse impact for various protected groups of
applicants and employees (Nelson 2010). Adverse impact
results from employing criteria that disproportionately dis-
advantages one group relative to another along a protected
category (e.g., race, sex, national origin; Guion 1998) if the
criterion is not job related (e.g., physical ability). Legally,
concern over adverse impact comes from the Uniform
Guidelines, which state ‘‘a selection rate for any race, sex, or
ethnic groupwhich is less than 4/5ths (or 80 %) of the rate for
the group with the highest rate will generally be regarded by
the Federal enforcement agencies as evidence of adverse
impact’’ (Equal Employment Opportunity Commission
1978, p. 38297).
In addition to the 4/5ths rule, researchers, organizations,
and courts often apply a test to determine if the adverse
impact produced by a selection procedure is statistically
significant (Agresti 1992; Biddle 2006; Roth et al. 2006).
Specifically, most research and practical sources (e.g.,
courts) use Fisher’s Exact Test (e.g., Roth et al. 2006).
With this test, adverse impact usually occurs if the mean
score on the selection procedure differs by two to three
standard deviations for majority and minority subgroups
(Biddle 2005; Siskin and Trippi 2005). A significance test
is typically used in addition to the 4/5ths rule because error
rates (i.e., false positives) in the former are typically lower
than in the latter. Consequently, most investigations into
adverse impact of selection procedures test for violation of
the 4/5ths rule and significance using Fisher’s Exact Test
(Bobko and Roth 2010; Siskin and Trippi 2005).
Ultimately, organizations are at risk if selection proce-
dures are found to produce adverse impact (if the selection
criteria is not job related; Williams et al. 2013). For
example, applicants excluded by the unfair procedure may
file discrimination lawsuits against the company that could
affect company reputation and hurt the bottom line. In fact,
the average cost for a company when an employee brings a
discrimination lawsuit is $75,000 (Chideya 1995) and it is
common for this type of litigation to settle for millions of
dollars (James and Wooten 2006). In addition, the orga-
nization is likely to experience reputation blemishes
(Karpoff and Lott 1993; Wentling and Palma-Rivas 1997)
that can affect customers’ willingness to purchase the
organization’s products (Pruitt and Nethercutt 2002;
Wright et al. 1995). Perceived unfairness of selection
measures also can affect employee morale and productivity
(e.g., decreased organizational commitment, job satisfac-
tion, and job performance), especially for those with pro-
motion and advancement aspirations (Kuhn and Nielsen
2008; Nielsen and Kuhn 2009; Terpstra et al. 1999).
Despite the serious ramifications of using selection cri-
teria that may produce adverse impact (if the selection
criteria is not job related), little research has been done to
assess the prospective effect of using credit checks or credit
scores as part of organizational selection procedures (Equal
Employment Opportunity Commission 2010). Specifically,
in a report to the Equal Employment Opportunity Com-
mission (2010, p. 2), Michael Aamodt, a principle con-
sultant at DCI consulting Group, cautions against the use of
credit-related variables in hiring decisions, citing that
‘‘there have only been five studies that investigated actual
credit history rather than self-reported levels of financial
stress.’’ As such, organizations have nothing but popular
media reports and mere speculation to guide them when
deciding whether to use this tool as part of their selection
systems. This is particularly troubling in light of a recent
finding linking employee credit scores to task and extra
role performance (Bernerth et al. 2012), which is likely to
lead more organizations to integrate credit information into
their hiring processes.
In an attempt to fill this important gap in the organiza-
tional literature, the present study contrasts integrating
simulated credit scores versus a random variable with no
group differences into a hiring system involving other
commonly employed criteria. We do so using Monte Carlo
simulations, a form of statistical analysis imitates various
selection systems and predicts outcomes (e.g., adverse
impact) that may result. The use of simulations in the
present study helps organizations to be proactive when
choosing selection practices, rather than have to react after
an adopted practice fails. These simulations can pinpoint
the situations where implementing credit scores as part of a
larger selection process might be most problematic in terms
of adverse impact, thereby providing much needed guid-
ance to those considering credit scores for their selection
processes. To realistically simulate selection systems, we
choose three (i.e., educational attainment, personality,
credit scores) of the top nine factors that were indicated as
the most important when making a hiring decision (SHRM
2010). In addition to examining the effect of combining
these three criteria on minority selection rates, we also
tested for adverse impact using methods that are used in
most court cases to evaluate the legality of selection tools,
i.e., the 4/5ths rule and a statistical correction to the Fisher
Exact Test (Agresti 1992), Lancaster’s Mid-P Correction,
which was recently introduced into the literature; Biddle
and Morris 2011]. Through this exploration, we hope to
shed some light on the extent to which using credit scores
in selection contributes to inequality.
The remainder of this manuscript is organized as fol-
lows. In the next section, we examine the three variables
we use in our simulation. First, we consider educational
attainment as a variable used to evaluate applicants during
selection scenarios. Second, we discuss the role that per-
sonality, particularly, conscientiousness, plays in hiring
decisions. Third, we investigate organizational reliance on
358 J Bus Psychol (2015) 30:357–372
123
credit scores during the hiring process. Then, we assess
multiple types of selection systems (i.e., single hurdle,
multiple hurdle, and multiple hurdle with cut-off score)
involving simulated educational attainment, conscien-
tiousness, and credit scores to determine if, and to what
extent, racial minority groups are impacted more adversely
when these three variables are utilized during different
types of hiring process. Because there is more information
available about Black-White than Hispanic-White or
Asian-White differences in the variables of interest (edu-
cational attainment, conscientiousness, and credit scores),
we restrict our focus to Black-White differences.
Simulation Model
Education as a Predictor
Educational attainment is often one of the first variables
used to screen applicants, as it judges competency and
potential productivity (Bills 1992; Hughes 2003). The
Society for Human Resource Management (SHRM)
recently reported that of the top nine factors considered
when making a hiring decision, three are directly or indi-
rectly associated with educational attainment. Specifically,
35 % of employers consider education directly and 29 %
consider other certifications. Moreover, 80 % of employers
consider skills as important. Seeing that employers typi-
cally presume that there is a linkage between educational
attainment and the acquisition of skills (Bills 1988), it is
obvious that educational attainment directly or indirectly is
something that organizations consider when hiring (Berry
et al. 2006; Devereux 2002).
Using educational attainment as a selection criterion can
work to the relative disadvantage of racial minorities
(Berry et al. 2006; Hargis et al. 2006). Specifically, a large
body of research exists that establishes long-standing rac-
ioethnic differences in academic achievement and educa-
tional attainment (Darling-Hammond 2004; Roscigno and
Ainsworth-Darnell 1999). For example, high school and
college graduation rates among Blacks are routinely lower
than those for Whites (Board of Governors of the Federal
Reserve System 2007). Further, given that a large number
of racial minority students are not receiving the academic
preparation necessary to attend college, it is not surprising
that Black students are underrepresented at a majority of
colleges and universities (Fine and Davis 2003; Freeman
and Fox 2005) and that the proportion of Black students
completing a bachelor’s degree is much lower than that for
Whites (Adelman 2006; Lotokowski et al. 2004; Thompson
et al. 2006). Consequently, Black students earn fewer
educational degrees and, thus, tend to be less competitive
in job markets wherein degrees serve as prominent human
capital indicators (McDaniel et al. 2011). As such, it is less
likely that Black, as compared to White applicants, will be
hired for positions in which employment is determined
largely by level of educational attainment.
Conscientiousness as a Predictor
Next, we examine the use of personality in hiring deci-
sions. Personality tests are a common component of many
selection systems (Rothstein and Goffin 2006) as their use
has risen steadily over the last 20 years (Tett and Chris-
tiansen 2007). Further, SHRM recently reported that of the
top nine factors that are considered when making a hiring
decision, two are related to applicant personality (e.g., fit).
Of the highly accepted five factors of personality (i.e., the
Big Five), conscientiousness is the aspect commonly con-
sidered most job relevant (Tews et al. 2010; Tracey et al.
2007). Of the Big Five personality factors (i.e., openness,
conscientiousness, extraversion, agreeableness, and neu-
roticism), conscientiousness is the most valid predictor of
performance across a variety of job settings (Dunn et al.
1995; Tews et al. 2010; Tracey et al. 2007). For example,
meta-analytic evidence reports correlations between con-
scientiousness and job performance that range between
0.18 and 0.22 (e.g., Barrick and Mount 1991; Tett et al.
1991). Thus, conscientiousness is likely to be a hiring
requisite for many organizations. As such, conscientious-
ness is considered often during selection processes, as most
organizations desire responsible, hard-working, and high
performing employees (Barrick and Mount 1991; Hurtz
and Donovan 2000; Judge and Ilies 2002).
Evidence from past studies suggests that personality
assessments have less adverse impact on racial minorities
than other commonly used criteria (Avis et al. 2002;
Bradley et al. 2002; Hough et al. 2001; Marcus et al. 2007;
Ones and Anderson 2002; Robertson and Smith 2001). For
example, recent meta-analytic evidence indicated no sig-
nificant differences between Black and Caucasians on the
dimension of conscientiousness, as the Black-White dif-
ference in conscientiousness was small (d = 0.07), non-
significant, and actually slightly favored Blacks (Foldes
et al. 2008). As such, it is not likely that the use of con-
scientiousness in selection systems will contribute to
adverse impact.
Credit Information as a Predictor
Finally, we examine the use of credit-related variables in
hiring decisions. SHRM recently reported that three of the
top nine factors that are considered when making a hiring
decision are directly or indirectly associated with infor-
mation obtained from credit background reports. Specifi-
cally, 47 % of organizations verify applicants’ employment
J Bus Psychol (2015) 30:357–372 359
123
background and references, 44 % look at applicants’
criminal background, and 9 % are interested in applicants’
credit background.
Credit Reports vs. Credit Scores
Many credit-related variables are often used interchange-
ably in popular press. However, there are differences in the
credit-related variables that companies can use during
selection. For example, some firms may use credit reports,
which include verification of educational or professional
history, contacting references, an individual’s criminal
history, and/or an individual’s credit history (SHRM 2010).
Recent reports indicate that 35–42 % of firms use some
form of a credit background check as a tool in their
selection process (Kuhn and Nielsen 2008; SHRM 2010)
though only 13 % of firms conduct credit checks on all
applicants (SHRM 2010).
Instead of using a credit report during the hiring process,
other firms use credit scores, a representative summary of
what is given in a credit report. To elaborate, a credit score
is a number that gives a snapshot of a period of time while
a credit report provides more detailed information regard-
ing applicants’ types of debt (Gurchiek 2011). Academic
research on credit scores is limited but authors have been
able to show that credit scores are significantly related to
organizational citizenship behaviors but not necessarily to
workplace deviance (Bernerth et al. 2012) and that no
significant correlations exist between credit scores and
employees’ job performance, or likelihood of being fired
(Thurm 2011).
In the present study, we choose to use credit scores as
opposed to more general credit background checks for a
number of reasons. First, credit scores are representative of
the information found on a credit background check. Sec-
ond, a credit score is actually tangible in that it is a stan-
dard, formulated number, whereas credit backgrounds are
not as quantifiable. Third, credit background checks
include things like educational attainment. Because we are
measuring such variables directly, we want to be sure that
the constructs in our simulation are distinct. As such, we
use credit scores to measure the use of credit-related
variables in selection procedures.
Credit Scores as Controversial Predictors
There is some controversy over the use of credit scores in
selection decisions due to their perceived fairness. Propo-
nents argue that many organizations use credit information
for legitimate reasons. Typically, employers rely on credit
scores because they are believed to relay information about
employee trustworthiness and responsibility (Brody 2010;
Gallagher 2006; Nielsen and Kuhn 2009). Additional
reasons why organizations use credit information for
employment purposes are (a) the organization is required
by an external agency (e.g., a state government) to do so,
(b) to reduce legal liability if an employee does end up
stealing, (c) research shows a relationship between finan-
cial distress and propensity to steal or accept bribes,
(d) belief that financially stressed employees will perform
poorly, and (e) a bad credit history demonstrates that the
employees are irresponsible and lack conscientiousness
(Equal Employment Opportunity Commission 2010).
Opponents of using credit-related variables in employ-
ment scenarios point out that credit scores do not provide
context (Gurchiek 2011). For example, when people are
out of work (typical in a slow economy), it is easy to fall
behind on bills, and this deters finding a job when decisions
factor in credit histories. Or, even for those with jobs, life
events like divorce or costly medical bills can impact
applicants’ credit reports (Deschenaux 2011; Gurchiek
2011). As such, the use of credit scores in employment
decisions could be detrimental for those that have lost their
jobs or experienced a major life event in the last seven to
10 years. The possible unfair nature surrounding the use of
credit-related variables has led legislatures in four states
(i.e., Hawaii, Oregon, Illinois, Washington) to pass laws
limiting the use of credit reports for employment decisions.
Further, at least 13 other states are considering similar legal
action (Deschenaux 2011). As such, policymakers,
researchers, and consumer advocacy groups alike have
raised concerns surrounding racial bias in credit scoring
(Nelson 2010).
Moreover, we contend that the use of credit scores in
selection may prove highly discriminatory. Racial minori-
ties, who reside in poor communities more often than
Whites, typically have less favorable credit reports due to
socioeconomic disadvantages (Arvey and Renz 1992;
Board of Governors of the Federal Reserve System 2007;
Gallagher 2006). Redlining, the illegal but pervasive
practice of denying financial services to residents of high-
risk, low-income neighborhoods, can limit access to credit-
lending institutions and, thus, negatively impact the credit
reports of individuals in those neighborhoods (Gallagher
2006). Minorities also have significantly less financial
assets and financial knowledge than Whites (Birkenmaier
and Tyuse 2005). As such, these practices especially hurt
racial minorities (Thurm 2011).
Research shows that racial minorities do, in fact, have
significantly lower credit scores than Whites have, even
when income, education, marital status, and residence are
controlled (Gallagher 2006; Smith 2007). Further, courts
have upheld that in certain circumstances credit checks can
violate Title VII of the Civil Rights Act because the dis-
advantaged conditions of racial minorities lead to poor
credit reports that may exclude them from the hiring
360 J Bus Psychol (2015) 30:357–372
123
process (Brody 2010). For example, the Equal Employment
Opportunity Commission sued Kaplan Higher Education
on the basis that using credit-based screening that is not
job-related discriminates against Black employees and
applicants (Thurm 2011). Therefore, based on this evi-
dence involving credit reports, it is likely that the use of
credit scores in selection systems will similarly contribute
to adverse impact.
Research Questions
In sum, although we know that a large Black-White mean
difference exists in credit scores, it remains unclear how
this mean difference will affect adverse impact rates in a
larger selection system. Credit scores are correlated to
other commonly used predictors (e.g., educational attain-
ment, conscientiousness) and these predictors also exhibit
mean differences themselves (e.g., educational attainment).
Thus, the exact nature of the impact of credit scores on
hiring rates in this larger context is unknown. As such, we
present a number of research questions to determine how
the inclusion of credit scores into a selection system using
more established predictors affect the hiring rates of Black
applicants across a multitude of realistic conditions.
To begin, any adverse impact resulting from credit
scores may vary depending on characteristics of the
selection system (e.g., selection ratio, top-down vs. cut-
score, etc.). As such, we pose additional research questions
concerning the sample size, selection ratio, the use of a cut-
score and the type of selection system (i.e., single hurdle,
multiple hurdle). Concerning sample size and selection
ratios, we believe that adverse impact will be higher when
sample sizes are larger and selection ratios are lower. In
previous research that used Monte Carlo simulation and
manipulated sample sizes and selection ratios, Roth et al.
(2006) found that adverse impact increases when sample
sizes increase and selection rates decrease. We expect to
find similar results in the present simulation. Therefore, we
pose the following research questions regarding the rela-
tionships between adverse impact and (a) sample sizes and
(b) selection ratios in our simulation:
Research Question 1: Does adverse impact increase
as a function of sample size?
Research Question 2: Does adverse impact increase
as selection ratios decrease?
Next, we believe that the use of a cut-score will decrease
adverse impact. Previous research supports that straight
top-down selection produces more adverse impact than
alternatives such as cut-scores or banding (Risavy and
Hausdorf 2011). Further, in research conducted by Ployhart
and Holtz (2008), these authors conclude that using band-
ing or score adjustments can reduce adverse impact. While
establishing a cut-score is not banding, it is essentially
doing something similar by saying that applicants above
the cut-score (i.e., those in the band) are indistinguishable
during the selection process. Based on this supporting
research, we pose the following research question regard-
ing the relationships between adverse impact and the use of
a cut-score:
Research Question 3: Is adverse impact lower if a
cut-score is used as compared to when top-down
selection is used?
Another key feature that can affect adverse impact is the
type of selection system: single hurdle versus multiple
hurdle. Prior work has demonstrated that the adverse
impact resulting from the use of a single composite pre-
dictor is a function of both the magnitude of the subgroup
differences on the individual predictors and the pattern of
intercorrelations between the predictors (Sackett and El-
lingson 1997; Schmitt et al. 1997). Similarly, the correla-
tions among predictors also play a critical role in multiple
hurdle selection systems (Finch et al. 2009). Given the
complex interplay between the predictor intercorrelations,
investigating the amount of adverse impact that results
from introducing credit scores into both types of selection
system conditions is important to evaluate. Moreover, the
effects may not be consistent across conditions or selection
systems as the selection ratios being applied at different
stages of the selection process are also vital to consider
(Finch et al. 2009). As such, based on this literature, we
pose the following research question:
Research Question 4: To what extent will adverse
impact occur in both single hurdle and multiple
hurdle selection systems?
Overall, we know that a large Black-White mean dif-
ference exists in credit scores. However, it remains unclear
how this mean difference will affect adverse impact rates in
a larger selection system. Thus, the exact nature of the
impact of credit scores on hiring rates in this larger context
is unknown. As such, we pose the following research
question:
Research Question 5: Does using credit scores in
larger selection systems produce more adverse impact
than a race-neutral random variable?
Methods
Overview
This study used three Monte Carlo data simulations to
explore the impact of using educational attainment,
J Bus Psychol (2015) 30:357–372 361
123
conscientiousness, and credit scores on the selection ratios
of majority (i.e., White) and minority (i.e., Black) job
applicants. To mimic multiple types of real-world selection
situations, each simulation explored a different type of
selection system. The first simulation represented a multi-
ple hurdle selection approach wherein employers first
evaluated applicants on educational attainment (e.g., from
their application or resume). Then, applicants were
screened based on their level on conscientiousness (e.g.,
from a personality test or as assessed from behavioral
interview questions). Finally, applicants were assessed
based on their simulated credit score (e.g., from a back-
ground check). The order of variables followed the order in
which applicants are assessed during the hiring process. To
elaborate, applicants’ resumes are usually evaluated first,
before hiring managers even meet applicants in person.
Specifically, hiring managers usually examine educational
attainment on a resume to determine if the applicant meets
the minimum qualifications for the job (e.g., have they
obtained a bachelor’s degree). After assessing if the
applicant has the minimum qualification for a job from a
resume, the hiring manager determines personality char-
acteristics (e.g., conscientiousness). Typically, hiring
managers would bring the applicant in for further evalua-
tion (e.g., testing or an interview) that tells the company if
the applicant would fit and do the job well. As such, we
included conscientiousness in our simulation second (after
education attainment). Next, evaluating applicants on
credit scores as the third variable considered, as opposed to
the first or second seems most realistic in the sense that
only the applicants that are deemed acceptable in regards to
educational attainment and conscientiousness would be
subjected to a final stage in the hiring process that includes
more fine-tuned assessments (e.g., background checks,
credit score check, calling references). We would suspect
that in most real hiring situations a hiring manager would
be hesitant to pay the money to a credit agency to obtain
credit scores for an entire pool of applicants when they can
wait until the pool is narrowed considerably from the first
two evaluations and pay considerably less (employers pay
per score). Thus, our choice of the ordering of the different
predictors was done as we believe, it most accurately
mimics the ordering that would be used in most hiring
contexts.
In the second simulation, we used a multiple hurdle
selection approach with a cut-score. In this simulation, the
multiple hurdle system described in the first simulation was
replicated for the first two predictors in the process.
However, instead of using top-down selection for the
simulated credit score variable, a cut-off score for the
credit score variable was used. A cut-off score was used
because organizations often apply a minimum threshold
that applicants must possess. Each organization chooses the
cut-off that they want to enforce. For example, a credit
score of 620 typically has been used by mortgage compa-
nies to distinguish prime and subprime loan applicants
(What’s In Your FICO Score 2012). Approximately 20 %
of Americans have a credit score below 620 (What’s In
Your FICO Score 2012). Further, private mortgage com-
panies use a minimum score of 660 to evaluate borrowers.
Between 30 and 40 % of Americans have a credit score
below 660 (What’s In Your FICO Score 2012). These
examples, however, describe loan applicants. In our anal-
yses, if applicants had a simulated credit score above 550,
they were selected. However, if their simulated credit score
was below 550, they were not selected. A cut-off of 550 is
equivalent to the cut-off for a D credit grade. Approxi-
mately 10 % of Americans have a credit score below 550,
or below a D credit grade (What’s In Your FICO Score
2012). Thus, we were extremely lenient in choosing our
cut-off of 550.
Finally, we wished to understand the impact that credit
scores would have in a single hurdle selection process. In
this instance, we formed a single composite by equally
weighting and combining educational attainment and
conscientiousness after they were both standardized. Then,
prospective employees were selected for employment in a
top-down manner according to their scores on this com-
posite as long as their simulated credit score was above the
550 cut-score threshold used previously.
Simulation Procedures
First, a separate population dataset for each subgroup was
simulated consisting of a sample size of N = 10,000 par-
ticipants. These population datasets were generated by
applying a Cholesky decomposition to an input correlation
matrix along with corresponding means and standard
deviations for each subgroup as specified in Table 1. The
values for the population parameters used to generate the
data were obtained from the published literature and are
described in more detail in subsequent sections. For each
iteration of the simulation, we sampled with replacement
from the appropriate population dataset to generate indi-
vidual sample datasets for each subgroup of appropriate
sample size as specified by our design.
The individual sample datasets for each subgroup were
then combined to form a composite sample dataset that
consisted of members of both groups. This composite
dataset was then sorted according to each of the three
predictors being used and individuals were selected in a
top-down manner at each stage of the multiple hurdle
selection process (or the only hurdle, as was the case for
the single hurdle approach). At the conclusion of the final
hurdle (or the only hurdle in the single hurdle approach),
the selection ratios for the majority and minority subgroups
362 J Bus Psychol (2015) 30:357–372
123
were computed. Using this information, we determined
whether the 4/5ths rule was violated and tested whether the
selection ratios were significantly different from one
another. The latter test was performed with the Lancaster’s
Mid-P correction (LMP). The LMP adjustment to Fisher’s
exact test has been shown to outperform the unadjusted
Fisher’s exact test when attempting to identify adverse
impact across a wide range of conditions (Biddle and
Morris 2011). This process was repeated for each of 1,000
iterations for each condition of our study.
Simulation Design
For each of the three simulations, each cell of the design
was evaluated across 1,000 iterations. The design of the
first simulation (i.e., the multiple hurdle approach using
credit scores in a top-down fashion) was a fully crossed 3
(sample size of applicant pool) 9 4 (selection ratio at
hurdle 1) 9 4 (selection ratio at hurdle 2) 9 4 (selection
ratio at hurdle 3) factorial design. Because our goal was to
understand the impact of credit scores as a selection tool,
we needed a standard to serve as a comparative referent.
The standard of comparison we choose was an identical
three hurdle selection system that used a randomly gener-
ated variable with no adverse impact across sub groups as
the third hurdle in the selection process instead of simu-
lated credit scores. While a comparison to a two hurdle
selection system could have been performed, this is not
advisable as the application of an additional hurdle (as was
done when credit scores were used) will apply another
selection ratio to the sample of prospective employees. The
application of another selection ratio can have an effect on
adverse impact rates independent of the type of predictor.
Therefore, we wished to disentangle the effect of an
additional selection ratio being applied from the effect of
the specific predictor being used at that third stage (i.e.,
simulated credit scores). Thus, our approach of comparing
credit scores to a random predictor will provide the fairest
assessment of the impact of credit scores as a selection
tool. We should note that this approach is highly similar to
that employed by Roth et al. (2006) who used a model with
hypothetical variables with no observable group difference
(i.e., d = 0) as their basis for comparison.
The second simulation (i.e., the multiple hurdle
approach with a cut-off score) used an identical fully
crossed 3 (sample size of applicant pool) 9 4 (selection
ratio at hurdle 1) 9 4 (selection ratio at hurdle 2) 9 4
(selection ratio at hurdle 3) factorial design. Since simu-
lated credit scores were used as cut-scores in this simula-
tion, the predictor used as the third hurdle in the selection
process was again a randomly generated variable. Indi-
viduals were selected at this stage of the selection process
based on their score on that randomly generated variable
according to the selection ratio specified by our design so
long as their simulated credit score was above the require
cut-off score of 550. Again, this approach allows us to
compare the use of credit scores as a cut-off score versus
credit scores as a hurdle while ensuring that differences
observed are not due to application of an additional
selection ratio to the pool of applicants.
The design of the final simulation (i.e., the single hurdle
approach) was a fully crossed 3 (sample size of applicant
pool) 9 5 (selection ratio at the single hurdle) factorial
design. Individuals were selected in a top-down fashion
based upon a single composite predictor according to the
selection ratio condition. Adverse impact rates were then
compared across situations where an additional cut-off
score requirement for the simulated credit scores either was
or was not applied.
Sample Size
The three levels of the sample size of the applicant pool
used in this study were 200, 400, and 2,000. These values
were chosen to be similar to other simulation work that has
been done to evaluate the adverse impact of a multiple
hurdle selection system (e.g., Roth et al. 2006). These
sample size values represent the size of the entire applicant
pool. So, the size of the minority and majority subgroups
making up the applicant pool would be less than this.
According to the Bureau of Labor Statistics (2011, 2012,
2013), Blacks comprise approximately 12 % of the U.S.
workforce. Therefore, within each sample size condition,
the data were generated such that the majority group was
88 % of the applicant pool, while the minority group was
12 % of the applicant pool.
Table 1 Simulated population data
Variable M SD 1 2 3 M SD
(1) Educational attainment 2.17 1.60 – -0.06** 0.06** 1.78 1.45
(2) Conscientiousness 48.40 9.90 -0.04** – 0.21** 49.60 8.70
(3) Credit Score 728.53 83.00 0.31** 0.31** – 616.10 112.70
N = 10,000; data for Whites are presented below the diagonal; data for Blacks are presented above the diagonal
* p\ 0.05; ** p\ 0.01
J Bus Psychol (2015) 30:357–372 363
123
Selection Ratios
In the single hurdle simulation, the selection ratio for each
hurdle was varied to be either: 0.1, 0.3, 0.5, 0.7, or 0.9. This
is consistent with the previous simulations on adverse
impact rates (Roth et al. 2006). Our choice of these values
not only mimics prior work on adverse impact allowing for
comparisons to that work, but these values also permit
examination of the entire range of selection rations. In the
multiple hurdle simulations, we had to modify the selection
ratio for each hurdle to be either 0.2, 0.4, 0.6, or 0.8.
Though past simulation work examining adverse impact
rates used selection ratios ranging from 0.1 to 0.9 (Roth
et al. 2006), those simulations consisted of only two hur-
dles. Because we have a three hurdle system, selection
ratios of 0.1 across all three hurdles result in too few hires
across the entire process, especially when sample sizes
were small. Thus, we chose 0.2 as our lowest value for the
selection ratio.
Measures
Predictors
Three hypothetical predictors were generated in this study:
Simulated educational attainment, conscientiousness, and
credit scores. Again, we choose these three selection criteria
because recent evidence has indicated that these are repre-
sentative of the top nine factors that are considered when
making a hiring decision about a candidate (SHRM 2010).
Simulated educational attainment was conceptualized as the
highest level of education completed and was represented
along the following six categories: 1 = high school degree/
GED, 2 = some college, 3 = two-year college degree,
4 = four-year college degree, 5 = some graduate or pro-
fessional education, 6 = graduate degree. We applied six
levels of education to measure this variable because the data
on means and standard deviations for educational attain-
ment, which were obtained from the Bureau of Labor Sta-
tistics (2011), were provided in this way. We felt that
keeping each of the six levels of educational attainment
provided gave us more variance in the construct than would
be possible if we condensed and combined levels (e.g., with
a dummy-coded variable coded as 0 = no college degree,
1 = college degree).
Simulated conscientiousness was conceptualized as
described in the NEO Personality Inventory (i.e., the
degree of organization, persistence, control, and motivation
in goal directed behavior; Costa and McCrae 1992). Data
on means and standard deviations by race for conscien-
tiousness were obtained from the study conducted by
Lockenhoff et al. (2008). These authors summed partici-
pants’ 5-point Likert scale responses to the 8-item subscale
and standardized them as T-scores (M = 50, SD = 10)
using the combined sex norms reported by Costa and
McCrae. Though a recent meta-analysis by Foldes and
colleagues (Foldes et al. 2008) estimated the Black-White
mean differences on conscientiousness, this study used the
Big 5 measure of this personality trait, a different measure
of conscientiousness that what we used in the present study
(i.e., the NEO). As such, we were unable to use this meta-
analysis in the simulation.
Simulated credit score was conceptualized according to
the Fair Isaac Corporation (FICO). FICO scores are the
most widely used type of credit score in the United States
and Canada. FICO scores range between 300 and 850. The
median FICO score for Americans in 2010 was 723
(What’s In Your FICO Score 2012). Data on means and
standard deviations for credit score were obtained from
Bernerth and colleagues (Bernerth 2012; Bernerth et al.
2012). Further, Jeremy Bernerth graciously supplied the
correlations among the predictor variables. To elaborate,
Bernerth and colleagues provided us with means and
standard deviations of credit scores as well as the predictor
intercorrelation matrix that was used to simulate the data.
The Bernerth data (Bernerth 2012; Bernerth et al. 2012)
were collected through employees and students at a uni-
versity in the southern U.S., as well as employees in the
same southern region that were not affiliated with the
university. The participants in their study [N = 142; 61 %
male, 37.8 years old (SD = 12.5)] were diverse with
regard to race (16 % Black, 4 % Asian, 2 % Hispanic, 3 %
other).
Because the Bernerth data were collected in 2010, we
collected data for other variables (i.e., educational attain-
ment, race break downs for labor force participation) for
this year, as well. As reported by the Bureau of Labor
Statistics (2011, 2012, 2013), the statistics associated with
our study (i.e., educational attainment, race break downs
for labor force participation) have stayed consistent since
2010, suggesting that the results from the present study are
both relevant and current.
Outcomes
We begin by considering the relative selection rate of
Black applicants to assess the impact of credit scores on
hiring. Subsequently, we employed two measures to
examine whether the use of credit scores produced higher
levels of adverse impact than when employing a random
variable. The first was the percentage of times within each
condition that the 4/5ths rule was violated. This was cal-
culated by first computing the selection ratio for each
subgroup by dividing the number of individuals in that
subgroup that were hired by the number of applicants in
that subgroup. Then, a ratio was created consisting of the
364 J Bus Psychol (2015) 30:357–372
123
minority selection ratio divided by the majority selection
ratio. If this ratio was less than 0.8, then the 4/5ths rule was
violated within that sample.
In addition to examining the percentage of times, the
selection ratio of minorities relative to the selection ratio
for the majority was in violation of the 4/5ths rule, we also
calculated whether there was a statistically significant
difference in the percentage of minorities hired when
compared to the majority. This was accomplished with a
2-way v2 test when sample sizes were 2,000. When sample
sizes were smaller, we relied on the Fisher Exact Test with
the Lancaster’s Mid-P Correction (Biddle and Morris
2011). The reason for these two different analytical
approaches is that Fisher’s Exact Test becomes suboptimal
when sample sizes become large, but would be more
appropriate in instances when sample sizes are small and
especially so when selection ratios are low. The percentage
of time that each statistic was rejected at the p = 0.05 level
was recorded for each dataset.
Results
Multiple Hurdles
Black Applicant Selection Rates
To explore the impact of using credit scores, we first
examined the selection ratios of Black applicants across the
various conditions of our simulation when credit scores
were or were not used. We performed a factorial analysis
of variance predicting the difference in selection ratio for
Black applicants when simulated credit scores are used in a
top-down fashion as the third hurdle in a multiple hurdle
process as compared to when a race-neutral variable is used
(see Table 2). We restrict our focus to the main effects and
two-way interactions, as the three-way interactions
accounted for less than 0.01 % of the variance in selection
rates. As the table indicates, sample size has no main effect
on the differences in selection ratios across the two
approaches (Blacks are consistently selected 8 % less often
when credit scores are used). In fact, Table 3 provides a
comparison of selection rates for Blacks comparing the use
of simulated credit scores in a top-down fashion as a third
hurdle vs. random variable as third hurdle (column 1),
simulated credit scores as cut-score vs. random variable
(column 2) and simulated credit scores used in a top-down
fashion as a third hurdle vs. cut-scores (column 3), and it
can be seen that there is virtually no effect of sample size.
Next, we examined the relative impact of applicant
selection ratios. For ease of interpretation, Table 4 displays
the effects of employing simulated credit scores at different
selection ratios in a simplified form. The table contains the
difference in selection rates for minorities when using
(a) simulated credit scores in a top-down fashion as
opposed to random variable as a third hurdle (column 1),
(b) simulated credit scores as cut-score vs. random variable
(column 2) and (c) the using simulated credit scores in a
top-down fashion compared to credit scores as a cut-score
(column 3). We show only the selection ratios of 0.2 and
0.8 on the first predictor (i.e., education) so as to make this
table more manageable for the reader. The main finding
illustrated in Table 4 is that credit scores lead to a lower
percentage of minorities being hired both when used in a
top-down fashion and when used as a simulated cut-score
(i.e., all values in table are negative) though the magnitude
of the difference is somewhat less for simulated cut-scores.
When selection ratios are very low (top of table), very few
minorities are being hired regardless of the predictor so the
Table 2 Summary of analysis of variance predicting the difference in
Black selection rates comparing credit scores used in a top-down
fashion to a race-neutral random variable
Source SS df MS F Sig. g2
Corrected model 0.77 56 0.01 280.30 0.00
Intercept 1.12 1 1.12 22795.67 0.00
Sample size (N) 0.00 2 0.00 0.06 0.95 0.00
Selection rate 1 (selr1) 0.30 3 0.10 2007.69 0.00 0.16
Selection rate 2 (selr2) 0.22 3 0.07 1499.11 0.00 0.12
Selection rate 3 (selr3) 0.13 3 0.04 895.77 0.00 0.07
N 9 selr1 0.00 6 0.00 0.02 1.00 0.00
N 9 selr2 0.00 6 0.00 0.11 1.00 0.00
N 9 selr3 0.00 6 0.00 0.31 0.93 0.00
selr1 9 selr2 0.06 9 0.01 128.94 0.00 0.03
selr1 9 selr3 0.04 9 0.00 87.34 0.00 0.02
selr2 9 selr3 0.03 9 0.00 59.96 0.00 0.01
Error 0.01 135 0.00
Total 1.90 192
Corrected total 0.78 191
Table 3 Impact of sample size on Black selection rate in multiple
hurdle system
N Credit scores—
top-down vs.
random
Credit scores—
cut-score vs.
random
Credit scores—
top-down vs. cut-
score
200 -0.08 -0.02 -0.05
400 -0.08 -0.02 -0.05
2,000 -0.08 -0.02 -0.05
Comparisons in each column represent the difference in mean
selection rates for Blacks when using credit scores in a top-down
fashion as a third hurdle vs. random variable as third hurdle (column
1), credit scores as cut-score vs. random variable (column 2) and
credit scores used in a top-down fashion as a third hurdle vs. cut-
scores (column 3); All data is simulated
J Bus Psychol (2015) 30:357–372 365
123
impact of using simulated credit scores compared to not
using simulated credit scores is not substantial (e.g., 1 %
difference). However, as selection ratios increase, the
impact of using simulated credit scores becomes quite
dramatic (e.g., the selection ratio for minorities is as much
as 28 % smaller when simulated credit scores are used in a
top-down fashion versus a predictor with no mean differ-
ence—see last row of Table 4). In short, as the selection
ratio increases, the percentage of minorities being hired
grows at a much faster rate for the random predictor with
no mean difference than the predictor with the mean dif-
ference (i.e., credit scores). For example, when the selec-
tion ratios are 0.8, the selection ratio for minorities is 0.52
when using the race-neutral variable as the third predictor
but is only 0.24 when simulated credit scores are used.
Thus, minorities are over twice as likely to get hired if the
third predictor has no mean difference compared to when
simulated credit scores are used instead. The pattern is the
same for cut-scores, albeit less extreme (0.52 vs. 0.41).
Adverse Impact
First, Research Question 1 asked if adverse impact would
be higher when sample sizes are higher. Though the
minority selection ratio was stable across sample sizes
(Table 2), as expected, the amount of adverse impact
observed increased as a sample size increased for every
selection system simulated (Table 5). The increase in
Table 4 Impact of selection
ratio on Black selection rates in
a multiple hurdle selection
system
Comparisons in each column
represent the difference in
selection rates for Blacks when
using credit scores in a top-
down fashion as a third hurdle
vs. random variable as third
hurdle (column 1), credit scores
as cut-score vs. random variable
(column 2) and credit scores
used in a top-down fashion as a
third hurdle vs. cut-scores
(column 3); All data is
simulated
Selection variable 1 Selection
variable 2
Selection
variable 3
Credit scores—
top-down vs.
random
Credit scores—
cut-score vs.
random
Credit scores—
top-down vs.
cut-score
M
0.2 0.2 0.2 -0.01 0.00 -0.01
0.4 -0.01 0.00 -0.01
0.6 -0.01 0.00 -0.01
0.8 -0.01 0.00 -0.01
0.4 0.2 -0.01 -0.01 -0.01
0.4 -0.02 -0.01 -0.01
0.6 -0.02 -0.01 -0.02
0.8 -0.03 -0.01 -0.02
0.6 0.2 -0.02 -0.01 -0.01
0.4 -0.03 -0.01 -0.02
0.6 -0.04 -0.01 -0.03
0.8 -0.04 -0.01 -0.03
0.8 0.2 -0.02 -0.01 -0.01
0.4 -0.04 -0.01 -0.03
0.6 -0.05 -0.02 -0.04
0.8 -0.05 -0.02 -0.04
0.8 0.2 0.2 -0.02 -0.01 -0.02
0.4 -0.04 -0.01 -0.03
0.6 -0.06 -0.02 -0.04
0.8 -0.07 -0.02 -0.05
0.4 0.2 -0.05 -0.02 -0.03
0.4 -0.10 -0.03 -0.06
0.6 -0.13 -0.04 -0.09
0.8 -0.14 -0.05 -0.09
0.6 0.2 -0.08 -0.03 -0.05
0.4 -0.15 -0.05 -0.10
0.6 -0.20 -0.06 -0.14
0.8 -0.22 -0.08 -0.14
0.8 0.2 -0.10 -0.03 -0.07
0.4 -0.19 -0.05 -0.14
0.6 -0.25 -0.07 -0.19
0.8 -0.28 -0.11 -0.17
366 J Bus Psychol (2015) 30:357–372
123
adverse impact tests of statistical significance is an obvious
outcome of the greater statistical power as sample size
increases. The increase in adverse impact according to the
4/5th rules is perhaps less obvious given the stable minority
selection ratio across sample size. Across all conditions,
minorities are being selected at very low rates relative to non-
minorities. However, at very small sample sizes, the change
in a hire/not hire decision for a single minority may be suf-
ficient to alter the adverse impact conclusion according to the
4/5ths rules. In contrast, violations of the 4/5ths rule are less
susceptible to these chance variations as sample size
increases. Given the consistent pattern of results across all
selection systems, Research Question 1 is confirmed, as
adverse impact is higher when sample sizes are higher.
We now focus our attention on how selection ratios
influence the level of adverse impact when simulated credit
scores are employed relative to when an alternative is used
(i.e., Research Question 2, which asked if adverse impact
would be higher when selection ratios are lower). In short,
it appears that differences in selection ratios affect the level
of adverse impact according to both the 4/5ths rule and the
Fisher Exact Test with the Lancaster Mid-P Correction.
These results confirm Research Question 2, as lower
selection ratios are related to higher levels of adverse
impact. In fact, looking at Table 6, we see that across all
conditions, there was adverse impact 43 % of the time at
baseline (when using a predictor with no adverse impact at
the third hurdle) according to the 4/5ths rule. But, the rate
of adverse impact when using simulated credit scores in a
top-down fashion as the third hurdle was 96 % of the time
(for a difference of 53).
Next, Research Question 3 asked if the use of a cut-
score would produce lower levels of adverse impact (as
compared to not using a cut-score). Employing a cut-score
approach with simulated credit scores produced adverse
impact 68.2 % of the time, a difference of roughly 25 over
the baseline model. As such, our results confirm Research
Question 3 because the use of a cut-score did produce
lower levels of adverse impact as compared to not using a
cut-score. We see a remarkably similar pattern (albeit,
slightly less dramatic) when looking at the statistical sig-
nificance of Black-White differences in selection according
to the LMP. In this case (see Table 6), we see that across
all conditions, there was statistically significant adverse
impact 6.9 % of the time at baseline (when using a pre-
dictor with no adverse impact at the third hurdle). This
difference between adverse impact according to the 4/5ths
rule and the LMP is not surprising given that past research
Table 5 Impact of sample size
on adverse impact rate in
multiple and single hurdle
systems
N % AI 4/5ths rule % AI statistically significant
Multiple
hurdle credit
scores top-
down
Multiple
hurdle credit
scores cut-
score
Single hurdle
credit scores
cut-score
Multiple
hurdle credit
scores top-
down
Multiple
hurdle credit
scores cut-
score
Single hurdle
credit scores
cut-score
200 91.81 60.90 61.88 22.15 4.61 37.72
400 96.78 65.43 72.58 44.30 8.57 51.84
2,000 99.96 78.31 93.22 94.26 43.15 93.92
Table 6 Effects of selection system on illegal Black-White differ-
ences in selection
Selection system % AI M (SD) % Sig. M (SD)
Random 43.43 (19.87) 6.88 (8.40)
Credit scores—top-down 96.18 (5.46) 53.57 (40.97)
Credit scores—cut-score 68.22 (14.37) 18.77 (22.75)
Credit scores—top-down vs.
random
52.75 (20.80) 46.68 (37.59)
Credit scores—cut-score vs. random 24.78 (14.36) 11.89 (17.26)
Table 7 Impact of sample size on Black selection rates in a single
hurdle selection system
N Black selection ratio
200 -0.10
400 -0.10
2,000 -0.10
The selection ratio column compares the difference in selection ratios
between a single hurdle selection system that uses credit scores as a
cut-score to one that does not; All data is simulated
Table 8 Impact of selection ratio on Black selection rates and Black-
White differences in selection
Selection ratio Minority selection ratio % AI % Sig.
M M M
0.1 -0.01 9.00 7.50
0.3 -0.04 33.77 23.90
0.5 -0.10 58.40 42.60
0.7 -0.15 68.77 58.77
0.9 -0.22 80.37 69.03
Each column compares the difference between a single hurdle
selection system that uses credit scores as a cut-score to one that does
not; All data is simulated
J Bus Psychol (2015) 30:357–372 367
123
has consistently found higher rates of false positives
according to the 4/5ths rule (Roth et al. 2006). Compara-
tively speaking, there was significant adverse impact
53.6 % of the time when simulated credit scores were used
in a top-down fashion as the third hurdle and 18.8 % of the
time when simulated credit scores were used as a cut-score.
Thus, it appears that involving simulated credit scores in
selection systems can cause substantially more adverse
impact than the use of a race-neutral alternative.
Single Hurdle
Next, Research Question 4 asks to what extent adverse
impact would occur for both single hurdle and multiple
hurdle selection systems. The single hurdle results essen-
tially mirror those of the multiple hurdle results reported
above. Again, we see that sample size has virtually no
impact on the percentage of minority applicants selected
with roughly 10 % fewer minorities selected irrespective of
the sample size (see Table 7). Furthermore, looking at
selection ratios, we see that as selection ratio increases, a
smaller percentage of minorities get hired when simulated
credit scores are used. This translates into more adverse
impact and more statistically significant group differences
(see Table 8).
Overall, our findings provide support for Research
Question 5 that asked if use of credit scores would lead to
higher levels of adverse impact than the use of a random
variable.
Discussion
The goal of this study was to determine the relative impact
of using credit scores or a race-neutral alternative on racial
differences in hiring outcomes. To examine our research
questions, we conducted three Monte Carlo simulations
that replicated different selection systems typically found
in organizations. Overall, we found support for the notion
that including simulated credit scores as a selection crite-
rion is related to hiring fewer Blacks and that this differ-
ence often resulted in adverse impact. To elaborate, our
results revealed that there were fewer Black applicants
hired and, often, considerably more adverse impact when
using a simulated credit score during selection than using a
random (race neutral) variable. As such, organizations
should exercise caution when using any of the selection
scenarios we considered. Even for situations when simu-
lated credit scores had a seemingly small impact on the
selection ratio for minorities (e.g., using simulated credit
scores as a cut-score in a multiple hurdle process), the rates
of adverse impact according to either the 4/5ths rule or the
LMP were substantial (on the order of 1.5 to almost 3 times
as large—see Table 6).
Implications
Research
The results of this study offer an important contribution to
the growth of the literature on the use of credit-related
variables during selection. Only a handful of studies have
examined the use of credit scores in selection systems
(Kuhn and Nielsen 2008; SHRM 2010). Fewer studies have
empirically demonstrated that the use of credit scores
during selection may result in adverse impact (Nelson
2010). Results from the present study confirm the little
research that exists suggesting that credit scores are detri-
mental for racial minorities when used as part of selection
systems (Birkenmaier and Tyuse 2005; Board of Governors
of the Federal Reserve System 2007; Gallagher 2006).
Further, our results extend the previous research in this area
in that we offer the first known analysis of the use of credit
scores (albeit simulated) in conjunction with multiple
predictors in different types of selection systems. Further,
our results extend the previous research in this area in that
we offer the first known independent applications of the
Lancaster’s Mid-P Correction when using Fisher’s Exact
Tests to test for adverse impact.
Practical
Organizations should be cognizant of the results of using
credit-related variables, especially credit scores, to evaluate
applicants. Specifically, including credit scores as a pre-
dictor, even when utilized during a final stage of the hiring
process, produces adverse impact in a majority of cases.
Further, including credit scores as a predictor (as demon-
strated in the present study using simulated credit scores),
even when done via a pass-fail mentality with a lenient cut-
score, typically produces adverse impact. Using a selection
tool that produces adverse impact can affect organizations
financially, through lawsuits, negative publicity, and loss of
reputation and customers (Chideya 1995; Pruitt and Neth-
ercutt 2002; Wentling and Palma-Rivas 1997).
Finch et al. (2009) discuss selection strategies aimed at
reducing adverse impact. Though their work does not take
into account the impact of credit scores in contributing to
adverse impact, it demonstrates that certain strategies are
much more effective than others when using multistage
selection procedures that may produce adverse impact. For
example, their work suggests that one may be able to
manipulate the selection ratio being applied to various
predictors such that one minimizes adverse impact while
retaining the same overall selection ratio. For example, in
368 J Bus Psychol (2015) 30:357–372
123
our three hurdle system, one could apply the selection ratio
of 0.4 at each stage of the selection system to produce an
overall selection ratio of 0.064. However, same overall
selection ratio can be achieved by applying a selection ratio
of 0.8 to educational attainment, 0.2 to conscientiousness,
and 0.4 to credit scores. Our simulation reveals that, when
using a cut-score, the former set of weights produce sig-
nificant adverse impact twice as often as the latter (19.56
vs. 9.33). Thus, while the use of credit scores will assuredly
produce more adverse impact than a race-neutral variable,
a carefully devised multiple hurdle selection system may
help minimize that effect. Their work, combined with the
results from the present study, can help practitioners avoid
the negative financial outcomes associated with lawsuits
and other negative outcomes associated with adverse
impact. Depending on the reasoning for using credit scores
as a predictor, it seems prudent to consider alternative ways
of assessing the constructs that employers are attempting to
evaluate by including credit scores in the hiring process.
For example, integrity tests can measure applicants’ like-
lihood of stealing without producing adverse impact
(Roberts 2011).
Limitations and Future Research Directions
As with any study, there are limitations that should be
acknowledged. First, our results should be interpreted with
the notion that they are based on simulated data. To mimic
real hiring situations, we tested three types of selection
systems (i.e., multiple hurdle, multiple hurdle with cut-
score, single hurdle), and included a wide range of sample
sizes (i.e., 200, 400, 2000; to simulate various organization
sizes) as well as a wide range of values for each selection
ratio (i.e., ranging from 0.1 to 0.9 in the single hurdle
simulation and from 0.2 to 0.8 in the multiple hurdle
simulations). However, our results only represent the val-
ues that we included in the simulation and not a full range
of possibilities in all types of hiring situations across all
types of organizations.
Second, we only used three predictors in our simulations
(i.e., educational attainment, conscientiousness, credit
score). Though we were careful to choose these specific
three selection criteria because they were representative of
the top factors that are considered when making a hiring
decision about a candidate (SHRM 2010), it can be argued
that hiring decisions may be based on more than three
factors. Further, hiring decisions may be based on different
criteria than the three included in this study. Though we do
not refute those claims, we argue that these predictors are
representative of criteria that many organizations use.
Third, one of the predictors used was simulated credit
scores, not credit background or credit history. It is
important to note that employment credit histories do not
include a credit score and thus it may not be accurate to
generalize findings from research on credit scores to
organizations that use credit history as part of their selec-
tion system. Further, the data used for the credit score
predictor were based upon the results of a single study (i.e.,
Bernerth 2012; Bernerth et al. 2012). As such, when using
the Bernerth et al. (2012) data for the credit score predictor
in our simulation, we inherently acquire the limitations of
their study. Further, the fact that our results rely heavily on
this one paper to generate our simulated data (due to the
lack of primary studies available on this topic) points to an
additional limitation of our study. Hence, due to these
limitations, the generalizability of our results could be
hindered.
Further, generalizability may be limited because our
data are based on employees across a number of jobs. In
some cases, organizations may only require the use of
credit scores during the hiring process for certain jobs (e.g.,
jobs that require financial responsibility). When interpret-
ing the results of the current study, it is important to note
that credit scores are used most frequently when hiring for
jobs with financial responsibility or for senior executive
positions (SHRM 2010). We did not restrict our investi-
gation to these specific jobs. As such, it is possible that our
results may be slightly different than the relationships
found in a sample that specifically investigates applicants
for a job that requires great financially responsibility.
However, we would add that our results would only be
different if the correlations between credit scores and the
other predictors (educational attainment and conscien-
tiousness) were a function of job type. While it is reason-
able to assume that the validities might differ for different
jobs, we are unaware of any reason why one would
anticipate that the correlations among the predictors would
be impacted.
Based on the limitations of the study, we urge future
researchers to replicate and extend the effects presented
here. For example, future research should attempt to rep-
licate our model through the use of non-simulated data.
Further, future research should expand our model by
exploring additional variables not assessed in the present
study (i.e., additional predictor variables). Then, research-
ers might investigate how the use of credit scores in
selection practices differs from the use of credit histories or
credit backgrounds. Specifically, researchers could inves-
tigate if the use of credit histories during hiring decisions
produces the same levels of adverse impact that the use of
credit scores do. It should be noted that while our results
suggest that the use of credit scores produces adverse
impact in a majority of cases for racial minorities, that
results may or may not extend to other minority groups
(women, older employees, etc.). Further, it is possible that
considering multiple minority statuses simultaneously
J Bus Psychol (2015) 30:357–372 369
123
would be an interesting avenue of future research. Spe-
cifically, race and gender may work together to affect the
relationships between credit scores and various hiring
outcomes.
Conclusion
In sum, despite the methodological limitations, there are
several important conclusions that can be drawn from this
study. Specifically, the results suggest that (a) when using
simulated credit scores, fewer Black applicants are hired
across nearly all scenarios compared to when simulated
credit scores are not used, (b) this difference in hiring rates
of Blacks when simulated credit scores are used resulted in
more adverse impact as compared to a random variable
(with no Black-White difference) being used, and
(c) multiple hurdle systems that used a cut-score demon-
strated lower levels of adverse impact as compared with
multiple hurdle systems that used a top-down approach, but
adverse impact rates were still meaningfully larger
regardless of how simulated credit scores were used. As
such, organizations should exercise caution when using
credit scores during the hiring process due to concerns
surrounding adverse impact.
References
Ababneh, K. I., Hackett, R. D., & Schat, A. C. H. (2013). The role of
attributions and fairness in understanding applicant reactions to
selection procedures and decisions. Journal of Business
Psychology.
Adelman, C. (2006). The tool box revisited: Paths to degree
completion from high school through college. Washington,
DC: U.S. Department of Education.
Agresti, A. (1992). A survey of exact inference for contingency
tables. Statistical Science, 7, 131–153.
Arvey, R., & Renz, G. (1992). Fairness in the selection of employees.
Journal of Business Ethics, 11, 331–340.
Avis, J. M., Kudisch, J. D., & Fortunato, V. J. (2002). Examining the
incremental validity and adverse impact of cognitive ability and
conscientiousness on job performance. Journal of Business
Psychology, 17, 87–105.
Barrick, M. R., & Mount, M. K. (1991). The big five personality
dimensions and job performance: A meta-analysis. Personnel
Psychology, 44, 1–26.
Bernerth, J. B. (2012). Demographic variables and credit scores: an
empirical study of a controversial selection tool. International
Journal of Selection and Assessment, 20, 242–246. doi:10.1111/
j.1468-2389.2012.00585.x.
Bernerth, J., Taylor, S., Walker, J., & Whitman, D. (2012). An
empirical investigation of dispositional antecedents and perfor-
mance-related outcomes of credit scores. Journal of Applied
Psychology, 97, 469–478. doi:10.1037/a0026055.
Berry, C. M., Gruys, M. L., & Sackett, P. R. (2006). Educational
attainment as a proxy for cognitive ability in selection: Effects
on levels of cognitive ability and adverse impact. Journal of
Applied Psychology, 91, 696–705.
Biddle, D. A. (2005). Adverse impact and test validation: A
practitioner’s guide to valid and defensible employment testing.
London: Gower.
Biddle, D. A. (2006). Adverse impact and test validation: A
practitioner’s guide to valid and defensible employment testing
(2nd ed.). Burlington, VT: Ashgate Publishing Company.
Biddle, D. A., & Morris, S. B. (2011). Using Lancaster’s mid-p
correction to the Fisher exact test for adverse impact analyses.
Journal of Applied Psychology, 96, 956–965.
Bills, D. (1988). Credentials and capacities: Employers’ perceptions
of the acquisition of skills. The Sociological Quarterly, 29,
439–449.
Bills, D. (1992). The mutability of educational credentials as hiring
criteria: How employers evaluate atypically highly credentialed
job candidates. Work and Occupations, 19, 79–95.
Birkenmaier, J., & Tyuse, S. W. (2005). Affordable financial services
and credit for the poor: The foundation of asset building. Journal
of Community Practice, 13, 69–85.
Board of Governors of the Federal Reserve System (2007, August).
Report to the congress on credit scoring and its effects on the
availability and affordability of credit. Retrieved from http://
www.transunion.com/docs/rev/business/clientSupport/legislative
Update/CreditScoreRpt.pdf.
Bobko, P., & Roth, P. L. (2010). An analysis of two methods for
assessing and indexing adverse impact: A disconnect between
the academic literature and some practice. In J. L. Outtz (Ed.),
Adverse impact: Implications for organizational staffing and
high stakes selection (pp. 29–49). New York: Routledge.
Bradley, J. P., Nicol, A. A. M., Charbonneau, D., & Meyer, J. P.
(2002). Personality correlates of leadership development in
Canadian forces officer candidates. Canadian Journal of
Behavioral Science, 34, 92–103.
Brody, R. G. (2010). Beyond the basic background check: Hiring the
‘‘right’’ employees. Management Research Review, 33, 210.
Bureau of Labor Statistics (2011). Labor force characteristics by Race
and Ethnicity, 2010. Retrieved on December 15, 2013 from
http://www.bls.gov/cps/cpsrace2010.pdf.
Bureau of Labor Statistics (2012). Labor force characteristics by Race
and Ethnicity, 2011. Retrieved on December 15, 2013 from
http://www.bls.gov/cps/cpsrace2011.pdf.
Bureau of Labor Statistics (2013). Labor force characteristics by Race
and Ethnicity, 2012. Retrieved on December 15, 2013 from
http://www.bls.gov/cps/cpsrace2012.pdf.
Chideya, F. (1995). Don’t believe the hype: Fighting cultural misinfor-
mation about African Americans. New York: A Plume Book.
Costa, P. T, Jr, & McCrae, R. R. (1992). NEO PI-R professional
manual. Odessa, FL: Psychological Assessment Resources Inc.
Darling-Hammond, L. (2004). The color line in American education:
Race, resources, and student achievement. Du Bois Review, 11,
213–246.
Deschenaux, J. (2011, February). States seek to limit employment-
related credit checks. Retrieved from www.shrm.org/LegalIs
sues/StateandLocalResources/Pages/CreditChecks.aspx.
Devereux, P. (2002). Occupational upgrading and the business cycle.
Labour, 16, 423–452.
Dunn, W., Mount, M., Barrick, M., & Ones, D. (1995). Relative
importance of personality and general mental ability in manag-
ers’ judgments of applicant qualifications. Journal of Applied
Psychology, 80, 500–509.
Equal Employment Opportunity Commission. (1978). Uniform
Guidelines on employment selection procedures. Federal Reg-
ister, 43, 38289–38309.
Equal Employment Opportunity Commission. (2010, October 10).
EEOC public meeting explores the use of credit histories as
employee selection criteria. Retrieved from http://www.eeoc.
gov/eeoc/newsroom/release/10-20-10b.cfm.
370 J Bus Psychol (2015) 30:357–372
123
Finch, D. M., Edwards, B. D., & Wallace, J. C. (2009). Multistage
selection strategies: Simulating the effects on adverse impact and
expected performance for various predictor combinations. Jour-
nal of Applied Psychology, 94, 318–340.
Fine, J., & Davis, J. (2003). Grade retention and enrollment in post-
secondary education. Journal of School Psychology, 41,
401–411.
Foldes, H. J., Duehr, E. E., & Ones, D. S. (2008). Group differences in
personality: Meta-analyses comparing five U.S. racial groups.
Personnel Psychology, 61, 579–616.
Freeman, C., & Fox, M. A. (2005). Status and trends in the education
of American Indians and Alaska Natives. NCES Publication No.
2005-108. Washington, DC: National Center for Education
Statistics.
Gallagher, K. (2006). Rethinking the fair credit reporting act: When
requesting credit reports for ‘‘employment purposes’’ goes too
far. Iowa Law Review, 91, 1593.
Guion, R. M. (1998). Assessment, measurement, and prediction for
personnel decisions. Mahwah, NJ: Lawrence Erlbaum.
Gurchiek, K. (2011, November). SHRM: Credit checks are legitimate
screening tool. Retrieved from www.shrm.org/about/news/
Pages/LegitimateScreeningTool.aspx.
Hargis, M. B., Baltes, B. B., Friend, Y., & Levi, A. (2006). Race
differences in termination at work: The role of educational
inequality. Journal of Business Psychology, 20, 587–598.
Hough, L., Oswald, F., & Ployhart, R. (2001). Adverse impact and
group differences in constructs, assessment tools, and personnel
selection procedures: Issues and lessons learned. International
Journal of Selection and Assessment, 9, 152–194.
Hughes, R. (2003). Skill or diploma? The potential influence of skill-
based pay systems on sources of skills acquisition and degree
programs. Work Study, 52, 179–183.
Hurtz, G. M., & Donovan, J. J. (2000). Personality and job
performance: The big five revisited. Journal of Applied
Psychology, 85, 869–879. doi:10.1037//0021-9010.85.6.869.
James, E. H., & Wooten, L. P. (2006). Diversity crises: How firms
manage discrimination lawsuits. Academy of Management
Journal, 49, 1103–1118. doi:10.5465/AMJ.2006.23478091.
Judge, T. A., & Ilies, R. (2002). Relationship of personality to
performance motivation: A meta-analytic review. Journal of
Applied Psychology, 87, 797–807.
Karpoff, J., & Lott, J. (1993). The reputational penalty firms bear
from committing criminal fraud. Journal of Law and Economics,
36, 757–802.
Kuhn, K. M., & Nielsen, M. L. (2008). Understanding applicant
reactions to credit checks: Uncertainty, information effects and
individual differences. International Journal of Selection and
Assessment, 16, 307–320. doi:10.1111/j.1468-2389.2008.00437.x.
Lockenhoff, C. E., Terracciano, A., Bienvenu, O., Patriciu, N. S.,
Nestadt, G., McCrae, R. R., et al. (2008). Ethnicity, education, and
the temporal stability of personality traits in the East Baltimore
epidemiologic catchment area study. Journal of Research in
Personality, 42, 577–598. doi:10.1016/j.jrp.2007.09.004.
Lotokowski, V. A., Robbins, S. B., & Noeth, R. J. (2004). The role of
academic and non-academic factors in improving college
retention. Iowa: American College Testing.
Marcus, B., Lee, K., & Ashton, M. C. (2007). Personality dimensions
explaining relationships between integrity tests and counterpro-
ductive behavior: Big five, or one in addition? Personnel
Psychology, 60, 1–34.
McDaniel, A., DiPrete, T., Buchmann, C., & Shwed, U. (2011). The
black gender gap in educational attainment: Historical trends and
racial comparisons. Demography, 48, 889–914.
Nelson, A. A. (2010). Credit scores, race, and residential sorting.
Journal of Policy Analysis and Management, 29, 39–68. doi:10.
1002/pam.20478.
Nielsen, M., & Kuhn, K. (2009). Late payments and Leery applicants:
Credit checks as a selection test. Employee Responsibilities and
Rights Journal, 21, 115–130.
Ones, D. S., & Anderson, M. (2002). Gender and ethnic group
differences on personality scales in selection: Some British data.
Journal of Occupational and Organizational Psychology, 75,
255–276.
Outtz, J. A. (Ed.). (2009). Adverse Impact: Implications for organi-
zational staffing and high stakes selection. New York:
Routledge.
Pfeffer, J. (1998). The human equation: Building profits by putting
people first. Boston: Harvard Business School Press.
Ployhart, R. E., & Holtz, B. C. (2008). The diversity–validity
dilemma: Strategies for reducing racioethnic and sex subgroup
differences and adverse impact in selection. Personnel Psychol-
ogy, 61, 153–172. doi:10.1111/j.1744-6570.2008.00109.x.
Pruitt, S. W., & Nethercutt, L. L. (2002). The Texaco racial
discrimination case and shareholder wealth. Journal of Labor
Research, 23, 685–693.
Risavy, S. D., & Hausdorf, P. A. (2011). Personality testing in
personnel selection: Adverse impact and differential hiring
rates. International Journal of Selection and Assessment, 19,
18–30.
Roberts, B. (2011). Your cheating heart. HR Magazine, 56, 55–60.
Robertson, I., & Smith, M. (2001). Personnel selection. Journal of
Occupational and Organizational Psychology, 74, 441–472.
Roscigno, V., & Ainsworth-Darnell, J. (1999). Race, cultural capital,
and educational resources: Persistent inequalities and achieve-
ment returns. Sociology of Education, 72, 158–178.
Roth, P. L., Bobko, P., & Switzer, F. S. (2006). Modeling the
behavior of the 4/5ths rule for determining adverse impact:
Reasons for caution. Journal of Applied Psychology, 91,
507–522.
Rothstein, M. G., & Goffin, R. D. (2006). The use of personality
measures in personnel selection: What does current research
support? Human Resource Management Review, 16, 155–180.
doi:10.1016/j.hrmr.2006.03.004.
Sackett, P. R., & Ellingson, J. E. (1997). The effects of forming multi-
predictor composites on group differences and adverse impact.
Personnel Psychology, 50, 707–721. doi:10.1111/j.1744-6570.
1997.tb00711.x.
Schmitt, N., Rogers, W., Chan, D., Sheppard, L., & Jennings, D.
(1997). Adverse impact and predictive efficiency of various
predictor combinations. Journal of Applied Psychology, 82,
719–730. doi:10.1037/0021-9010.82.5.719.
Siskin, B. R., & Trippi, J. (2005). Statistical issues in litigation. In F.
J. Landy (Ed.), Employment discrimination litigation: Behav-
ioral, quantitative, and legal perspectives (pp. 132–166). San
Francisco: Jossey-Bass.
Smith, A. (2007, March). EEOC urges caution on unnecessary credit
checks. Retrieved from www.shrm.org/Publications/HRNews/
Pages/CMS_020975.aspx.
Society for Human Resource Management (SHRM) (2010, January
22). Background checking: Conducting credit background
checks. Retrieved from http://www.shrm.org/Research/Survey
Findings/Articles/Pages/BackgroundChecking.aspx.
Terpstra, D., Mohamed, A., & Kethley, R. (1999). An analysis of
federal court cases involving nine selection devices. Interna-
tional Journal of Selection and Assessment, 7, 26–34.
Tett, R. P., & Christiansen, N. D. (2007). Personality tests at the
crossroads: A response to Morgeson, Campion, Dipboye,
Hollenbeck, Murphy, and Schmitt (2007). Personnel Psychol-
ogy, 60, 967–993.
Tett, R. P., Jackson, D. N., & Rothstein, M. (1991). Personality
measures as predictors of job performance: A meta-analytic
review. Personnel Psychology, 44, 703–742.
J Bus Psychol (2015) 30:357–372 371
123
Tews, M., Michel, J., & Lyons, B. (2010). Beyond personality: The
impact of GMA on performance for entry-level service employ-
ees. Journal of Service Management, 21, 344–362.
Thompson, M. S., Gorin, J. S., Obeidat, K., & Chen, Y.-H. (2006).
Understanding differences in postsecondary educational attain-
ment: A comparison of predictive measures for Black and White
students. The Journal of Negro Education, 75, 546–562.
Thurm, S. (2011, October). Next frontier in credit scores: Predicting
personal behavior. Retrieved from online.wsj.com/article/
SB10001424052970203687504576655182086300912.html?mod
=googlenews.
Tracey, J., Sturman, M., & Tews, M. (2007). Ability versus
personality. Cornell Hotel & Restaurant Administration Quar-
terly, 48, 313–322.
Wentling, R. M., & Palma-Rivas, N. (1997). Diversity in the
workforce: A literature review (MDS 934). Berkeley, CA:
National Center for Research in Vocational Education.
What’s In Your FICO Score (2012) Retrieved June 2, 2012, from
http://www.myfico.com/CreditEducation/WhatsInYourScore.aspx.
Williams, K. Z., Schaffer, M. M., & Ellis, L. E. (2013). Legal risk in
selection: An analysis of processes and tools. Journal of
Business and Psychology, 28(4), 401–410.
Wright, P., Ferris, S., Hiller, J., & Kroll, M. (1995). Competitiveness
through management of diversity: Effects on stock price
valuation. The Academy of Management Journal, 38, 272–287.
372 J Bus Psychol (2015) 30:357–372
123
Copyright of Journal of Business & Psychology is the property of Springer Science &Business Media B.V. and its content may not be copied or emailed to multiple sites or postedto a listserv without the copyright holder's express written permission. However, users mayprint, download, or email articles for individual use.
